|
Post by Nostra1 on Oct 7, 2006 22:41:52 GMT 1
As I recently discovered, building a PC from individual components is not too difficult these days. If you can put together a bookshelf, you can put together a PC. Unless you want to go extreme builds, the only tool you will need are good cross head and flathead screwdrivers. What I found however was that building a PC is intimidating until you actually build one. Even trying to order the correct parts is not easy unless you have PC building experience. Before building my PC, I did not even like opening the side panel thinking I was going to break the computer. So I decided to take what I have learned over the months and try to put together some thoughts regarding how to build a simple PC. As I am right now building my wife's computer, I will eventually use that as a photo guide. However, I am currently modding the case which means it is in pieces and all over the house. I will probably not start the actual build (which only takes a few hours) for awhile. In the meantime, I wanted to go over some of the key points to a PC and hopefully describe certain areas in layman terms. If I make a mistake, please correct me since I am no expert in this area and my experience is new. This thread will grow over time since I cannot put the whole thing together at once. If you were curious in building a computer and found it intimidating, I hope this thread helps. Edit: I will be writing whatever I can about the background of the topic before discussing the issues for the build. I will put a "build flag" in the post so that you can skip the background stuff and go straight to the discussion on the build.
|
|
|
Post by Nostra1 on Oct 7, 2006 22:56:19 GMT 1
First step: Determine your budget. Computers can cost anywhere from $300-400 for a budget bare bones build to several thousand dollars for a top of the line everything beast. Unless you are trying to show off your system and have deep pockets to build your spotlight PC, cost is one of the important factors for most people. So determining how much you can spend is essential when determining what to buy. The basics of any computer fall into the following components. 1. Case 2. CPU 3. Motherboard 4. RAM 5. Hard Disk 6. Power Supply 7. Optical Drive (CD, combo, DVD) 8. Video Processor 9. Sound Processor 10. computer management 11. Operating System Of the above components, 1-6 and 11 are basically essential for running a computer. 7 used to be optional when software came in floppy diskettes but these days a CD read write drive is a minimum requirement. 8 and 9 can often be found built into the motherboard and this can reduce cost but if you want to have any sort of video and audio performance (can we say games?), you may want a decent vid card and sound card. 10. Is more based on how much you want to control your system and is optional. It is hard to build a computer without the above components. The budget comes in determining how expensive components you want to get. Depending on where you live and how you like to shop, you can get computer components from various places. The standard places, often called "brick and mortar" shops are the local retail electronics shops such as Best Buy and Circuit City in the United States and equivalent places in the rest of the world. These places are convenient but often expensive and do not always carry all the parts. The one exception to this in the US is Fry's electronics. If you are lucky to have a Fry's near you, you can often find a large amount of parts at very low prices. They also have many nice sales during the year. Edit: Another shop in the US that has good prices on components is MicroCenter.The more common way to shop for parts is via the Internet. There are now a myriad of shops that sell parts all over the world. I don't have a list of these shops nor do I want to start gathering one. So I have provided a link to a very comprehensive list that someone has already made of shops. forums.extremeoverclocking.com/showthread.php?t=152446I do not know how good these stores are so shop with caution. In the United States the most popular Internet shop to get components is Newegg. Formerly Egghead Software, they reorganized themselves in the mid 1990's and established themselves as the leader in PC component sales. Their prices overall are good and they ship real fast. Their shipping can be expensive though, especially for small parts. Its no fun buying a $1.00 part and pay $6.00 for shipping. Still the sheer number of parts is exhausting and their service is good. Other smaller parts shops often carry hard-to-find items. This is especially true for cooling components. Many good coolers are from small shops that are out of the country (such as Noctua in Austria and Microcool in Italy) so they are often found in the smaller stores. Small stores are also more friendlier than the big stores but sometimes not as well disciplined. In the US, you can check resellerratings.com to see a rating of a store. One useful point for Newegg for any person (whether or not they shop at NewEgg) is their well organized website. If you wanted to look for background information on a specific part, you can go to Newegg's site and find out the background information on the part. There is also a "5 egg" product rating where customers rate the product (be careful of this since not all customers rate things well. I have seen customers give a product a 5 egg even before they even received the product). They also have a good menu system where you can fine tune the search to see what components are available for a given need (ie want to find a pci-e vid card and not a AGP one? That is possible. Want to find a ATI vid card within the pci-e vid cards? That is possible too.). So you may want to use Newegg to see what components are available and then see if equivalent components can be found in the place you shop. Finally, the last way to get parts is to go to a local small computer shop and have them order parts. It is probably more expensive than ordering yourself but you get the benefit of having a pro get it for you (which means if anything goes wrong, you can just take the part back to the shop for a replacement/refund). You can determine how much piece-of-mind you want when purchasing. Build flag:Ok so find out the prices of the components you want to buy. Cases: An ATX case can cost anywhere from $25 - over $300. Typical case price is $50 - $150. CPU: A cheap Celeron can cost as little as $40 and an Core 2 Duo Extreme Edition can cost $1000.00. Most CPU's cost in the $80 - $400 range. Motherboard: Mobos can cost anywhere from $50 for a micro ATX board to over $300 for a nice SLI or Crossfire based mobo. Server motherboards can cost more but I am going to stay just in the standard PC design. RAM: Depending on what you want, RAM can cost around $80 - $200/Gb. If you want 4 Gb of RAM, be ready to pay a lot. You can run Oblivion with 1Gb of RAM so that can be a reference point. Edit (June 2007): With the introduction of 2Gb RAM sticks, RAM has come down in price significantly. RAM prices are as low as $35/Gb now.Hard Disk: Hard Disks cost less per Gb as you go up in size but get more expensive overall the bigger they get. So long as you don't ask for outrageous speeds (like the 10,000 rpm Raptor, or the 15,000 rpm Cheetah) the typical hard drive will cost anywhere from $40 - $300 depending on whether you buy retail or OEM and the size and speed. A 20Gb IDE hard drive purchased OEM can be as low as $40 or so. In contrast a 400Gb+ SATA 3.0 hard drive purchased retail can cost $300 or more. Power Supply: Power supplies can cost anywhere from $30 for a low priced Sparkle brand PSU to over $300 for an Enermax Galaxy 1 kW power supply. People tend to cut costs here when building a computer since many don't see any benefit from getting an expensive power supply. DO NOT SKIMP COST HERE. If you buy a cheap power supply and it blows (they can literally blow up) during a power up, it can take out your cpu, motherboard and RAM with it, costing you way more than the amount that you saved for the power supply. Optical Drives: So long as you don't go into the esoteric models such as the new Blu Ray drives, a typical optical drive can be purchased for $15 - $100. If you have the fortune of buying OEM instead of retail, get this OEM, it will be half the cost. You can also use old drives from older computers you have kicking around if you want. A CD read write OEM is about $15. A DVD +/- RW burner can be purchased for about $30 OEM (and they are often much better quality than what you get in a brand name computer that uses no-name drives). If you have to buy retail, the same drive will cost about $70 or so (though you do get a cardboard box with it ) Vid card: Probably one area that people spend a lot of money on since games these days are so video intensive. A vid card can cost as little as $20 for low end cards to over $500 for a high end Radeon HD2900XT or GeForce 8800GTX. Sound card: Sound cards can cost anywhere from $15 for a low end card to over $200 for a top of the line card. Computer management: Since this is optional and the degree of control equipment you want can range anywhere it will be left off. You can spend as little as $2.00 for an additional 80mm case fan all the way up to $100 for a single Matrix Orbital control panel. Water cooling can cost up to $500 - $1000 if you go all out crazy. Operating System: This is easy. Check out how much Windows XP or Vista home or pro costs from your local shop. In the US, if you are willing to go OEM, XP home can cost only $90.00. XP home retail is about $190. EDIT: Update, now with the Vista package out, you may rather go this route than to load up XP. Note: Some of your old programs may not work with Vista.If you are shopping from the internet, make sure to include shipping cost in your budget (about 5 - 10%). In the US, the money you spend on shipping is about the same as the money you save not paying sales tax (assuming you live in a state where on-line shopping is not taxed). With a budget determined and basic price level you want to stay below for each component, you can then determine what you want to make. An HTPC (home theater PC) will have different requirements and components to a gaming PC which in turn is different to a office PC. So think what sort of PC you want to make as well.
|
|
|
Post by Nostra1 on Oct 10, 2006 15:13:04 GMT 1
Next Step: Choose your Central Processor. Well if you have gone to a store to buy a Windows based desktop PC, you have noticed that the majority of the desktop PCs fall into two basic categories, Intel based systems and AMD based systems. The difference is basically only found in the difference of the CPU. As you know already, the CPU or Central Processor Unit is the basic brain of the computer. It does the majority of arithmetic calculations, deciphers input signals from the various input devices (keyboard, mouse, etc) and keeps the information flowing in an ordered manner. In the past, it was the main part of the computer with the rest of the computer having more a support function. With increasing strength and responsibility given to the GPU and other peripheral devices, the CPU is less dominant in the PC of present. However it still is the foundation component of any computer. A CPU can cost as little as $30 for a low end Sempron to as high as over $1000 for an Extreme Edition Quad core Core architecture chip. In general, Intel is still king of the office with the majority of systems built around the Pentium line. Intel offered a series of services that helped the office work well with business related requirements (LAN, communications, etc). Also the general public regarded "Intel Inside" as the sign of quality. AMD or Advanced Micro Devices in the past was considered a "me too" company that brought out cheaper versions of the Intel chips. However with the development of the Athlon 64, AMD took over domination in two areas, extreme computing and gaming. The Athlon 64 (A64) had some very good thermal management and very strong overclocking capabilities which were valued by the extreme computing crowd. Also the A64 was found to give faster gaming capabilities than the equivalent Pentium chips. This made AMD become the favored processor for home entertainment PCs until just recently. Now with the new Core 2 Duo's out from Intel, they have taken over the lead once again in performance. In response, AMD has drastically decreased their processor price so that one can now build outrageous systems for little money these days. A low end gamer system can cost as little as $350 and a nice system can be made for about $600 - $800. Aside from the processor type there is the performance of the processor. In the past, the performance of the processor could be compared using the processors' frequency. However, with a variety of processors having different relationship between its clock frequency and its performance, one can no longer use the processor frequency to compare performance. For the past three years, the central processors used the Pentium 4 performance as the comparison value. For example, an Athlon 64 often has two values in its product description. One is the actual processor frequency, the other is the performance equivalent Pentium 4 processor speed. An A64 3000+ Venice 1.8Ghz has a processor speed of 1800Mhz or 1.8 Ghz but is equivalent to a 3000 MHz or 3.0 Ghz Pentium 4. Mobile Pentium M's have a performance level of Pentium 4's having twice the process frequency. For example a 1.5 Ghz Pentium M is similar in performance to a 3.0 Ghz Pentium 4. Dual Core processors: With the increased use of multitasking applications, a central processor often has to take the burden of doing more than one application at a time. Similar to you having to wash the dishes at the same time as writing a report, this is not easy for the processor to do at once and performing multiple tasks simultaneously can slow the processor down as compared to performing those tasks separately. Thus dual core processors, which contain more than one processor core in a single chip has now become quite popular. For Intel, they are denoted as Pentium D or Core Duo. For the AMD systems they are denoted as Athlon 64 X2 as well as some of the high end FX chips. Each single chip is actually slower than a single core counterpart (for example an X2 3800+ has slower individual cores than an Athlon 64 3800+). Their performance is measured with the two cores working in tandem. For many applications, the program only uses a single core to run and so under that circumstance a single core system is faster than its corresponding dual core counterpart. Most games for example still fall under single core operations so a dual core processor can be slower than a single core processor. However, when multitasking, the dual core beats the single core easily. Also more and more programs are actually leveraging a dual core design by multitasking within a single application. Thus for systems of the future, the expected architecture is dual core. However, the dual core chip is more expensive than its single core cousin so all this falls into what chip one may want to buy. Edit: In the future, games are also expected to become multi threaded to take advantage of multiple cores. Aside from the the new cpu's being multi cored, the other main reason for multi-threaded games is because the newest generation consoles (PS3, XBox360) are multi cored platforms. With so much overlap between console and PC games, games designers will make the games to meet both the console and PC design requirements.What the heck are these processor names? You often see a name associated with a chip such as a Sempron Manila, a Pentium E2140 Allendale or an Athlon 64 X2 Brisbane. These names are often used to describe a specific design and process of the chip. When manufacturers set out to make a chip, the equipment necessary to make the chip are very expensive and difficult to design. Thus when making a new process architecture, the manufacturers make specific changes in the chip design that can benefit the performance of the chip. They then manufacture a lot of chips from that particular processing system. Then they give a chip process system a name so that customers know which processing system the chip came from. For example when AMD revised the design of their Athlon 64 chip to a 90nm design with significant lower power and better efficiency (25% lower heating) they named these chips, Winchester, Manchester, Venice, San Diego, Windsor and Toledo. Winchester was the original single core A64 with the new 90nm design while Manchester was the original dual core. Of the latter processes(Venice, San Diego, Windsor and Toledo) , the difference between the names were the number of cores (2 for Windsor and Toledo, 1 for Venice and San Diego) and the size of the ram cache (512 kB for the Venice and Windsor (x2), 1024 kB for the San Diego and Toledo (x2)). This way customers can eventually read up on the performance reviews of these chips and know that when they buy a specific chip, they are getting one from that design architecture. The one really great news is that whichever processor manufacturer you decide to go with, you can make a very nice system at quite low cost. Intel or AMD, the winner in the end is the customer. Build Flag:So when choosing a chip, determine what sort of price range you need to stay within. You will probably not buy a $500 CPU for a $800 computer. You probably want to stay below $100 if you plan on a very budget build such as a $500 system. A $1000 system may want to go to a $100 - $200 chip. Again, this is somewhat dependent on your needs. If you do not expect to play games but want to do office work or surf the net, you can reduce your graphics processor cost and increase your central processor cost. First determine if you want to go Intel or AMD. This will determine what you need to get for the rest of your build. Then determine the chip you want to buy within a specific manufacturer. Intel chips: (if you have chosen AMD, skip this section and go to AMD chips). The latest Intel chips prior to the recent introduction of the Core architecture is all based upon the NetBurst architecture. This was first introduced in 2001 but become more evident after 2003 when the P4 came down in cost to replace the P3 that was being phased out. Now all the Netburst Pentium 4 and its offshoots (Pentium D, Xeon) are made with the smaller scale 90nm or 65nm design. The smaller size was supposed to bring in less power consumption but then to promote performance, added so many transistors in each chip that the chip became literally a space heater for a home. The high end Pentium EE Pressler chips can consume 130W of power and much of this goes into heat. This resulted in very esoteric cooling methods such as water cooling and phase change cooling to become more mainstream. The basic CPU in the Netburst line are the Celeron D, The Pentium 4, and the Pentium D. The Xeon line was mostly set up for server application and Pentium M was for laptop architecture. Some people went out to use Pentium M's for desktops because it generated less heat than its desktop cousin. Celeron D: Basically a Pentium 4 with a lower level 2 cache. Originally they may have been lower quality Pentium chips but as manufacturing got better, they just made these chips lower performance. For a budget build they can be good. The standard name of the chip these days is Celeron D "Prescott". This chip gives a 256kB level 2 ramcache. Core speeds range from 2.5 to 3.3 Ghz. Note: some chips do not support 64 bit so will not be usable with Vista. Check to make sure it is 64 bit supported. To do this check the socket number. Socket 478 is older and will not support 64. Socket LGA 775 is newer and will support 64. Pentium 4: The basic Pentium chip over the last couple of years. The significant difference between the PIII and the PIV is the processor speed. The recent systems can go anywhere from 2.6 to 3.8 Ghz. Most are 64 bit ready but there are a few that are not. Again check to make sure the chip you are buying supports 64 bit. Avoid Socket 478 if possible. The common process names are Pentium 4 "Prescott" and "Cedar Mill". Prescott has a 1MB level 2 cache and is 90 nm design while the Cedar Mill has a 2 MB cache and is 65 nm design. Cedar Mill gives better performance and runs cooler but is more expensive. Both are single core processors. Pentium D: The dual core version of the Pentium 4. There are basically two computer cores in tandem within each chip. The newest single cores are often said to be just dual cores with one core made disfunctional. As discussed above, dual cores perform multitasking better and thus for the office applications, are really nice. For games, the advantage is not yet appreciated yet though future games will probably exploit the dual core architecture. The two product names for Pentium D's are Smithfield and Pressler. You can think of Smithfields as the dual core model of the Prescott while the Pressler is the dual core model of the Cedar Mill. As such the Pressler is the higher performance chip. The processor goes from 2.6 - 3.8 Ghz. Edit: Intel has recently brought out Pentium chips using the Core design (given below) instead of the older Netburst design. They are dsignated as Pentium E and the first two chips are E2140 and E2160. They have better performance than their Pentium D cousins and run much cooler. With the introduction of the Pentium E's, Intel will be discontinuing their Netburst archtecture for the new Core design.Core 2 Duo: On July 2006, Intel has introduced its newest architecture processor called Core 2 Duo. With this chip, Intel has went away from their Netburst design. They also went away from their standard Pentium naming system. This chip is dual core from the get go. It is said to have 40% better performance at 40% less power consumption than its Pentium counterpart. The preliminary performance results seem impressive with Core 2 Duo beating not only its own Pentium D line but also AMD's Athlon 64 FX series. The Core 2 Duo chip goes by the name Conroe and initial products have cpu frequencies of 2.4 and 2.6Ghz and the extreme edition can go up to 2.93 Ghz. These frequencies seem low but again, frequencies do not mean much these days, as they can outperform the best of the Pentium chips. With such performance the power requirement is only around 80W and the chip runs at around 25 - 40C depending on the load. As a side note, the mobile version of Core 2 Duo has also been released and goes by the name Merom. Laptops will soon appear with these cores. Edit: Update: Intel has recently also added the Allendale and Kentsfield chips to the Core 2 Duo lineup. Allendale is similar to the Conroe but has a 800Mhz front side bus to Conroe's 1066 Mhz. Kentsfield is not technically a Core 2 Duo but a Core Quad and is the first line of quad cores to come out. With Intel's campaign to promote their flagship product, the Core architecture is now the most common CPU available.Summary: At this time it is hard to find any reason not to choose Core 2 Duo as the CPU. It is blazing fast, runs very cool, overclocks like crazy and is cost effective. The lowest priced Conroe chip is around $180 or so at Newegg. The only reason one may choose a different chip is if $180 is too high (which it can be for a budget build). If so then the other choice is Pentium D Smithfield. Get one with a dual core for only $80 range at Newegg. Edit: Update: In April 2007, Intel is expected to lower the price of their Core line. At that point the low end Allendale and possibly Conroe chips may fall to the lower $100 price range. In addition, the new Pentium E's are now out for anyone that wants a dual core Intel Core chip below $100.Conroe base cost chip; www.newegg.com/Product/Product.asp?Item=N82E16819115005Pentium D base cost chip: www.newegg.com/Product/Product.asp?Item=N82E16819116001The Pentium D is said to overclock very well and some people have taken this to near 4 Ghz levels (albeit with extreme cooling). Overclocking to about 3.3 Ghz seems routine. So this is a nice stable chip. If you need to go even lower in budget, get a Celeron D for like $40 range. It has the ability to upgrade to Pentium D and Core 2 Duo later. www.newegg.com/Product/Product.asp?Item=N82E16819112207If you want better performance than the two chips above, hey just ride up the Conroe ladder. The extreme edition or the quad core chips cost around $1000 and should satisfy anyone's needs. The thing is if you are going for a lower cost chip, make sure you can upgrade to Core architecture later. Both the Pentium D and Celeron D will do so as long as it is a LGA 775 socket type CPU. However, just having LGA 775 socket is not good enough. Conroe only works on the latest chipsets from Intel, ATI and NVidia. So when choosing a motherboard, choose a socket LGA 775 mobo that states that it can accept Core 2 Duo (Conroe) as a chip. You can use Newegg as a reference point to find out which motherboards runs Pentium D, Celeron D and Core 2 Duo. From the page below choose CPU type as "Core 2 Duo...". Typical chipsets that can transition from Netburst to Core is Intel's P965 chipset for example and NVidia's NForce 6 series chipset. www.newegg.com/Product/ProductList.asp?Submit=ENE&N=2010200280(go to next section for AMD lineup.
|
|
|
Post by Nostra1 on Oct 19, 2006 4:41:40 GMT 1
AMD chips: (For people who chose Intel chips, this section can be skipped)
Advanced Micro Devices has historically produced central processors that followed behind Intel. In the past, the thought was that if you did not have enough money to buy Intel Inside, you bought (somewhat grudgingly) an AMD clone. These days, there are many people who are die hard AMD fans and they have earned their position as being a competitor of Intel rather than a clone. Much of this came with the development of the Athlon 64.
The Athlon 64 is AMD's eighth generation processor. It is the successor to the Athlon XP and as its name indicates, it supports 64 bit architecture (AMD's Opteron was the first to support 64 bit but it was geared toward servers and thus the Athlon was the first 64 bit to hit the general public). Of course at the time, there was very little 64 bit software for the core to support. However, the fact that AMD came out with the 64 bit core gave it a position of being a technology leader. Because this is the eight generation chip, it is often called the K8 architecture (K standing for Kryptonite...the one material that could take down Superman (implying Intel)).
Aside from the 64 bit support, AMD's other key feature of the Athlon 64 was its Hypertransport (not to be confused with Intel's Hyperthreading technology). In the K8 architecture, AMD got rid of the front side bus by bringing in the memory management into the CPU itself. In the motherboard, the chipset that controls the motherboard is divided into the Northbridge and Southbridge. The Northbridge controlled the communication of the CPU with the RAM memory. The Northbridge had its own speed, separate from the CPU and thus could often become the bottleneck of the communication. By incorporating the memory management into the CPU itself, the K8 chips could move the memory management control speed at the same rate as the chip itself. In other words, the K8 architecture got rid of the front side bus (though people still call the memory management this by habit). This made the chip more efficient in its communication and this is one reason why the K8 chips could outperform the X86 Intel chips even at a slower chip frequency. To make sure people understood this performance the K8 chips had two speed ratings; one was the actual speed of the CPU core processor and the other was the speed of the Intel Pentium 4 chip having similar performance.
The AMD line presently falls into four separate processors; the Sempron 64, the Athlon 64, the Opteron, and the X2. The FX which is AMD's counterpart to Intel's Extreme Edition is the notation they give to their highest performing chip at the time. The socket types for the chips are socket 754, socket 939 and socket AM2. Socket 939 and AM2 are the newest and one should only consider these socket types when building a system at this time.
Finally, though there is all the hype of Core 2 Duo with it's impressive performance, the K8 series chips earned their respect by providing the best gaming performance for many years. Its game performance is very good. And to be honest, the whole reason we are on this wonderful forum is because of gaming. It also runs very cool compared to Intel's Netburst chips such as the P4, Pentium D and Celeron D. Because it runs cool it also was notoriously good at overclocking. With the recent price drops, you can build an incredible system at very low prices. An Athlon 64 3000+ (which is in the "recommended CPU performance to run Oblivion" level) can be purchased for as low as $55 these days. One can then immediately turn around and overclock it to 3500+ performance without breaking a sweat (stock voltage, stock heat sink). Its definitely a good series of processors to build a game machine on.
Sempron 64: As the Celeron is to the Pentium, so is the Sempron to the Athlon. It was originally the low performance chips from the Athlon batch but as manufacturing became better, it became the Athlon with a lower level 2 cache. It has 128 kb - 256Kb of cache compared to the 512Kb - 1Mb of cache of the Athlon. As its speed was comparable, it was a nice budget build for the budget gamer. As with the Athlon, it ran games well. The Sempron usually has the name "Manilla" (128kb)or "Palermo" (256kb). The typical speed is around 1.6 Ghz (2800+) to 2.0 Ghz (3600+). Sempron only comes in socket 754 and AM2. Semprons are only available in single core.
Athlon 64: AMD's best known chip. Athlon 64 are single core chips and come in socket 754, 939 and AM2. The typical name of thie chips are "Venice" that has 512kb of level 2 cache and "San Diego" which has 1 Mb of level 2 cache. Sometimes you may find the A64 in the older "Clawhammer" name which is the same as the "San Diego" but instead of the 90nm design is 130nm which means it uses more power and runs hotter. Stick with the San Diego if possible. The AM2 Socket goes by the name "Orleans" which is basically the same as the "San Diego". Typical speeds range from 1.8Ghz (3000+) to 2.6Ghz (4000+).
X2: This is the dual core version of the Athlon 64. This is only available for socket 939 and AM2. The processor names go by either "Manchester" (dual core Venice) or Toledo (dual core San Diego). For socket AM2, the processor name is Windsor (dual core Orleans). Typical speeds like the Athlon are 1.8Ghz (3800+) to 2.6 Ghz (5000+). Edit: Update: AMD has now introduced their 65nm Brisbane line of processors for socket AM2. These are smaller and so should run cooler than the 90nm Windsor line.
Opteron: This is AMD's CPU designed for servers. However, unlike Intel's Xenon which only fits on a different socket, the Opteron comes in Socket 939, 940 and AM2. As these chips had excellent overclocking ability and good performance, many people chose the Opteron chip over the Athlon. Opterons came in both single and dual core and had 1Mb level 2 cache per core. They are typically more expensive than their Athlon counterpart. Opterons come in a zillion names such as Denmark, Venus, Troy, Sledgehammer, Italy, etc. I don't know the differences between them. These processors are not used by brand name PC's but on professional servers. By putting on your PC it can provide the added performance boost for any do it yourself builder.
FX: This designation is the name given to AMD's most powerful Athlon chips (similar to Intel's Extreme Edition). As faster chips come out, they get higher names. The FX chips often found these days are FX-55, 57, 60 and 62. FX 55 and 57 are high end single core chips and FX 60 and 62 are dual core.
Summary: If you are starting to build a PC from scratch and do not need to use old parts from other computers or have other people using older sockets such as 939 in your family, then one should consider the AM2 socket. AM2 came out this year and though there is little change from the socket 939, it is said to be ready for when AMD comes out with their next generation processor. You can now find many motherboards and DDR2 RAM has come down in price so AM2 is no more expensive than 939. Edit: Update: Well it seems that with the recent price drops, AMD has decided to reduce the AM2 processors below their 939 counterparts. This may be one more incentive to go to AM2 if you are making a brand new build vs upgrading an old one.
Also, there is little reason to go to Sempron in my opinion these days. As nice as the Sempron is, the recent price drop of the Athlon as has made the cost advantage minimal. Might as well start with the Athlon. As for the argument of whether one wants to get dual core or single, this is more to your preference. If you multitask, dual core is nicer. If you only want to play present games, then single core may give you faster performance. If you need to remain below $100 then single core is all that you can pay. Once again, its up to your preference in what you want to get. With AM2, you have many choices to choose from. Edit: Update: With the recent decrease in price of AMD processors, the price of the low end Athlon X2 are now below the $100 price range in the United States.
(next section, motherboard)
|
|
|
Post by Nostra1 on Oct 25, 2006 1:16:43 GMT 1
Motherboard:
If there was any part of the PC that discouraged one from touching the PC, let alone build it, it's the motherboard. With its exposed circuits, numerous connections and the bunch of unrecognizable parts, it's the one part of a PC that makes people scratch their heads and then decide to buy another system instead of upgrade. However, for an end builder, there are a few key points that are important and so long as those points are considered, a motherboard becomes not much more difficult to shop for than a video card or a central processor.
If the Central Processor Unit is the brain of the PC, the motherboard is the nervous system. It controls the communication to and from the various components in the PC including the CPU, the RAM, the GPU, the hard drive, the optical drive, the monitor, speakers, mouse, etc. As such, it needs the hardware and electronics to do so.
The heart of the motherboard is the chipset. It contains the firmware and circuits to manage the communication between the CPU to the other components. In the Intel based systems, the chipset is divided into the Northbridge and the Southbridge. The Northbridge manages the communication between the CPU, the RAM and the Video Card. The Southbridge manages the communication between the CPU, the optical, hard drives as well as the PCI peripherals such as the sound card. For AMD based systems the RAM management is done within the Hypertransport of the CPU so there is only one chip in the chipset that controls the communication to the other components. Still people often call the chipset the Northbridge even in the AMD boards.
The manufacturer of the chipset for Intel based motherboards is often Intel itself. However, other manufacturers such as N'Vidia and ATI also manufacture popular chipsets for the Intel based systems. And with SLI and Crossfire being so popular for graphics, one often will choose the N'Vidia or ATI option over Intel. Conversely, some people often feel more comfortable purchasing the chipset from Intel. For AMD, they do not manufacture the chipset, so the most popular brands on AMD based motherboards are N'Vidia and ATI. Other manufacturers include VIA and ULI. They are often a bit less expensive and can be a good cost alternative. Edit: With the recent merging of AMD and ATI, the ATI chipsets are now labelled AMD. Beyond this recent brand change, there is no significant difference between the old ATI chipset series and the new AMD ones.
Along with the chipset comes the firmware that interfaces the chip with the user. This is called the BIOS or Basic Input Output System. The BIOS does the initial startup and tests of the motherboards to determine its configuration and the type of components connected to it. Once it performs these tests and setup, it hands the control of the system over to the operating system contained in the master hard drive. The BIOS is your link to controlling the basic functions and setup of your PC. For brand name computers the BIOS is purposely made to only give limited control as computer manufacturers do not want users pushing components they should not. However the BIOS from a do it yourself motherboard is much more functional. There may be the possibility for causing problems but in most cases it also provides the controls that can help you configure, control and diagnose the system.
The BIOS is usually contained in a ROM chip on your motherboard. It can be replaced but does require one to do some specific tasks to do so (often described as flashing a new BIOS).
Aside from the chipset and BIOS, there are the various connections on the board and depending on the capabilities of the chipset, one has a variety of combinations of connections available on different motherboards.
One connection is the connection to the Hard Disk. Traditionally the hard disk was connected to the motherboard via the ATA or Advanced Technology Attachment connection. It is also often called the IDE port or Integrated Drive Electronics port. Basically IDE describes the physical port and the ATA describes the port's specifications but for most part one can think of ATA and IDE to be the same. If one opened an old PC, one saw a bunch of flat ribbon wires connecting to the hard drives (not to be confused with the smaller one connecting to the floppy drive). The connection on the motherboard that connected to this flat ribbon was the ATA or IDE port. Even today, IDE (or ATA) ports are often used for the connection to the internal hard drive, and they are often described as ATA, PATA (parallel ATA), Ultra ATA or ATA 100 connection. The fastest data transfer from an ATA port is 133 Megabyte per second.
The more recent connection interface between a motherboard and a hard disk is the Serial ATA or SATA port. Serial interfaces are just that, serial and thus don't have to coordinate between multiple lanes of data. This makes the connector more narrow and the speed of communication can be faster. The two standard speeds for SATA are 150 Megabyte per second (often called SATA 150 or SATA 1.5) and 300 Megabyte per second (often called SATA 3G). These communication ports are faster but the bottleneck is not the ports but the drives themselves. At present no single hard disk drive (including the Raptors) can transfer data at the high 300 Mb/s rate. So SATA 3G may be overkill (except possibly under RAID configurations which I will not get into). However, if the motherboard is willing to give the capability to you, you might as well take it. Motherboards that are SATA capable also have separate ATA or PATA connections so you get both options. Even if your Hard Drives are all SATA connections, optic drives (such as CD and DVD) are still PATA or ATA connections. Edit: Recently some of the DVD's have come out with SATA interface.
One also needs to look at the connections for the RAM. This is mostly for making sure one gets the right RAM type for your system as well as understand the total memory capacity of the RAM. The typical RAM connection types these days falls into two groups; DDR SDRAM (Double Data Rate Synchronous Dynamic Random Access Memory)and DDR2 SDRAM. The difference of the two ram is the amount of bitsize in the prefetch buffer space. DDR and DDR2 are incompatible with each other so cannot be interchanged (though a very few motherboards will let you use both (not interchangeably but separately)). More important, the hardware connection of DDR and DDR2 are different with DDR having a 184 pin connector and DDR2 having a 240 pin connector. So a DDR2 RAM stick will not even fit into a DDR connection and vice versa. The motherboard chipset controls the RAM for Intel motherboards while the internal memory controller in the CPU controls the RAM for AMD. So for a specific motherboard and CPU there are a variety of RAM speeds one can use. Usually if one buys a faster RAM than a motherboard can handle, it will just default to a slower speed. For OEM motherboards in brand name computers, this may not be the case and may only be able to work with a specific speed RAM. So aside from the RAM type (DDR2, DDR) one should look at what speed RAM a motherboard can accept. Most stores selling motherboards will provide all this information when shopping around. These days the amount of RAM memory a motherboard can carry is usually 4 Megabytes with two sets of two Megabytes in dual channel. For some smaller micro ATX motherboards, the RAM capacity may be 2 Megabytes. Edit: Recently, RAM has started to appear in DDR3 format and as high as 2Gb per stick for 8Gb total memory.
Another important set of connectors to consider are the peripheral connectors (often called PCI slots or Expansion slots). PCI stands for peripheral component interconnect and it is the standard connection architecture that is used by additional boards that connect to the motherboard (such as the sound card, the video card, the network cards and the recently popular physics cards). If you look at a motherboard, you will see a series of connectors running down the motherboard, parallel to each other. They will be near a set of slots openings in the back of the computer case. These are the PCI connectors.
PCI connections have been around for quite awhile and are the standard area for computer expansions. In the old days, when the graphics of the computer was limited, the PCI connection was fast enough for data transfer (the PCI slot speed is 33 MHz). However, with the recent power accelerated graphics, PCI no longer can adequately handle the video card's speed. Thus a faster (100Hz) connection was developed called PCI express or PCI-e. The PCI-e runs at a faster speed of 100 Mhz. The PCI Express c connection has several configurations and can be available in PCI-e x1, PCI-e x4, and PCI-e x 16. The preferred connection for modern day video boards often use the PCI-e x 16 slot for fast communication.
In some boards, the video is connected to an AGP or Accelerated Graphics Port. AGP was developed when PCI slots could no longer handle the modern video requirements. Now with the PCI-e x 16 available, more and more motherboard companies are switching from the older and slower (though not by much) AGP slot to the PCI-e x 16. Much of this is more toward group conformity than technology requirements since the AGP is still fast enough for modern graphics boards. Yet due to mass appeal, manufacturers are migrating to PCI-express at the common graphics interface connection. Still many of the boards use AGP as the graphics board interface.
In addition, there are the miscellaneous connections that are either on the motherboard itself or situated out the back connection panel of the motherboard. They include everything from PS2 connections for the mouse to USB 2.0 connections for peripherals. The more complex the chipset, the more connections you are given. Some such as USB connections are quite useful for everyone and with more USB devices these days, one cannot have enough of these connections. Typical motherboards have like eight to ten USB connections divided with around four connectors out the back panel and four to six more onboard connections on the motherboard. Other connections such as optical connections and Firewire ports are more optional to each user. Onboard graphics and sound can add more connections to the motherbard and there are several fan connections on modern motherboards for the CPU, chipset (if not passively cooled) and case fans. The BIOS will often provide the rpm information of many of these fans so that you can check to see if the fans are working correctly.
Also, there are options such as the availability of onboard graphics (such as ATI Radeon X200 and N'Vidia GeForce 6150) in which case you get some basic graphics capability without having to purchase a separate graphics card. Similarly one often finds some sort of on-board sound capability so that one does not have to purchase a dedicated sound card. These on board systems are often not as good as dedicated cards but can provide capability for budget builds. The one on-board system that is quite good is the network interface connection or NIC and typical motherboards provide gigabyte rate NIC for ATX boards and 100 megabyte rate NIC for micro ATX boards.
Finally there is the size of the motherboard. Motherboards come in standard sizes so that they can universally fit most computer cases. The two most common sizes for desktop computers are ATX and micro-ATX sizes. Most do-it-yourself systems have ATX sizes while most smaller systems and brand name systems (such as Dell, Gateway, HP, Compaq, Emachines etc) use micro-ATX boards. Micro ATX has less capability than their ATX cousins but they are also smaller and will fit smaller cases.
Build Flag:
By choosing the CPU type and corresponding Socket (LGA 775 for example for Intel chips and AM2 for AMD chips) you have narrowed the choices of motherboards thus making it easier. Still there usually are a whole myriad of motherboards one can choose from. Your other choices that can further narrow the fields are:
1. Chipset used on the motherboard. 2. Onboard graphics or not. 3. micro ATX or ATX.
The Chipset is probably the most important part to check. This will determine much of the capability of the system such as the maximum RAM memory one can have, number of PCI peripherals, graphics connection type, network capability, etc. If one was to read up on anything, it should be the chipset.
For example one of the popular chipsets for AMD systems is the N'Vidia NForce 4 series. This is one of N'Vidia's most recent chipsets (recently taken over by N'Vidia's NForce 5 series). It is quite popular since it gives a wide array of capability: 4 Gigs of RAM in dual channel pairs, 10 USB expansion, PCI express connectors, works with all K8 CPU's (though AM2 uses NForce 5), has SATA 150 (NForce4) or SATA 3G (NForce4 Ultra) and a whole slew of control capability in the BIOS. It was also the first chipset to use SLI capability for multiple graphics cards. Of course if one wanted to go ATI crossfire then one wanted to get a ATI chipset instead and if so the Radeon Express 3200 was the chipset used. Micro ATX boards had their own version which was N'Vidia NForce 410 and ATI Radeon Express 200.
The choice of whether one wants on-board graphics tends to depend on whether one wants to spend money on a graphics board or not. For a game system, you really need a separate graphics board so on-board may just be paying for something not needed. However for people who do not want to buy a separate graphics card or for a simple system only used for web surfing it may just be the value choice.
Once one chooses a chipset and some of the peripherals needed, one can then choose a motherboard manufacturer. Since the chipset determines much of the motherboard's capability, the different brands really determined quality of build. You could go to different forums and see what were the popular brands. If you did, you found people to be fans of one brand or another. Some are die-hard ASUS, Gigabyte, Abit or MSI users while others did not much care and went with the cheapest brand or the one that had the most bells and whistles. DFI has a specific fandom among overclockers due to its very versatile but temperamental BIOS capability. Again, doing some due diligence in this area is useful.
You can use web retailers such as NewEgg to get basic information on a motherboard. You can also use it to narrow down the choices by filtering to your specific CPU, socket type and some capabilites (such as graphics connectors, hard drive connectors, etc).
(Next section RAM)
|
|
|
Post by osiris on Nov 5, 2006 21:32:44 GMT 1
Sorry if i'm disturbing you, but leave you all alone in this land seems very bad to me. Thank you very much for this guide, Nostra1: i hope that the other members will find it interesting as well. By the way, i'm thinking to start a similar one by myself about "how to clean up the inside of a computer case", 'cause the tons of soil collected by my PC, which are causing strange rumors and probably also the famous overheating. Yes, it's time to make a good cleaning.
|
|
|
Post by Nostra1 on Nov 6, 2006 3:09:44 GMT 1
Thanks for the kind words Osiris. Sorry that the writing is taking time.
I think your "how to clean out an old PC" would be great. I recently cleaned out my wife's old Pentium II system for upgrading, and it would have been useful for that. One thing I learned from that project was "wear a dust mask or at least a towel". When I opened up the front panel, the dust bunnies attacked and I got a face full of dust. I had a sore throat the next day. So now I use a mask as often as possible though it makes me sound like Darth Vadar (which I guess is why my son keeps stealing it).
The system is now upgraded and now is a beater wagon web surfing computer. It needs a bit more RAM (I used some sticks lying around from an old Emachine) but it runs quite nicely now.
Looking forward to your thread.
|
|
|
Post by Nostra1 on Nov 14, 2006 16:38:16 GMT 1
RAM RAM or Random Access Memory is probably the most common upgrade part for your computer (though the video card is coming up fast in upgrade popularity). It is probably the easiest part in terms of physical installation since all you do is pop it into the RAM slot and you are done. Unfortunately it is also one of the most confusing parts to get to work. More than one person has tried to upgrade the RAM only to find their system not to boot, to become unstable and crash, see the infamous Blue Screen of Death (BSOD) or in worst cases corrupt Windows such that you had to reformat your hard drive and re-install Windows. Much of this comes from the fact that there are a lot of different formats for RAM and much of the architecture is not universal. In the case of CPU, there are only two manufacturers and thus the architecture for hardware compatibility is better defined. If one chooses a Pentium or a Athlon, one immediately knows the voltage of the CPU, the core frequency, the front side bus frequency (or the hypertransport frequency) and the chipsets designed to run these processors can design in the compatibility. For RAM, there are many RAM chip manufacturers (some common companies include Samsung, Hynix, Winbond, Micron, etc). And RAM sticks are not as universal in design. The physical RAM architecture may vary (SDRAM, DDR SDRAM, DDR2 SDRAM) RAM voltages can vary, the RAM speeds can vary, the latencies of the RAM can vary, it can be single sided or dual sided, and of course the RAM capacity can change. This leads to a lot more variations in the architecture of the RAM and a lot greater possibility for incompatibility. To keep things from going out of control, memory design is governed by the Joint Electron Device Engineering Council (JEDEC). Even so, there are still enough variations to create incompatibilities. Thus the main point is to make sure that the RAM you purchase is compatible with the motherboard chipset of the system you are building or upgrading. It may only take a few seconds to put in RAM but it can take time to make sure you buy the correct RAM. For brand name computers (Like Dell, Gateway, Compaq, and others), there is one additional constraint. To reduce costs, these computers often have their OEM motherboard designed ONLY to the system that is being made. This means that they only will design the chipset/BIOS to be compatible with the system being built. If you upgrade, you can hit incompatibility and RAM is especially susceptible to this. Thus there may only be a narrow choice of timings, voltage and speed of RAM that can be used with a specific brand computer and when upgrading capacity (adding more RAM), one must stay within the tight parameters. If you own a brand computer, before you go buy additional ram, find out what sticks are compatible. To do so, you can go to some popular RAM manufacturer's websites such as Kingston, Corsair, etc and type in your computer model number. They will then tell you which model RAM works with your computer. This sure beats buying randomly, paying a good amount of money, happily sticking in the RAM and then watching your system crash. Do it yourself computers and systems built by smaller stores, use Retail motherboards (such as ASUS, ASRock, EPoX, MSI, Gigabyte, ECS, Jetway, etc) and these boards can accept a larger variety of RAM (since the board must be able to handle the variety of systems that are built by do it yourself builders). Even so, there is a limit to the variety of RAM that can be installed into these systems. For example in Athlon 64 CPU's the memory frequency is directly related to the CPU core frequency. Since the default frequency is 200Mhz for the CPU and DDR SDRAM gets two memory actions per cycle, the default memory frequency becomes 400MHz. That is why you often see DDR-400 (also called PC3200) memory as the default memory of modern computers (that is until DDR2 recently took over the mainstream market). Unless you increased your core frequency (and thus the memory frequency), you could buy DDR 500 RAM at a higher price and stick it in your computer and see that it runs only at stock DDR 400 speeds. So understanding RAM is quite important if you want to build your own system or even if you want to upgrade a system you bought at a store. RAMS have changed much over the past few years. If your system is old enough, it may still be sporting SDRAM. Synchronous Dynamic Random Access Memory is a type of RAM that works in place with the internal clock signal of the processor (the CPU). It will wait for the clock signal of the processor and perform its function according to the timing of the processor. The old SDRAM performed one memory function (triggered by the rise in voltage portion of the square wave) per cycle. Since the processor ran at 100Mhz, so did the SDRAM. So they were often designated as PC100 SDRAM. They had a pin layout of 164 pins per memory stick. As this memory is now obsolete, I will not go any further into this. Up until recently, the most common RAM type used was the DDR SDRAM (Dual Data Rate Synchronous Dynamic Random Access Memory), which was the next generation to the SDRAM. Unlike the SDRAM that performed one function per clock cycle of the processor the DDR SDRAM performed two operations (one on the voltage rise of the square wave and one on the voltage fall of the wave). Thus for the same 100Mhz cycle of the processor, the memory could perform 200 operations thus giving it a 200Mhz effective frequency. The amount of information passed during an operation was 8 bits. So at 200Mhz, the memory could move 1600 Mbits/second. The processors these days move at 200Mhz clock cycle so the corresponding memory is 400Mhz (thus the name DDR 400) and can move 400Mhz x 8bits/hz = 3200 Mb/s (thus the other name PC3200). So if you have a computer that is about a year old, it most likely has DDR 400 memory. Older computers had slower processor clocks and the memory could be DDR 333 (PC 2700), or DDR 266 (PC 2100). For do it yourself motherboards, most could accept any of the above memory. For brand name computers, you often had to get the exact correct RAM timing (once again be careful about this if you want to upgrade a Dell, Gateway, etc). The physical architecture of DDR RAM was 184 pins. Thus SDRAM did not even fit into a DDR SDRAM slot. The RAM used for Celerons, Pentiums are DDR SDRAM. Similarly, Sockets up to 939 in AMD ,motherboards use DDR SDRAM. The most recent memory is DDR2 SDRAM. The difference between the DDR SDRAM and DDR2 SDRAM was the size of the prefetch buffer (the temporary storage in the RAM during I/O). The DDR2 buffer size was 4 bits verses DDR's 2bit. In both the DDR and DDR2, the memory is double pumped, or performs 2 actions per cycle. So the frequency of a DDR2-800 was similar to a DDR-400. However, since the prefetch buffer was twice the size, a 200 Mhz processor double pumped gave DDR2 twice the data rate and thus DDR2 was denoted 800 while DDR was denoted 400. Intel's Core Duo and Core 2 Duo (Conroe) used DDR2 and Socket AM2 on AMD motherboards used DDR2. The architecture was a 240 pin design so DDR and DDR2 did not fit each other. In addition to the architecture of the RAM (SDRAM, DDR SDRAM, DDR2 SDRAM), and speed of the RAM (PC100, DDR 400, etc), there is the RAM timing, also known as latencies. Aside from the bandwidth or speed of the RAM, there are certain timing or (increments of time) that a specific RAM will perform a specific action when acquiring or storing data. RAM requires time to find a specific data, read it, go to the next data block necessary, read it, return to provide the data to the CPU, go to find an open spot to write the data from the CPU, etc. It has been said that these functions are not unlike a person's ability to quickly read and write data in a spreadsheet. The faster the RAM can perform these functions, the faster the data operation. As the RAM trades data to and from the CPU via the front side bus (or the Hipertransport), it has to synchronize with the timings set to be able to provide the adequate data. If the data timing does not match, the RAM may fail to provide the data in a single cycle, thus slowing down the performance, or may not be able to provide the data in any cycle, thus hanging up the system (can we say system failure?) or worst of all give bad data that can result in data corruption. The timing of the RAM is given by a specific set of values often called RAM latency. though there are many values, the most common values given are the Command Rate, the Common Address Strobe or Common Address Select (CAS latency), RAS precharge delay (tRP), RAS to CAS delay (tRCD), and Active Precharge delay (tRAS). Each of these sets a specific timing for the RAM to operate. For example the Command Rate is the time delay between the time the RAM is chosen to the time the actual command is sent to the RAM and is given by 1T or 2T. 1T is one cycle, 2T is two cycles. As you can imagine 1T is faster than 2T but some RAM cannot move this fast so would miss the commands sent out causing RAM failure. The most common referred value, the CAS latency is the time (number of cycles) between when the command is sent to when the data arrives at the data bus. You can understand that lower values means faster RAM. The faster the RAM, the faster the system can function. The timing of the RAM is often given in CAS-tRP-tCRD-tRAS-Command Rate sets. You can often hear people talk about their ram being of the timing 2-3-3-6-1T. This talks about the specific timings within the RAM. For DDR RAM a timing of 2 for the CAS latency (the first number) and 1T for the command rate (last number) is considered fast while a CAS of 3 and a command rate of 2T is considered slow. Even for the same timing RAM (for example DDR 400) a RAM with 2-2-2-5-1T will usually be more expensive than a RAM having 3-3-3-8-2T. A tighter timing also suggests the RAM to be more stable and causing less failures. Thus when looking at RAM, one needs to know the architecture, (SDRAM, DDR, DDR2), speed (DDR-333 (aka PC2700), DDR 400 (aka PC3200), DDR2 800 (aka PC2-6400)), timing, as well as the standard memory size (512Mb, 1Gb, etc). Build Flag:As most old systems only use SDRAM, and this is a new build discussion, I will not go into this area. For newer systems the RAM used is either DDR or DDR2. If you bought a brand name computer, your owner's manual (I hope you kept this) should tell you what the memory type is. If you don't know, then you can download two freeware, one called CPUz and the other either SiSoft Sandra or Everest Home (Google these and you can find them. If you have trouble, go to www.majorgeeks.com to download). These diagnostic software will tell you exactly what type of RAM you have. Make sure you buy a similar RAM type. If you are starting a build, then its easier. Just find out what RAM your motherboard will take and buy that RAM. Typical values are DDR 400 and DDR2 800. If you are not planning to overclock and just want to run the system at stock speeds then get a nice cheap value RAM that can run the computer at reasonable latencies (CAS 2.5 or 3 for DDR and CAS 5 or 6 for DDR2). Some good RAM companies are Corsair, OCZ, Patriot, GSkill, Crucial, Geil, Kingston, Pqi and PNY. Buying a RAM that runs faster than stock speed is a waste of money if you are not planning to run faster than stock speed. There is only a 4 slots usually in a given motherboard (sometimes only 2) so buy a decent size RAM so that you can upgrade later without having to pull out old RAM. RAM can be expensive so many people buy slowly and build up their RAM capacity. Finally and quite important, just because Windows booted up does not always mean your RAM is running stable. If the RAM is partially unstable, it can still partially run Windows. What you will see however is that as you continue to run the computer, it will continue to get harder to run and crash. What is bad is since it ran for an extended time with instabilities, it may have given corrupted data which got written into the hard drive. If this data is related to Windows settings, you may end up corrupting Windows such that you have to re-install the hard drive. The best way to avoid this is to get a memory testing software. The most common software is one called Memtest86. This software will run your computer hard by sending several mathematical commands that severely uses the RAM. It checks to see if there are failures in the mathematical calculations. If there are, it reports these failures. Memtest is especially good for RAM as its tests RAM extensively (more than software such as Prime which tests the CPU). So if you just bought a stick of RAM and put it into the computer. Run Memtest for several hours and see if it produces any error. If it does, immediately return it to the store for a refund or exchange. If it passes, you can feel quite comfortable that the RAM you bought will not cause you problems down the road. (Next section: Hard Disk)
|
|
|
Post by Nostra1 on Dec 2, 2006 22:07:30 GMT 1
Hard Disk:
Now that we have the CPU, Motherboard and RAM, we have passed the most difficult part of choosing parts for the build. The rest of the build will be easier to choose. In fact when upgrading to a new system, it is usually these components that together can set an upgrade. Most everything else can usually be kept from the old system (provided the hardware connect-ability is compatible). One such component is the hard drive.
The hard drive has changed little in basic architecture over the years though the performance of the hard drive has increased dramatically. It is still the main component for data storage in your computer (including the operating system). As such, even 10 year old hard drives can still be connected to a modern computer through the IDE connector and though it cannot run the operating system, can be used as a secondary storage system. The only thing is that a 10 year old hard drive often only has 1 or 2 Gigabytes of storage space and is quite slow. A small memory stick can outperform such a drive. The point is, when upgrading, you can just pull out your hard drive from your old system and use it with your new system. If it has a minimum of 20 Gigs, it still can function well in modern systems (even 10 Gigs can work).
Hard drives are just magnetic disks capable of holding a large amount of data in a single disk. In addition, there are several read/write heads (basically a thin film induction coil that can generate or read a magnetic field). The capacity and speed of these drives have increased over the years. It is typical to see 100 Gb or greater on modern hard drives and a spin rate of 7200 rpm (disk spin speed) vs the 5200 or less of older drives. Usually a notebook computer will have a slower hard drive speed. Finally the hard drive has mechanics that protect any data loss during physical movement and the data bus that transfers the data to and from the drive.
Aside from the physical capacity of the drive (given in gigabytes) and the speed of the drive (given in RPM's), the other area of choice is the mode of data transfer. The two basic differences are ATA and SATA. If it is a modern motherboard (such as for a new build) it will most likely have both ATA (aka IDE) connectors as well as SATA ports. The only thing to check is if the motherboard accepts SATA 150 or the faster SATA 3G (3.0Gb) as hard disk drives can be purchased for either. Usually SATA 3G drives can be configured as a SATA 150 using jumpers but SATA 150 will not be able to run SATA 3G. On the other hand motherboards that can run SATA 3G will usually run SATA 150 as well (since it is slower).
Build Flag
Unless you are a hard core competition gamer where every second counts, you will not need the fastest hard drives available (such as the 10,000 rpm Western Digital Raptors and 15,000RPM Seagate Cheetah's). A typical hard drive running 7200RPM is plenty fast for the majority of us and is a LOT less expensive than the fastest drives. In the United States the most popular of the drive manufacturers is Seagate and Western Digital. However, for other countries it may be better to find a more domestic brand such as Samsung for Korea and Hitachi for Japan. Find what is commonly available in your area.
Aside from capacity and speed, the most important aspect of any hard drive is its reliability. The most traumatic problem often seen is when one loses all their data in their hard drive after a crash. Of course having a backup drive and setting your drive up in a RAID array can help reduce risk but another way is to buy a drive that is known to be reliable. You can check various forums to see what drives are popular. In the United States Seagate and Western Digital makes reliable drives and thus are quite popular.
Finally cost is a factor so find a drive that meets your capacity requirement and budget. Make sure it can connect to your motherboard (check your motherboard manual or specifications) and make sure the drive you choose is known to be reliable and you will in most part be fine.
One thing you may find when shopping for hard drives is the physical size of the drive. You may find both 2.5" drives and 3.5" drives. The 2.5" drives are for notebook computers. For desktops, you want the 3.5" drives.
If you can find disk drives sold as OEM products verses retail, you may want to consider this. OEM drives are often sold to companies that manufacture computers and as such are perfectly good drives but only come as a drive. They do not have a box, a manual or connectors. On the other hand, unless you like cardboard boxes, this is not too important and there is little to read to connecting a hard drive (connect the SATA or IDE, and connect the power plug and you're done. Its more difficult to connect the back of your PC than it is to connect a hard drive). Finally, if you bought a retail motherboard, it comes with IDE and SATA connectors so you don't need these. OEM drives are usually less expensive and can save you money. I may hesitate in buying an OEM motherboard but an OEM hard drive is a good option.
(Next section: Power Supply)
|
|
|
Post by Nostra1 on Dec 6, 2006 23:36:14 GMT 1
Power Supply: The power supply is often considered a necessity more than a nicety. It's the one that most people only consider as making the computer run and nothing more. People think it really doesn't do anything to the programming or direct actions of the computer. If one was to open the computer up, one finds the power supply as the ugly gray box that sits at the back of your computer. Thus most people do not even care about the power supply when choosing a computer. When one does decide to choose a power supply when building their first computer, the most common action is to buy the biggest power supply at the lowest price. If you find a 700W power supply for $20, you feel like you were a smart shopper and wonder about the people who buy $100 power supplies that only supply a mere 500W. The power supply is basically the foundation of your entire computer. Similar to the foundation of a house, it many not be the most visibly impressive part of a computer but it provides a very important function. This function is to supply a very stable amount of electric power to your valuable computer components. Bad power can cause system instabilities that can result in your computer hanging up, restarting, crashing and in worst cases making other components not work properly (such as the hard disk corrupting Windows). The worst case however is when the power supply completely fails and sends a surge of voltage to your computer, destroying your components. There is more than one story where a person destroyed their brand new built computer by saving money and buying a cheap power supply and watching it "pop" and destroy their system in the first power up. Even if the power supply works, higher quality power supply can provide a steady level of power on the 3.3V, 5V, and 12 V rails, while lower quality power supplies have voltages that can supply ripples, noise or sag in the voltage. Finally, the rated power of a power supply can be determined by different criteria depending on the manufacturer. Quality power supplies rate the power as the power level where the power supply can provide efficient power. When the power supply goes beyond the power rating, the power efficiency may start to decrease but the power supply may still function properly. Poor quality power supplies may rate the power as the power level where the supply will actually fail. So the same 500W value may be different between two different supplies and some supplies may not be able to provide stable clean power at 500W even if they are advertised as such. So be careful and don't assume bigger is better. A PC power supply is basically a power rectifier that transforms the 60Hz (in the US) AC power to a steady state DC power. This DC power is supplied at three different voltage levels; 3.3V, 5V and 12V. Since the output is DC, the voltage needs to remain a steady flat value as close to the above three voltages as possible. All components in a PC then runs from one of the three voltages. These days there are often multiple rail power supplies (dual rail, tri rail and quad rail). These are supplies where the current path is split into multiple rails so that the total current down any single rail remains low. With lower current, there is lower heat dissipation and thus the power supply can run cooler. The trick is to never connect too many components on one single cable (don't continue to add adapters to connect more stuff onto one cable). In general multi rail is better than a single rail. However, it more important to have quality rails than to just have more rails. A high quality single rail power supply may be as good as a dual rail power supply. Also, though the power rating of a power supply is important, its ability to carry current is just as important. The more components one connects to a power supply, the more current the power supply has to provide. If one goes over the current capability of a power supply, one can burn out the supply and make it fail. Thus one can look at the current ratings for a power supply to see how high a current it can handle. Much of this information can be found on the power supply itself. There is a table written on many power supplies that will give the current rating on all its rails. This is especially true for higher quality power supplies. This current rating, compared with its power rating can give a good indicator of how stable the power supply will be. For example even if a power supply is advertised as a 600W power supply, if it only can carry 16A on the 12V rail, its probably not going to be able to supply a lot of power. On the other hand a 400W power supply having 24A on the 12V rail may be more stable at the higher power. Multiple rail power supplies can have lower current ratings on the individual rails since the power supply will split the current among the various rails (this is the reason why one should not connect too many components on a single rail). So a 15A 12V rail and a 14A 12V rail on a dual rail power supply may be similar to a single 25A 12V rail for example. Finally, there are really only like 10 power supply companies that make the power supplies and the rest of the companies repackage these supplies and sell them as their own brand. This does not mean the repackaged supplies are bad. They can sometimes be lower cost than the models sold by the original manufacturer. The point is to make sure the power supply is manufactured by a reputable manufacturer. Build Flag:Most rigs do not need more than 500W to run them. In fact a simple system with a single core CPU, a non-SLI or Crossfire mobo, 1Gb of RAM, one Hard Drive, one DVD burner and one video card will usually only use up to 300W of power. So buying a gonzo large 700W PSU is not necessary. If however you are building a high end system having a dual or quad core CPU and a dual SLI rig, you will need that sort of power. Determine what you want to build and that will determine the power supply requirements. You can go here to find out the approximate power requirements of your setup. www.extreme.outervision.com/psucalculator.jspAs for single or dual rail, try to get a dual rail power supply if possible. If its single rail, make sure its one that has at least 24A on the 12V rail. Of the brands of power supplies to get you should stay with a reputable manufacturer. These include Seasonic, Silverstone, Fortron Source, Corsair, Antec, PC Power and Cooling, Zippy, Sparkle, Enermax, OCZ. There are others and you can look at the reviews to see how good the supply is but the above list is pretty good. Of the above the brands that make their own power supplies are Seasonic, Enermax, Zippy and Fortron Source and all four companies are very good. In fact Seasonics parent company Etasis actually makes power supplies for high level server machines. The rest get other company's power supplies but can be fine. For example PC Power and Cooling, Silverstone, and Corsair get their power supplies from Etasis as well. OCZ gets its power supply from Fortron Source. Thermaltake makes a very good line of power supplies in their high end Toughpower series. However their lower end power supplies are questionable. They are not bad but usually Thermaltake overrates their power capability. I have one of their low end 430W power supply but its only really a 300W power level as reviewed by Xbitlabs. If you want a basic PSU for a low cost, look at Sparkle and Fortron Source. Another decent cheap power supply is the XClio Goodpower power supply. The Sunbeam Nuuo seems to be pretty good as well (don't buy any of the other Sunbeam lines though). Also the Hiper Type R 580W power supplies are pretty good supplies. A good list of power supplies can be found here in the DFI forum. DFI motherboards are quite picky due to their tough demands for overclocking (which requires very stable voltage) so their level of satisfaction can satisfy most users. www.dfi-street.com/forum/showthread.php?t=10854One big thing to remember is don't use the power supply that comes with your case. This is especially true if your case is really cheap. The one exception to this is Antec's cases that can have a decent power supply inside. Even then, make sure the power supply inside is one of Antec's good models (such as the Truepower line). Modular PSU's are all the rage but they can be a hassle if you have to connect a lot of components to the system. Also though many talk about PFC or Power Factor Control, I would not make that a deciding factor considering the 10% greater efficiency only means a couple bucks savings at the most on your electric bill. Choose a price range and power level and then shop the brands for one that looks good. (next Optical drives)
|
|
|
Post by Nostra1 on Dec 15, 2006 5:58:03 GMT 1
Optical Drives Well, this is one section I am not really familiar with. Optic drives are one of those very visible components on your computer. However, a typical DVD is not that expensive a product and is relatively simple to connect. The optic drive is even more simple than a hard drive. Optical drives are come in the old style CD drives, Combo drives (CD RW, DVD ROM) and DVD Burners (CD RW, DVD RW). With the exception of the most recent drives most CD drives are ATA (also known as IDE) connections. They really do not exceed the speed of an ADA connection. Most optical drives all work in similar manners and with the exception of new drives such as Blu-Ray, they are basically the same in construction. The difference is mostly in the quality of the product. This is determined by the burn quality, the read quality and the reliability. Also the noise level of a drive is quite important to many. The standard drive is a DVD +/- RW burner. Even the standard 16x DVD burner can rip and burn a DVD very quickly compared to the drives of a few years past. What is nice is that the typical drives are quite inexpensive. Higher cost often doesn't mean better quality as the higher cost drives are just repackaged lower cost drives. There are additional versatility capability such as bitsetting that I do not understand so I will not discuss. For 16x DVD burners some popular brands are BenQ (no longer available as the company was bought out), Lite On, NEC, LG and Pioneer. Samsung seems to be nice but a notch lower and Sony does not seem to be popular when one reads the DVD forums. Plextor seems to be nice but much more expensive than the others for little return. In any case, one should look at reviews and forums for popular drives. As the modern optical drive function is similar to the drives of old (but faster) you can still connect up your old CD RW to your new computer. Just connect it to the ATA connection and your motherboard should recognize it right away (just check in the BIOS). Also, since there is not much to a DVD drive except the drive itself, it is also another product you can buy as OEM (bulk) over retail. It will not come in a box but should have no problem. Just connect the power and the ATA and you are done. There is a major fad for Blu ray these days. These new drives use a blue laser vs a red laser. The blue laser has a smaller wavelength and thus can write smaller. This means more data can be put on a single disk. However, the price of these drives are very high (sever hundred dollars or euro vs 40 dollars or euro for a bulk regular drive). It may be best to wait for the price to come down if you are just a casual user of DVD as I am. Build Flag:The newest drives (aside from Blu ray) are x18 drives. Once again, look at forums and reviews to determine what drive is popular at this time. In most cases, it should probably work better than many no name drives in brand computers. Once again, if you can get bulk, buy bulk. Edit: One forum you can visit for DVD info. club.cdfreaks.com/(Next Video Card)
|
|
|
Post by Nostra1 on Jan 2, 2007 6:14:04 GMT 1
Video Cards
If there was any piece of hardware that influenced your ability to play a game, it is the video card. That is because most every game has some level of 3D rendering and almost all games have high levels of 3D graphics. One can play games with some of the older generation CPU's but most require a modern video card to play a game.
The Graphics Processor Unit or GPU is basically a dedicated processor whose sole function is for graphics processing. A simple image is not the issue for such processors. It is basically the 3D animation where such processors are required. And since we are talking about games here, a graphics processor becomes an important part of any gaming computer.
As the name implies, the graphics processor is designed to process graphic images on the computer. As the most process intensive tasks are 3D images, the GPU is made as a sophisticated vector image processor. As such the graphics processor has many components that are required to process such images at high speeds. Of these components, there are three that are often focused when talking about a video card's capability. These are the Vertex Shader, The Pixel Shader (or Pixel Pipeline) and the Raster Operator.
The basic 3D image is a configuration of many many polygons that are oriented to form a large solid image. As the image moves or rotates, each of these polygons are transformed via mathematical functions to move in the exact manner that makes the whole solid image look like it is moving. In order to do so, the GPU needs to be able to process millions of these polygon images each second.
To keep things simple, lets just consider the polygons to be triangles. Then to draw a triangle in space, you basically need to identify the three points that make up a triangle. Then these points can be connected to form the triangle. So long as one knows the location of these three points in space, one can continue to draw that triangle. If the triangle moves, that can be denoted by the movement of these three points. So the GPU has to keep track of these three points and how they move per frame. This is done by the Vertex Shader. The Vertex Shader can be considered a triangle drawer and in a single frame will draw all the triangles necessary to make the image. Since each single item in the 3D image consists of many triangles, the more items in an image, the more triangles are required and the faster the Vertex Shader has to work. This is why grass in Oblivion is so painful on the GPU since each blade of grass has to be drawn individually using triangles and then these triangles has to move independently each frame.
In addition to the objects shape is the objects color. Each one of these polygons have to be filled in with a specific color of an object. Things get more complicated when there are complex lighting (such as shadows, flickering lights, sparkles and explosions) that need to add layers of colors onto the base color of these polygons. These specific layers of colors are filled by the Pixel Shader. The more triangles the image and the more color complexity (such as lighting effects) the more colors the Pixel Shader has to fill in. So adding complex shadows and high dynamic range in a high resolution, high complexity image having lots of grass can be quite taxing to the Pixel Shader. The more complex the image coloring (such as in Oblivion (especially at the OB gates)) the more the pixel shader has to work to make the image.
Finally aside from color and the object generation, there are some post image generation processing that are done to enhance the image. For example, there are often situations where the triangles that are used to make the objects don't fit exactly smoothly. This creates a zig zag shaped image with sawtooth like patterns on the edges. This is called aliasing. So a post image processing to smooth between these patterns is called anti aliasing. As this smoothing requires mathematical processing, this operation is done by the Raster Operator (or ROPS). The more ROPS a processor has, the more complex image manipulation processing can be done.
Aside from the graphics processor, all modern graphics cards have some level of dedicated RAM used only for graphics processing. Just like more RAM can speed up the bacis functions of a computer, more VRAM can speed up the functions of a graphics card.
Finally with new processors coming out, they are able handle the latest graphics software such as Direct X or Open GL. Some of the older cards often cannot play a game not only due to the limitations of the hardware but its inability to run the 3D imaging software.
Build Flag:
First and foremost before dong anything, determine what connection your motherboard has regarding video cards. It can be PCIe x 16 or AGP or sometimes even PCI. Make sure you get a card with the proper connector.
The majority of the most popular graphics cards uses a GPU from one of two manufacturers, ATI or NVidia. Both companies have very powerful video cards for processing the most sophisticated graphics around. These graphics cards can be made to work well either as a graphics design card (such as the NVidia Quadro series), a game card (such as the NVidia GeForce series) or a TV tuning card (such as ATI's Radeon All in Wonder series). Make sure you choose the appropriate card for the application.
Once you choose the processor manufacturer and the card type, you can determine how much power you want. Usually this comes down to how much power you can afford. Usually when a manufacturer comes out with a new processor (such as NVidia's G80 processor), they create a new series of cards based on the processor (such as the GeForce 8000 series using the G80 chip). Within the series, the same processor architecture is used. However, there are difference in the performance of video cards within a specific series. A new series usually determines the architecture of the vertex shaders, the pixel shaders and the raster operators for that series. Much of the performance is separated into four aspects. One is the actual speed of the processor itself (just like with CPU's) and higher speeds means better performance and higher cost. The second is the number of vertex shaders, pixel shaders and ROPs the card may use (thus people often talk about the number of pixel pipelines a card may have) with more shaders and ROPs giving higher performance. Another is the bitsize of the information transferred during a single cycle for the GPU. The larger bitsizes means a faster operation. Finally, the amount of VRAM a card may have with higher RAM capacity to be a faster processor.
What is nice is that both ATI and NVidia sets standards on how a video card can be designed using their processors. So for example a typical ATI X1600XT will perform similar for all video card manufacturers. You can expect similar performance for a given type card from any of the manufacturers.
One final point to make is that GPU's like their CPU counterpart have specific power consumption and heat cooling requirements. Some of the newer boards are quite power hungry and run very hot. For example, if you were to planning to put two GeForce 8800 GTX boards in SLI you better have a decent 600 - 700W power supply (Such as the OCZ GameXtreme 700) to be able to run such cards. Also you may want to consider an aftermarket cooler to keep your card cool.
So the basic is to choose the proper connector and read up what processor (or card series) has the strongest capability within your budget. Then you can decide which video card manufacturer you want to purchase.
|
|
|
Post by Nostra1 on Feb 14, 2007 10:07:41 GMT 1
Sound Cards: Cannot give any input on this since I have never used a sound card yet. Any experts on sound cards, any comment you can provide would be appreciated.
Operating System: Ok besides using XP, I have no comment I can give. Any Linux users out there, an input would be appreciated.
Computer Management: Since this is a vast subject area, I don't want to go into it too much. However, there are two specific areas of management I would like to comment on, one is thermal control and the other is freebie diagnostic software.
Of the many issues that can affect the computer, the two that often can result from hardware is voltage instabilities and temperature issues. The biggest contribution to voltage instabilities is the quality of the power supply and much of this has been covered in the PSU section. Sometimes it can also be in the quality of the motherboard although most motherboards can handle the basic voltage capability of a home build (assuming you are not doing any esoteric overclocking).
Temperature in the most part has recently appeared due to the performance increases in CPU and GPU chips. During the Pentium II processors and old graphics boards, the performance of the system was not to a point where thermal issues were a major problem. However, now with high level processors, many of these chips can get very hot. Thus adequate thermal management is needed to ensure proper performance and longevity of the system. This is especially so if you are running multi core systems and playing lots of games (which stress the graphic processors).
Stock cooling can be adequate in most situations but sometimes may not be enough if one is increasing performance of the PC (such as upgrades and such). If one is running the computer very hard, it may be necessary to increase the cooling performance within your computer.
There are many cooling methods used these days including air cooling, water cooling, thermal electric cooling (aka peltier coolers) and phase change cooling. As the majority of these cooling methods are quite esoteric and extreme, I will only focus on air cooling at this time.
The majority of computers these days use air as the medium for thermal transfer. As such the ability to keep your PC cool depends on how well the PC can transfer the heat from the hot points in the computer to the air, and then move the warm air out of the PC. As such, air flow becomes one of the most important points of the PC. Thus all air cooled improvement designs comes in the ability to rapidly move the heat from the processors to the air and the ability to move the hot air out of the PC. Any restrictions in this area can cause heat issues in your PC.
The above objective then requires two action. One is the determination of the heat sources in your PC and its ability to remove the heat to the air. The other is to maximize the airflow to move the hot air out of the computer.
For airflow, there are two basic variables to consider. One is to determine how the airflow is being made. This requires the use of fans and the careful arrangement to produce good flow. Just like the flow of water requires an inlet (source) and outlet (sink), the airflow also has to have decent inlet and exhaust. Typical PC design puts the inlets in the front and side of the case and the exhausts in the back and top of the case. Modern cases provide the ability to mount fans to provide the flow of air in these locations. The larger the fan, the more airflow for the same rpm. 120mm fans are the preferred fans these days and replacing any smaller fans with such bigger fans (if possible) often can help with the airflow.
The other variable is the minimization of any restrictions to the airflow. Restrictions can be anything from a badly mounted component, tangled cables, heavy dust and even walls. If one places a PC into a very tight cubicle, one should expect the airflow to become more restricted than keeping the PC in an open area. So be careful of tight spaces. Clean up wires to minimize air restrictions and make sure there are no clogged fans due to dust.
Once airflow has been improved, the other factor is the maximize the thermal transfer between the hot points in the PC to the flowing air. The main areas of heat production in a computer are as follows:
The CPU, the video processor, the motherboard chipset, the power supply, the RAM, the Mosfets on the motherboard, the hard drive and the optical drives. Of those listed, the biggest sources of heat are the CPU and the video processor. Any improvements in heat transfer should be focused on these two areas.
Both the CPU and the video cards come with stock heat sinks that assist the transfer of heat to the air. Though most are active (ie assisted by a fan) some can be fully passive. For any passive cooling, it is even more important that the airflow within your PC is adequate. As the heat output of these components increased with increasing performance, the method of transferring the heat has steadily increased. Modern, heat sinks are often very large objects whose large surface area is necessary to ensure adequate thermal transfer.
Heat sinks usually are divided into two basic designs, those that only use metallic vanes for thermal conduction and radiation and those that use a combination of metal vanes and heat pipes. Heat pipes are pipes that contain fluid having low boiling temperature. When the hot point transfers its heat to the heat pipe, the fluid in the pipe vaporizes and carries the heat to the other end of the pipe. The other end, being cooler, will then release the heat and change the vapor back to fluid to return to the hot end. It is quite an effective method of thermal transfer and air conditioners and refrigerators have often used this technology for many years. This is now used in PC coolers. If your heat sink has copper tubing passing through it, these are most likely heat pipes.
The standard material used in heat sinks are aluminum and copper. Both materials conduct heat very well and thus are favored for heat sinks. Of the two material, copper is a better heat conductor. However, copper is heavier than aluminum and is also more expensive. So to cut costs and make a heat sink lighter, many companies opt for aluminum. Most modern heat sinks use a combination of copper and aluminum to maximize heat conduction while minimizing the weight and cost. Some popular heat sink manufacturers are Arctic Cooling, Thermaltake, Thermalright, Noctua, Sunbeam (under their Tunic brand name), Scythe, Gigabyte, Zalman and Coolermaster.
As good as a heat sink can be, it is no use if there is inadequate thermal bonding between the heat sink and the hot source. The important point is to make sure there is a strong bond that is contacting a large area of the heat source to the heat sink. This is done by a pressure contact of the heat sink to the heat source. Most CPU processors have a intermediate metal plate on top of the actual processor to protect the processor. Most GPU's do not and thus care must be taken when applying pressure to the GPU. Without proper contact, you can have the biggest heat sink and still have poor cooling.
To maximize heat transfer at the bond, manufacturers use thermal interface material (aka TIM) between the heat source surface and the heat sink surface. Thermal interface material is a paste like substance that has high heat conduction. It is less conducting than pure copper or aluminum, however, it is much better than air. The point of TIMs are to fill in any air pockets that may be between the heat sink surface and the heat source surface. Popular TIM manufacturers are Arctic Silver and Shin Etsu.
Finally, these heat sinks are assisted by fans. The typical fan size of older heat sinks were 80mm or less. However, many modern heat sinks use 120mm fans to cool the heat sinks for a CPU. The 120mm push more air and rotate slower so they are much quieter than their 80mm brethren. However, they have made the modern heat sinks gigantic. Some examples of big heat sinks include Thermalright's SI 128, Thermaltake's Big Typhoon, Zalman's 9700, Scythe's Ninja and Infinity, Noctua's U12, and the biggest baddest air cooled heat sink of them all, the Tunic Tower. If you use any of these heat sinks, make sure your case is large enough to fit them in.
Build Flag:
If you are not planning to stress your computer a lot, then stock heating may be fine. However, the moment you start upgrading your CPU or adding video cards, be careful of heat issues. Some of the recent vid cards such as the ATI X1950XT and the GeForce 8800GTX radiate a LOT of heat and upgrading your vid card can turn your formerly fine computer into a space heater.
If you have heating issues (which will probably be noticeable since the air inside your PC will become quite warm, then consider upgrading your case fans). If you can upgrade from 80mm fans to 120mm, do so. If the is no place to add fans, you can always use PCI slot fans to help cool your case. They are cheap and though they take up one PCI slot, they can cool quite well. Also consider upgrading both your video card cooler and your CPU cooler. Just make sure these large coolers fit your system.
You can also purchase a fan controller to help control all these fans. This can be very helpful if you want to control the rotation of the fan to balance noise reduction with cooling power. Many fan controllers come with thermal sensors as well so that you can measure the temperature of whatever the thermal sensor is attached to.
(Next: Example Build)
|
|
|
Post by Nostra1 on Mar 6, 2007 17:57:19 GMT 1
Free downloadable software: Before going on to the example build, I wanted to cover the other part of computer maintenance, this is the downloadable utility software. With the increasing amount of people working on their computer, there are now a large number of utility software that has been developed to help diagnose your computer. Some like CPUz, Everest and Sandra 2006 are made to help determine what hardware is in your system. Others like Prime 95 and memtest are available to test the stability of your system. Programs like Speedfan can help diagnose the performance of your system. Software like PCMark and 3DMark are available to benchtest your system. Finally programs such as AVG, Spybot Search and Destroy, and Adaware help protect your system from malware. When you buy a computer, you often get a bunch of free software to help run your computer. Of course when you build your computer, such products are limited. This can be a good thing as much of the software you get free can be useless and slow down your system. To help run your system, you will get some basic software with the components you purchase. Motherboards will come with utility software (such as ASUS PC Probe that monitors all voltage, temperature and fan rate) to help run the computer. Video cards come with video utility software (like ATI catalyst) to help your video work better. Similarly for sound cards and input devices (mice, keyboard etc). This will help your computer run better. Between the freeware and the software you get with your components, you can do a lot with your computer. I will not go over all the software available but will focus on some software you should get to help run your computer. At the minimum, you should download CPUz and Sisoft Sandra to get the information regarding your hardware settings. CPUz will provide some basic information regarding your processor and RAM. It is good to run when you quickly want to get information regarding your CPU and RAM. Sisoft Sandra is a more detailed diagnostic software that will provide information to most of the components in your PC. It is so detailed, you may get lost in all the information it provides. Sandra can also benchtest the speed of your processors and RAM to see how your system compares to others. It will also help you burn in your components such as your CPU if you want. A similar program to Sandra is Everest Home Edition. Between the three software, you can get a very detailed printout of what is in your computer and how its set up. CPUz download site:www.cpuid.com/cpuz.phpSisoft Sandra download site:www.sisoftware.net/Everest Home Edition download site:www.majorgeeks.com/download4181.html(note, Lavalys no longer updates or supports the freeware edition. If you want to buy their corporate edition, you can from their site. However, the free edition is still useful) Another good software to get is Speedfan. Speedfan is a freeware that will display many of the on-board diagnostic sensors in your system. It will report temperature of the many thermal sensors. It will also report the voltage readings from your 12V, 5V and 3.3V lines. For dual core systems, Coretemp is another good software to get as it accesses the actual sensors at each CPU core (vs the temperature sensor at the motherboard (which is less accurate)) to provide a temperature. Speedfan download site:www.almico.com/sfdownload.phpCoretemp download site:www.thecoolest.zerobrains.com/CoreTemp/(Note, as mentioned Coretemp accesses the temp sensors at the core itself and not the sensors on your motherboard. As such it often shows temperatures hotter than the standard temp readings. Just a fyi in case you see two different temps from Coretemp and Speedfan). Stress software are also very useful to determine the stability of your system. These software usually stress your computer to working at its maximum output to see how well they work when pushed to the limit. Most of the time, your cpu is only working like 10 or 20% of its max output. This does not always tell you if your computer is running well since many problems occur only when the PC is pushed (such as playing games). These stress software will push the PC to its limit for you to see if any problems arise. The three main programs that people use to stress test their system are Prime 95, Orthos and memtest. All these programs are mathematical modeling programs that are done to calculate values. Prime 95 for example is a prime number generation program that mathematicians use to find new prime numbers. However, it can also be used to test the PC since it checks its math to see if the machine is generating the correct values. If there is any error generated due to a bad process, it will send a warning flag. Prime is generally used to check the CPU since it stress the CPU primarily. Memtest on the other hand stresses the RAM operations by constantly accessing the RAM space of your PC. It also checks for inaccuracies during the operation and sends a warning flag if there are problems. Orthos is a program that runs two instances of StressPrime. It is used to test each core of a dual core system. Using these systems and making sure your PC does not generate any errors is very useful to diagnosing the stability of the system. They are also great to test your PC once your overclock your processors. Prime95 download site:www.mersenne.org/freesoft.htm(Yes the site looks kind of suspicious. What do you expect from a bunch of mathematicians) Orthos download site:sp2004.fre3.com/beta/beta2.htmMemtest download site:www.memtest.org/(Note Intrepidacious writes a more detailed explanation and link to memtest in this thread: osiris01.proboards86.com/index.cgi?board=hs&action=display&thread=1170300401&page=2)Many times you may want to stress test the stability of your system before it enters Windows so that any instabilities does not corrupt the OS. For such occasions, there is the Ultimate Boot CD. UBCD basically puts a bunch of stress, diagnostic and other software that you would rather run outside of Windows into one compact diskette. The way to use this is to go into your BIOS and make your CD drive the first boot item and disable the hard drive boot option. Then drop this CD in and run the tests. If it passes then you can reconfigure your PC to boot from your hard drive again. Note: Be careful when running these programs as some can do some detailed stuff to the PC (such as the burn in software). I haven't used the majority of software in the CD so I cannot say how good they are but it seems to be a well reviewed utility. UBCD download site:www.ultimatebootcd.com/download.html(UBCD explanation page: www.ultimatebootcd.com/index.html)Once you build your precious computer, you want to compare its performance against others in the public. This is to make sure your system is running well and to many, gain braggin rights ;D. For that there are many benchmark software (such as Sandra) available. One company that provides many benchmark software is Futuremark. They are known for the 3DMark and PCMark software. 3DMark is a series of programs that test the capability of your video card setup while PCMark is more geared towards your processor. They are two versions available in general, the bigger one that you can buy and the smaller one that's free. The smaller one is a basic benchmarking software and many use it to record the performance of their system. These are very large programs due to the large video test animation that is included in each version. 3DMark comes in 3DMark 3, 3DMark 5 and 3DMark 6, the higher versions being harder on the graphics. There is also a 3D Mark 1 but it is published by Mad Onion and not by Futuremark. Futuremark provides PC Mark 5. Another graphic benchmark program to test your vid card is Aquamark 03. Some of these programs are interesting to run just for their test graphics. Return to Proxycon is a pretty cool animation short. Futuremark download site:www.futuremark.com/download/(Note you can download 3D Mark 01 from the discontinued section of this page) Aquamark download site:www.majorgeeks.com/download.php?det=2105Finally to protect your system, you can download some useful malware protection products. If you happen to build a computer that came with an Nvidia chipset (such as the NForce 4 or NForce 590) you will often get a Firewall program from Nvidia as well. However, this firewall has not been too popular as it can cause the PC to act up when you install the firewall. So you may not want to install such software when you get it. Three of the many software available to protect your PC that have been found to work well are AVG antivirus/antispyware protection, Spybot Search and Destroy and Adaware spyware protection. There are other software out there that work quite well such as Avast and Advir. I have not used Avast but it seems to have a good rep. I have Avira on my daughter's rig and it seems to work ok. One thing to be careful is to make sure you don't accidentally download a spyware diguised as a spyware remover. Some software such as Spy Sheriff and Spyware Soft Stop are actually spyware and if you download these, you become infected. So be careful what you download. Spybot Search and Destroy download site:www.spybot.info/ Lavasoft (Adaware) download site:www.lavasoftusa.com/download_and_buy/product_comparison_chart.phpAVG download site:free.grisoft.com/doc/1www.ewido.net/en/download/Note: Schadenfroh has a wonderful thread that lists a whole bunch of anti malware freeware on the Anandtech board. Here is a link to it. From it I decided to download and try out Comodo. forums.anandtech.com/messageview.aspx?catid=40&threadid=2013358&enterthread=yThere are a lot of other freeware and sites such as Majorgeeks has a whole list of freeware and shareware available to the public. (Edit: Links now included in text)
|
|
|
Post by osiris on Mar 7, 2007 16:22:37 GMT 1
Thanks for posting this, Nostra; useful as usual. Can't wait for the links, as i'm probably the most "download - addicted" person here... And i'm always glad to discover new useful PC stuff.
|
|