New System Parts Recommendations

This is a draft page used for proposing changes to the the article on before becomes final.
Some rules: If you need to discuss this draft, please do so at talk or contact the person whom create/contributed. You can make changes providing your additional is correct and make sense Correction to any spelling, grammar and punctuation errors are much appreciated welcome.

Seeing as though a lot of people ask about what computer parts they should buy for a range of budgets, I decided to make a Wiki on systems that I think are the most bang for buck. Discussion is encouraged of course, and anyone can edit the parts/prices.

After each system I will give a brief summary of why I chose particular parts over others. You will note that these systems do not include monitors. In that case, say your budget is $1000, look at back one range i.e. $750 and add the 17" BenQ FP73G ($245). This is a cheap but good screen. If you're after a 19" screen, buy the LG 1952T-SF - $318.

You may also notice that none of the systems include extra peripherals like keyboards or mice, or operating systems. If you want a keyboard and mouse, basic kits start at about $30 and go up from there. For an operating system, some will choose to use a Linux distribution (or similar free OS), others may have their own copy of Windows lying around, although a copy of Windows Vista Home Basic (purchased with the system) is about $130 or so. Windows Vista Home Premium is closer to $160 for an OEM copy, and Ultimate is about $260.

Every week I will update the prices (hopefully) and every two weeks I'll adjust the parts if need be. You may note that some of the systems are slightly over - I try to keep within 1% of the nominated price range. Note that many of the later systems are under budget by $50 or so - this is to allow for postage of various items (as you may or may not know, enthusiast parts aren't always readily available at the local PC shop).

Criteria and Part Selection Philosophy

What's Bang for Buck?

It is quite simple to determine if Product A is better value than Product B.

• Read up on the two products - you must read
• Evaluate performance on the two products, obtain a performance margin (e.g. 40 FPS vs 50 FPS) - always obtain the difference and divide that by the lower performing product, then multiply by 100 to get percentage.
• Research price of two products, obtain price margin (e.g. $130 vs $160) - likewise obtain difference and do similarly
• Take both evaluations and compare - the product that gives most performance for the least amount of money is the Best Bang for Buck

CPUs and Graphics Cards

Lets take an example that I started above. Lets say Product A is a graphics card that costs $130 and gives 40 FPS on a particular game. Product B on the other hand is priced at $160, and gives 50 FPS.

• Using the steps above, we look at the performance difference - 10 FPS.
• Divide this by the lower performing product, 40 FPS, which gives us 0.25.
• Multiply this by 100 to get percentage - 25%.

This means that Product B is faster than Product A by 25%. But at what cost?

• $160 - $130 = $30
• $30/$130 = 0.23
• 0.23 x 100 = 23%

Hence, the price premium we pay is 23%. 25% (the performance gained) is higher than 23% (price premium), thus, Product B here is the most Bang for Buck, by a small margin. When this happens, we look at our wallet. Do we really have those extra $30? Of course, if we do spend the extra $$$ we get a slightly better deal, but for the increase in performance it is hardly worth it (in this case). There are other cases where this is not the case, and a small price premium brings much greater performance gains as you will see below in the summaries.

Also to look in a CPU and Graphics card is overclockability. Some people say - hey I don't overclock! Even if you don't overclock, by the time it comes to selling off the old hardware, overclockable hardware always goes for more $$$. Why? It's simple. When done correctly, you can get extra speed for free. So why not? Sometimes the ability of a particular piece of hardware to overclock will make it more Bang for Buck than others - a good example is a 7300GT DDR3. Although it can be beaten at stock by the 7600GS, it can overclock much better than its counterpart (thanks to GDDR3 memory) and thus, though costing around the same it gives better theoretically achievable performance and makes it a better buy.

Motherboards

Arguably the key component of any machine, the criteria for selecting a motherboard is usually quite simple.

• Feature set
• Reliability
• Overclockability

If you're not going to use 8 SATA Ports, then why buy a board that has them? If you're not using RAID, then why buy a board that has it? You may notice that I use a single board, namely the 'Gigabyte 965P-S3' in many of the systems. Why? Because it has enough features for more than the average user, its reliable, and overclocks well. The DS3 version of this board has solid state capacitors. So what? It costs much more ($60+) but the feature set is the same.

Hard Drives

Hard Drives come in a variety of different speeds and sizes. For our purpose, finding the best Dollar per Gigabyte would allow for an economical and high performing computer.

Eg: Seagate 250gb sata 3.5" 8mb cache costs 103 on 31-10-06 at umart.com.au $103 / 250gb = 0.412 $/gb

which is currently the best bang for buck HDD.

Common Comparisons - Which to Buy?

General/Miscellaneous

Future Proofing

The systems I suggest are designed to have some future proofing - AMD CPUs alledgedly will work with AM2 boards, so bye bye Socket 939 or 754. Socket 939 and Pentium D's are still viable upgrade options, but they lead to a dead end. I believe it is in the best interests of buyers to have some sort of upgrade path to head down if required.

Non-Future Proofing

The alternative strategy is to buy the most cost effective solution (for your needs), with little regard to future proofing, and upgrade the whole machine on a shorter time frame (say 9-12 months). To use this stradgy it is best to understand your needs first. If applied properly, you will have whole machines to hand to relatives (or keep as development severs) and still get decent performance. The reasoning is that what good is a AM2 CPU to you if u do not have another motherboard to run it on?

More CPU Mhz vs Faster Graphics Card

If you're gaming, a faster graphics card will impact the performance more than the CPU would. This is primarily because graphics cards are designed to do the graphics work, whilst the CPU is designed to do computations etc. Thus, if there is a bit more room in your budget and you're wondering whether to upgrade the CPU or get a faster graphics card, go with the faster graphics card.

To prove my point, you can try doing this (not for people new to overclocking):

• Run a 3DMark benchmark, note the score
• Overclock your graphics card, run the benchmark, note the score
• Return graphics card to stock, overclock the CPU, run the benchmark, note the score
• By comparing the three scores, the overclocked graphics card score should be higher than the overclocked CPU score.

Note that you don't overclock like nuts to see the effect, even 5-10% is sufficient.

Flagship Products

WIP here but you can read this for the meantime.

"What do you mean? When I bought the cpu, it was the best on the market, AM2 and Conroe came out after! I'd like to see a better graphics card." - dmandn

OK, let's look at difference between X6800 and E6600.

From this Guru3D Review, the X6800 costs 1600 AUD approximately. The E6600 is around 500.

Looking at FEAR running at 1024*768, we have 168 vs 188 FPS. 168/188 = 0.89. There is an 11% difference in performance. Yet you are paying a 200% premium.

In Tom Clancy's Advanced Warfighter there is a mere 1 FPS difference. Again, in Far Cry, we see 124 vs 133 FPS. 124/133 = 0.93 = 7% difference in performance. Again you are paying an extra 200%.

Then again, you could argue that the graphics card was bottlenecking the system. But who the hell needs to play games at 2500x1600 resolution... you'd need a 30" LCD for that.

So basically what I'm trying to say is that buying flagship products means that you pay an unjustified premium for the performance you gain, from the average users perspective.

However, these top-of-the-line CPU's almost always include one important feature: an unlocked multiplier. This means you are no longer restricted to the cheaper CPU's locked FSB multiplier, so your overclocking options just got a whole lot friendlier. Those who seek every last iota of performance, through extreme cooling methods like phase-change, often purchase these flagship products.

CPUs

AMD vs Intel

Currently (Q4 2006) Intel is owning AMD in terms of CPU architecture. Intel's new Core 2 Duo is faster, and requires less power. However it is costly. That said, any AMD system below $750 or so, I believe, will beat any Core 2 Duo system in terms of overall value and performance. Thus, budget conscious buyers should turn to AMD instead of Intel (at this point in time). Also, at the lower end, the AMD offerings are generally superior to the Netburst-based Intels (Celeron, Pentium 4, Pentium D).

E6300 vs E6600

Essentially, 2MB L2 Cache vs 4MB L2 Cache. Read here. Need I say more? If you're gaming, you'll see such a small increase in performance you won't notice, but if you're doing multimedia encoding then go for it. Its damn pricey though $255 vs $485. Personally I'd just stick with the E6300, and go with the E6400 if theres still some room left in your budget.

It's also worth considering the E4300 if you wish to overclock, since it has a 9x multiplier, but only an 800MHz FSB standard.

X2 3600+ vs X2 3800+

Some may believe that with half the cache as the 3800+, the 3600+ is blown away in performance. However, in the reality it is quite the opposite. Tests by X-bit show that the performance of the 3600+ is only around 3-5% behind the 3800+ - hardly noticeable in real-world tasks. However, the price difference is $30 ($125 vs $155) - thus the 3600+ is more bang for buck and is included all the middle to low range systems.

Motherboards

Gigabyte 965P-S3 vs DS3

It is common belief that the DS3 will overclock better than the S3. The only difference between these boards is that the DS3 uses solid state capacitors. This has no effect on overclocking performance. Please read here near the bottom. Also read here, right on OCAU :P. The caps only affect the board's life (bulging caps etc) and being significantly cheaper (S3) than the DS3, the S3 is clearly the more bang for buck option.

Gigabyte 965P-S3 vs ASUS P5B Deluxe vs ASUS P5W DH Deluxe

A $375 board should absolutely obliterate a $185 one right? Wrong. Though the board being compared here is a DS3, as noted above, the DS3 and S3 are exactly the same board bar the solid state capacitors. In the real world tests (the ones that will really affect your experience) there is only about a 2-3% difference in performance, e.g. 134 vs 137 FPS, 549 vs 562 FPS. So how is a 49.33% increase in cost justified? Even the unlocked multipliers going down for the ASUS boards seem to provide little extra performance.

Hard Drives

Hard Drives in RAID and vs Raptors in RAID

A lot of gamers believe that RAID will give them a performance boost in games. Wrong depending on the situation - different games behave differently. As we can see from here, your game loading times may improve by 5 seconds in one game, but 20-30 in another. However, the high cost premium for Raptors probably makes it worth while to stick with standard SATA drives and spend the money elsewhere in the system.

Hard Drives in RAID vs Single Large Drives

Sometimes it is cheaper to buy several smaller drives than two larger ones (e.g. 4 x 250GB vs 2 x 500GB). The opposite can also be true. In this example, the price of one Seagate drives buys you one the space of two for less.

• 2x 80GB SATA HDD $122
• 1x 160GB SATA HDD $79

As we can see here, clearly one 160GB is cheaper than two 80GB drives. Power consumption and noise is also reduced. However, for server machines quick hard drives are a must, and this is where RAID is recommended.

If you're looking for a lot of storage space, you may want to consider the gigabyte per dollar (GB/$) value of your drives. For example, take some more Seagate SATA II drives (MSY prices Jan 2007):

• 80GB SATA II HDD $61.00 (1.31 GB/$)
• 160GB SATA II HDD $79.00 (2.03 GB/$)
• 250GB SATA II HDD $99.00 (2.53 GB/$)
• 320GB SATA II HDD $134.00 (2.39 GB/$)
• 400GB SATA II HDD $185.00 (2.16 GB/$)
• 500GB SATA II HDD $283.00 (1.77 GB/$)
• 750GB SATA II HDD $505.00 (1.49 GB/$)

The sweet spot here is the 250GB drive, so if you want MORE than 250GB, it's cheaper to get a bunch of these drives and RAID 0 them (you should also factor in the cost of a RAID controller and possibly a bigger PSU if you need them). From these simple calculations, it is important to recognise that more can be had for less, and less can be had for more.

16MB Buffer vs 8MB Buffer

As we can see from benchmarks there isn't too much difference between a 16MB and 8MB buffered hard drive. However, the 16MB drives are based on newer technology and should therefore be faster than their predecessors.

SATA 3.0Gb/s vs SATA 1.5GB/s vs PATA

Though the specifications may sound impressive, in reality the SATA interface itself does little to improve performance as shown here compared to the older PATA interface. However, SATA2 and SATA drives often come with more cache (16MB vs 2MB) and this can help boost writing speeds somewhat. Furthermore, the SATA interface uses less power, and the cables used are physically smaller, thus easier to manage. SATA2 drives are designed to be backwards compatible with SATA motherboards, so it is not necessary to purchase a new motherboard to support it.

Additionally, SATA2 drives support NCQ (see below)

NCQ (Native Command Queueing)

See this Wikipedia article.

Graphics Cards

SLI / Crossfire

Some people want to have two graphics cards in their systems. Why? Because it gives them a performance boost. However, not all games support multi-gpus and more importantly you could easily buy a single card later in the future that is not only faster, but probably cheaper and certainly will support new hardware features. Also to note is the running cost and the power requirements for the PSU.

Overclocking / Volt Modding

Overclocking and volt modding are ways to improve the stock performance of a graphics card (much like the CPU) quite easily.

For example, for those keen on voltmodding, a 7900gt 256mb with the reference PCB can be easily voltmodded to a 7900gtx with little risk. Which greatly improves the speed of the card from 450mhz @ 1.2v to 650mhz @ 1.4v which are what the 7900GTX boards run at. Because of the additional heat produced by overclocking and/or volt modding, an after market cooler is recommended for this undertaking.

7300GT GDDR3 vs 7600GS

Though the reference 7300GT specifies DDR2 memory, a few manufacturers such as Palit and Inno3D have instead used GDDR3 memory and increased the clock frequencies on both the core and memory (375Mhz vs 500Mhz Core, 666Mhz vs 1400Mhz Memory). Though the 7300GT GDDR3 has 4 less pixel pipelines, it keeps up with the 7600GS and in most cases beats it by a considerable margin (R76GS meaning Reference, G76GS meaning Galaxy 7600GS - comes with GDDR3 RAM - note that the 7600GS GDDR3 is a boosted version of the 7600GS). The 7600GS can be had for $147 picked up from a shop, whilst the 7300GT DDR3 can be had for around $112 plus postage. Thus the 7300GT DDR3 is a better buy for the money.

7600GS vs X1600 Pro

Again, I draw a comparison based on price. An X1600 Pro can be had for as little as $110 plus shipping. 7600GS, $132. Price difference before shipping, $22 - equating to a 20% price premium. DirectX Benchmark - 27 vs 39 FPS - this is an almost 31% performance increase. In an OpenGL Benchmark the performance difference is the - 21 vs 30 FPS - 30%. Thus, for 30% more performance at only 20% more $$$, the 7600GS is clearly more bang for buck.

7900GS vs 7900GT

The 7900GS is only $289 vs $365-415 for the 7900GT. This makes a 21-30.37% difference in price. However, it is clocked at the same speed as its GT counter-part and benchmarks have proved that there is little difference in performance, at most around 9% in 3DMark. However Real-World tests is where it counts, and as you can see from 3DAvenue's review the 7900GS only lags behind by a few frames. Thus, the 7900GS is a very good buy.

512MB 7950GT vs 256MB X1900XT

These cards are similarly priced, however, as we can see from the benchmarks here, the 256MB X1900XT outperforms the 7950GT (thus blowing away that more onboard RAM = higher performance at high res and AA AF theory). The X1900XT is also cheaper, and therefore is a better buy.

RAM

1GB or 2GB

In game benchmarks, load times don't affect the score and even a bit of lag as something gets loaded to/from disk mid game won't drop the average FPS by much. In reality, games (especially BF2/BF2142) load faster with 2GB and rarely decide to lag out while stuff gets swapped out onto the hard disk. Alt-tabbing in and out of a game to other programs is far smoother.

Video editing with 2GB lets you have a lot more background programs and thumbnails open without hitting the physical memory limit and start swapping to disk. Again, the benchmarks measure encoding speed and 2Gb makes no difference as memory use is only 100-200Mb while doing it.

Given a tight budget, go for a slower CPU and 2GB of RAM rather than a fast CPU and 1GB of RAM. The extra Gb of RAM (for a total of 2Gb) as an upgrade path is reasonable if you're planning on having more money available later.

A bit slower CPU decreases performance by a few percent if at all, but sitting there with memory maxed out and the hard disk thrashing on any computer is a very noticeable performance problem. This is particularly important if you wish to use Windows Vista, which is particularly RAM-hungry.

DDR2

You may ask why I include A-DATA RAM in many of these systems. The answer is simple: they are cheap, yet they overclock well. The 800Mhz Vitesta kit can reach 1000Mhz DDR as shown by DriverHeaven and. techPowerUp!

With the 667Mhz Vitesta kit, RbMods.com and Anandtech have hit past 900Mhz DDR.

And finally, with the 533Mhz Vitesta kit, Anandtech were able to hit 800Mhz DDR.

Complete Systems

Guidelines when updating

Please follow naming conventions, i.e. Size/Capacity/Speed, Brand, Model, Staticice Price

Use conservative Staticice links to widen search results, e.g. Instead of 2GB 667 Kit, 2GB 667 will bring up more results

Allow $50 for postage in each price range, e.g. in the $1500 range, spec up to $1450

Upgrade Kit

The purpose of these upgrade kits are to either extend the life of a current computer or provide a 'quick fix'. The AM2 and 754 systems here will allow you to upgrade later on - this should be considered when deciding. However, if you do own an older system such as a 939 (thus limited to DDR RAM and/or AGP graphics) then there are still cheap options available (compared to buying a complete/near complete system). Note that if you are spending more than $750 you are probably better off buying a completely new system.

Workstation

Depending on on what it's required for, workstations can be relatively simple, low-performance computers (Such as an AMD Sempron system) for use in word processing and small tasks, or high-performance dual-core machines (such as an AMD dual-core Opteron or Intel Core 2 Duo) made to crunch numbers and render complicated details.

HTPC

If you're trying to build a Home Theatre PC, this is for you. (Taking inspiration from here. :)