When it comes to convenience, you can't go past a pre-built PC. Bennett Ring checks out ten of the best systems that money can buy.
No, you haven't accidentally picked up a copy of PCs for Parents -- it's not often that Atomic review pre-built, off-the-shelf systems, but we thought it was about time that we took a look at the biggest, baddest mofos built by a variety of companies. Most Atomic readers prefer to build their own machines and save themselves a bundle of dosh in the process, but when it comes to purchasing or recommending a system for one of your many relatives/friends/neighbours/geek wannabes, it's often much simpler to go for an off-the-shelf PC. And here's why.
For starters, the drama of making sure all of your components are compatible is usually non-existent in a pre-built PC, provided you don't buy it from Dodgy Brothers Incorporated.
It also means that you won't be getting calls at 11pm at night from your Uncle Jake if he accidentally types in 'format c:' at the command prompt. Darn that uncle, darn him to hell.
And when it comes to convenience, it's hard to beat a pre-built system. Simply place your order, and within days a working, fully-specced PC is delivered to your doorstep.
We contacted 25 of the more popular PC builders and asked them to send us the most feature-packed, speed-demon, high-end system that they sell. We imposed absolutely no price ceiling on the entrants, to encourage the builders to go a little crazy. A whopping 15 of these companies chickened out, perhaps due to fears that the Atomic testing procedures would show them to be unworthy of your hard-earned dosh. This left us with ten of the meanest PC systems money can buy. So even before the benchmarking began, we knew that the builders of these beasts had the utmost confidence in their machines.
We also made it clear to the builders that if they were to supply a system for this roundup, it must be available for purchase in exactly the same configuration. So if you decide to buy one of these machines but the manufacturer can't come through with the goods, drop us a line and we'll give 'em hell.
How the hell do we do a PC roundup?We'd never done a PC roundup before, and so had to devise a totally new method of testing each of the submitted PCs. While Ben's suggestion of throwing it off a really high cliff to see which one hit the ground in the shortest time would have looked really cool (mmmm. . . exploding PCs), we finally realised that thanks to that silly Newton chap, each machine would score identically. We ended up settling on a variety of benchmarks and examinations of the PCs construction instead.
The first benchmark to be run was SiSoft Sandra's memory benchmark. What the? Well, this synthetic benchmark proved to be invaluable in seeing which manufacturer had set up the memory most aggressively, as well as seeing who used the best memory modules.
Due to the fact that the memory subsystem is one of the most temperamental and tweakable areas of the PC, this measurement showed which manufacturers knew their BIOS settings back to front, and accounted for 10 per cent of the overall performance score.
The next test was SiSoft Sandra's file system test, which was used to benchmark hard drive performance. Yes, this is another synthetic benchmark, but we wanted to reward manufacturers who had gone to the trouble of setting up high performance RAID setups, or even those who knew how to get the most out of their IDE drives. Weighting for this test was 10 per cent of the overall performance score -- like the memory benchmarks these low weightings represent their lesser value over the remaining benchmarks.
Next up was the most commonly used benchmark this side of Uranus -- MadOnion.com's 3DMark2001SE Pro, patched to version 330. This synthetic test taxes the graphics subsystem most, although the CPU does play its part. Thirty per cent of the overall performance score was derived from this benchmark.
The new Unreal Warfare engine is easily the most visually attractive engine currently available on the PC, and as a result can bring a Ninja PC to its grazed knees. With a little help from [H]ardOCP's UT2003 benchmarking application, we extracted the benchmark results for two UT2003 flyby tests: Citadel at 1024x768 and Inferno at 1280x1024.
These different maps were chosen as each is the most taxing at these different resolutions. The UT2003 botmatch benchmarks were not included due to their inherent inconsistency. The results of the UT2003 benchmark accounted for 30 per cent of the overall performance score.
The final benchmark had to be a real world applications-based productivity benchmark. Yes, these parts of the test were just as boring as they sound. The choice was between PC Mark or BAPCo SYSmark2002, which isn't a hard decision to make. While SYSmark2002 does have a few concerns regarding bias towards the P4 over the Athlon, it is still the only viable productivity applications based benchmark. We would've liked to use SYSmark2001, but this has proven to be very difficult to get running successfully on newer systems. In fact, even the newer SYSmark2002 benchmark proved to be difficult to get to complete on the Hyper-Threaded Pentium 4 systems. Every machine equipped with a Hyper-Threading enabled Pentium 4 simply would not complete SYSmark2002 until we disabled Hyper-Threading via the BIOS. However, Hyper-Threading was left enabled for every other benchmark.
The final performance score, derived from each of these tests, accounted for 50 per cent of the overall score for each PC, as performance was the number one focus of this roundup. What about that other 50 per cent then?
Another 25 per cent of the final score went to build quality. This was broken down into several areas: cabling, airflow/cooling, modifications (Neon, Perspex, blowholes, etc), overall sound level of the PC (a quiet PC is a good PC), and finally Windows setup.
Windows setup included driver installations, and the lack of or presence of clutterware and BIOS settings.
Next up was the feature list, which accounted for the final 25 per cent of the score. We wanted to see the most feature-rich beasts possible, so rewarded those builders who had managed to squeeze as much good stuff in as possible. Features that we were looking for included high-end sound cards, CD or DVD burners, RAID, total hard drive and memory capacity, as well as all the other assorted tidbits that go together to make an Uber PC. Monitors and speakers were not included in the roundup, as we believe Atomic readers will have their own ideas about which of these is suitable for their own unique situations.
Price also factored into the equation, and thanks to some tricky Excel work from the spreadsheet Maestro Tim Dean, we ranked each of the machines based on overall bang for buck.
It must be remembered that the manufacturers of these machines generally threw caution to the wind when installing expensive components, so a PC that didn't do too well in this regard probably isn't the best value for money system that the manufacturer builds.
We were quite surprised to see just how closely these machines were matched, especially considering the wide range of prices. Even the slowest performer wasn't far behind the leader of this high-octane pack. We had originally intended to overclock each of these PCs to see which had the most headroom, but due to the immense amount of time needed to properly overclock ten different PCs, this proved to be an impossible task. Now, let's see what a cool $5,000 or so can get you these days.
Rank guideThe 'Atomic beefiness' score refers to the ranking of each system in terms of the combined ratings for performance, build quality and features. We&q
Issue: 133 | February, 2012