Can a £300 gaming PC compare to a £3,000 one?
18th Mar 2012 | 08:00
We pit a cheap gaming PC against a high-end rig
Benchmark results are all very well, but can you feel the numbers? When it comes to PC performance and especially gaming grunt, that's the key question.
Of course, everybody knows high-end PC components are piddle-poor value for money. You don't need us to tell you Intel's latest £800 six-core monster, the Core i7 3960X, isn't eight times as good as a £100 quad-core AMD Phenom II.
The same goes for every other component class. Whether it's graphics cards, motherboards or especially storage, your bang for buck plummets horribly at the top of the price scales. None of that, however, is the same as saying high-end clobber doesn't deliver any benefits for the cost.
There's little doubt extreme edition processors and the latest multi-billion transistor GPUs will spew out unholy benchmark numbers. The thing is, we're less sure how much impact that has on subjective gaming pleasure. Putting it another way, if we plonked you down in front of two PCs, one trimmed out with the finest kit known to man, one built down to a modest price point, and fired up your favourite games, would you actually be able to see or feel the difference?
So let's be clear about this. Our task this month isn't merely to prove that spending a shedload on the finest PC components isn't cost effective. It's to find out whether it makes the slightest difference to the way games look and feel.
To do that, we've built two very different systems, focussing on the three core components that influence performance. So that'll be the processor, graphics card and hard drive.
Our money-no-object rig weighs in at nearly £3,000 for those components alone. And our budget-oriented alternative? It's less than £300. Can a system costing just one tenth the price of another really deliver an indistinguishable gaming experience? Time to find out.
First up, let's lay out some ground rules for this intriguing contest. Critically, our main focus is gaming. That doesn't mean we'll be excluding overall system performance entirely. We'll chuck that into the mix for context. But it won't influence the overall result.
However, what will effect the outcome, and something we're contriving to be quite specific about, is the display. The first thing you'll have noticed is that we're not including it in the core set of components. The reasons for this are the frankly enormous array of possible PC monitors and the related issue of user preference.
For most gamers, if money was no object they'd probably run the most powerful possible PC. When it comes to screens, the choice isn't so obvious. Would you go for the largest possible display? Perhaps, but only if the system in question doesn't double as a multi-purpose desktop.
Then again, maybe it's maximum resolution you should be aiming for? In which case, you'll have to compromise on screen size, since the highest resolution monitors are not the largest.
With all that in mind, we've opted for a single control monitor. While there is a niche of gamers who demand the highest possible resolutions, it's debatable how much difference it makes to the visuals when you extend beyond 1080p Full HD.
Moreover, 1080p has become the de facto standard for many games. It's also a resolution that dominates regardless of screen size. The vast majority of PC monitors from 22- to 30-inch and beyond now sport a 1,920 x 1,080 pixel grid. The control screen we've gone for is a 24-inch example.
The make and model are not important. What does matter is that 1080p pixel grid. It puts a limit on the number of pixels any gaming PC is likely to have to pump out.
Immediately, that plays into the hands of the cheaper system. There's little doubt that life would be much harder for it at a resolution like 2,560 x 1,600. But our argument is that even at the high end and regardless of screen size, the most likely resolution for gaming is going to be 1,920 x 1,080. So, that's what we're sticking to.
Taking that logic and running with it a little further, we've removed another image-related stipulation. The two systems are not required to run at the same image quality settings in-game. If you think that tilts things even further in favour of the cheapo rig, you'd be absolutely right.
But we think our reasoning makes sense and that is this comparison isn't about the numbers. It's about the gaming experience - nothing more, nothing less. Our plan is to set the two systems up and put gamers in front of them in a blind comparison test.
So, what matters isn't making sure that the two systems are running exactly the same level of anti-aliasing, anisotropic filtering or shader complexity. What matters is whether gamers can tell the difference. If dialling down the eye candy doesn't make a difference that's noticeable, then that's exactly what we'll do.
However, one thing you should be very clear about is that we won't be compromising the settings on the high-end rig. For each and every game we'll tune it to look as spectacular as physically possible. The challenge will be getting the low-end system right.
As ever, it's a question of playing off image quality and frame rates with the latter being particularly critical. Knock the anti-aliasing down from 8x to 4x and few, if anyone, will notice. Chop the frame rate in half and that's a different matter, especially if the result ever dips below 30 frames per second.
With all that in mind, it's clear that this entire experiment is heavy on the subjectivity. But that, frankly, is the whole point. Like we said, we all know how a benchmark fisticuffs would turn out. And it wouldn't be pretty. What we didn't know going in was exactly how the real world gaming experience compared.
So what are the hardware variables, how did we set the systems up and what games did we go for?
Component wise, our focus is on the three components that have the biggest influence on gaming performance: the processor, the graphics card and the hard drive. Obviously, there's a little more to it than that.
But beyond those three, not only do the performance implications drop off, the price delta shrinks enormously, too. Our reasoning was to allow roughly £100 each for the low end rig. We didn't put a price limit on the top-end monster, but in the end the average wasn't a millions away from a nicely symmetrical £1,000 per component.
With our systems specified, built and saddled up with a fresh copy of Windows 7, the next task was game title selection.
This is a subject ripe for hang-ups, so we elected to go with some very straightforward criteria. Elder Scrolls V: Skyrim made the cut because it's a super hot title, it looks great, it's fairly demanding and it ticks the large-levels and long-draw distance boxes.
Next up is Crysis 2. Okay, as a game to play it's pretty crappy. And its execution reeks of console-port compromises. But in many ways its graphical fidelity can be regarded as the benchmark. In parts it's absolutely stunning.
Our final candidate is DiRT 3. We wanted a driving game in the mix and there's no doubting DiRT 3 is a looker. Granted, this trio is hardly comprehensive. But obvious omissions such as, perhaps an MMORPG such as WOW don't tend to be hugely performance intensive. They're coded to support very broad user bases. So, while there will always be a few exceptions, we reckon our chosen trio gives a pretty decent picture of overall gaming performance.
Tuning the tests
Of course, choosing games is only half the solution. We next had to select image quality settings. From the outset we'd decided that running at 1080p native resolution was essential, so that bit was easy.
Configuring the high-end system was very straight forward, too. The system has performance to burn so it was simply a case of maxing everything out and then jumping in-game to make sure everything was running smoothly.
Like we said, we didn't want our budget rig getting an easy ride. It would have to compete with the very highest possible image quality settings currently available. Indeed, tuning up the cheapo machine was intriguing.
Going in, we'd expected that it would be a finely balanced process of playing off image quality settings and frame rates. One thing we were certain about was that smooth frame rates were essential. Any keen gamer is going to pick up on jerky, unresponsive performance.
Image quality, on the other hand, is a different matter. How many people, for instance, can tell the difference between 16 times anisotropic filter and eight times filtering? Likewise, for any given game, does knocking settings such as shader or shadow quality down make a big difference to perceived image quality?
We were in for a long haul. Or so we thought. It transpired, it was much easier for the simple reason that it handled the highest image quality settings much better than we expected. One reason for this is the undeniable influence that console games have on game engines. The Xbox 360 and PS3 are positively ancient but retain an iron grip over game development. That's lead to stagnation.
The bottom line is that we needed to do relatively little to get our affordable PC running smoothly. Mostly it was a case of dropping the anti-aliasing from 8x to 4x (though Crysis 2 doesn't actually offer proper AA, which is probably a big help) and tweaking just a handful of further settings. For the most part, both rigs ran with ultra settings across the board.
The final part of the puzzle is the actual blind comparison. The key here was to keep the identities of the systems and the components included completely concealed.
In fact, we didn't even want our testers knowing what an enormous gulf in hardware specification they were experiencing. Instead, all our testers knew was that they were comparing the gaming experience on two different systems.
We also wanted to get impressions of different aspects of gaming performance including in-game frame rates, graphical quality and game load and level load times. With that in mind, the test went something like this.
Both machines were up and running at the same time on the same desk and via the same 24-inch high quality 1080p monitor. Switching between the two was a matter of the few seconds required to move the DVI cable across, allowing an almost instant comparison.
Our first test involved playing each title from an identical pre-loaded and paused game state without observing level loads or desktop performance. We wanted an unbiased assessment of the in-game experience, something that can be coloured by sitting through a lengthy level load.
Our testers were allowed around ten minutes per title to jump back and forth between machines, as often as they liked. For this comparison we didn't make any specific demands, we simply asked the testers how they thought the two systems compared.
Next was a test starting from the desktop and involving loading, first, the main game interface and then a game level, after which further game play and general impressions were taken. as well as adding storage performance into the mix, we also asked our testers more specific questions regarding performance and image quality.
Which, if either, of the two systems had better graphics? Did one or the other deliver a noticeably smoother frame rate or superior response to mouse and keyboard inputs?
Finally, we allowed our testers a little time to try out basic PC usage to get a feel for desktop performance: firing up a few apps, surfing the web and watching some high definition video.
Among all this there's a commonly encountered and gaming-specific issue that isn't covered and that's installation. There's absolutely no question the high-end machine with its ultra-fast SSD storage was much quicker during the installations. Without a doubt, this is something our testers would have noticed.
All of the above, of course, concerns subjective experiences. When you get right down to it, that's all that actually matters in the real world. If it feels the same, who cares if it runs 10 times faster?
That said, we also ran a suite of benchmarks to provide context and to underline the significance or insignificance of the comparative numbers. If our testers couldn't feel the performance difference, we looked at whether that reflected in the numbers. Likewise, if the gap in system performance was big enough to be picked up subjectively, was that also reflected in the benchmark numbers?
So there you have it. Two systems. one for those with money to burn. Another for those on a tight budget. Tested by keen gamers and as you'll see, the results are frankly astonishing.
Three thousand pounds has rarely looked so humble
This ladies and gentlemen, is what dreams are made of. You may have thought that this much cash would get you a computer core a little more sexy looking, but we're not interested in basic looks here, oh no, the beauty of this rig is very much on the inside.
The main components we've considered vital to the gaming experience are the graphics card, processor and SSD. The other components we've used could easily be replaced with similar models and aren't quite so vital.
We've gone into a little detail on those other components purely for system completeness. It's all decent kit, just not essential for the testing.
Intel Core i7 3960X
Intel hasn't seen a lot of competition at the high-end recently. This has left us feeling underwhelmed by this, it's latest processor. That doesn't take too much away from the fact that it's the fastest darn slice of silicon money can buy though and here there's no other choice.
That big chunk of notes gets you six physical cores to play with, capable of handling 12 threads concurrently. The basic frequency clocks in at 3.33GHz, capable of hitting 3.9GHz in Turbo mode. You also get a healthy 15MB of L3 cache and support for quad-channel DDR3 RAM.
AsRock X79 Extreme4
This may not be the best example of the plethora of X79 motherboards around but it just goes to show that you don't need the most expensive board to get serious performance out of your pricey CPU purchase.
That said it won't give you the same sort of overclocking performance as the Asus pairing of RoG Rampage and Sabertooth X79 boards, but when you're spending that much on a chip do you want to shorten its life by waving the overclocking stick at it? This sub-£200 board then is a decent partner and home for your Sandy Bridge E processor.
AMD Radeon HD 7970
Right now this is the fastest single-GPU graphics card on the market bar none. In fact it actually gives the dual-GPU cards a run for their money too. It may be a pricey ol' beast but it almost has the performance chops to make you forget about the price. It also does some smart power play too, keeping power draw low in down time.
But while it is definitively the fastest card right now, the performance lead over the last generation isn't great enough for us to think it's going to still be the fastest once Nvidia gets its Kepler cards out on the shelves.
Corsair Vengeance 1,866MHz
The jury is still out on how important an advantage the quad-channel memory of the X79 platform is, but if bandwidth is your bag then this awesome Corsair kit is a definite winning pack.
Rated at 1,866MHz out of the box, the Corsair kit is a doddle to set up, mainly thanks to the only real benefit of the latest quad-channel RAM kits. That's the latest iteration of the XMP initiative and means you don't have to go through the minutiae of your board's BIOS in order to make sure you're getting the most out of each of your modules. The XMP 1.3 technology then is a God-send.
OCZ RevoDrive 3 X2
Storage technology has had a major speed bump over the last few years thanks to the advances in SSDs. Your straight SATA-based drives are quick enough, but if you want crazy speeds then a PCIe-based device like the RevoDrive is the only way to go.
OCZ's current RevoDrive uses the latest controllers from SandForce to create the ultimate in desktop storage. At 480GB it's also easily big enough to use as the main boot drive in your machine without having to worry about what applications or games you have installed on it. And boy does it make your PC react quickly.
Corsair AX 1200
Corsair has moved into a whole host of new markets, most recently with its excellent new peripherals. But its old school areas of expertise, that of memory and PSUs, are still very much in evidence.
The AX 1200 is one of the finest PSUs we've ever tested. It's a little chunky but is still incredibly efficient and displays minimal electrical interference at either 75 or 100 per cent loads. It can also be quite loud compared with other 1,000 watt (and above) power supplies. This is a supply yearning for a multi-GPU setup, but has enough efficiency for any load.
We may be talking last gen but this rig still packs a punch
This machine is obviously not as technically-gifted as the previous system, but then its core components also cost around the same sort of price as the single graphics card we've opted for in Rig One. Should you buy the makings of a full PC or a Radeon HD 7970?
Again, the main focus here is the graphics card, processor and storage, in terms of gaming. The quad-core AMD chip at its heart will deliver a healthy chunk of gaming performance and when paired with a £100 HD 6850 you're looking at impressive pixel-pushing chops.
The HDD may seem a little retro, but it doesn't hold things back too much.
AMD Phenom II X4 960T Black Edition
As much as AMD wants you to believe that its latest Bulldozer architecture is the way forward, realistically an eight-core Phenom would have it beaten. The old Phenom II then, though not world-beating, is still a decent bargain quad and can be encouraged to hit 4GHz.
You could pick up a Sandy Bridge CPU for around the same money, but then you're looking at a dual-core CPU with zero over-clocking potential. This Phenom II X4 965 is a fully-fledged quad-core and so gives a decent balance between gaming pedigree and general computing.
This AMD 970-based motherboard from Gigabyte is one of the few concessions to future-proofing that we allowed. It may not be the top-end 990 platform, but the 970 is still a capable platform and has that AM3+ socket, which means that Bulldozer or Piledriver will fit the board when it's time to upgrade.
But until then it's still a great home for the Phenom II chip and will also allow for an X6 upgrade too should you feel the need for those extra two cores. It's DDR3 compatible too, though lacking SATA 6Gbps but it does offer a decent shot at overclocking.
AMD Radeon HD 6850
We aren't too proud to change our opinions on a product if subsequent changes warrant it. And AMD's HD 6850 was a card we took an immediate dislike to originally, but it's fast become a budget favourite.
When it was first released it wasn't that fast for AMD's asking price. At around £100 though you're getting a rather powerful little graphics card, capable of hurling polygons around a medium resolution screen at quite a rate. In fact, it can even do a job at full 1,920 x 1,080 resolutions too. The latest driver releases have also given it a noticeable speed boost.
G.Skill Trident DDR3
There's a reason that only the X58 and X79 platforms use triple or quad-channel memory and that's because as essentially server components they rely on bandwidth. For the rest of us the old school dual-channel DDR3 is all the memory we need.
G.Skill has produced some of our favourite memory over the last couple of years with the RipJawsX kit being our current favourite thanks to its hefty overclocking chops. This Trident kit is cheaper and already comes rated at 2,000MHz. And we all know what faster memory means, just a plain better PC experience.
Hitachi Deskstar 7K1000
The Hitachi Deskstar was the first 1TB HDD that we ever saw in the PCF offices, what a momentous day that was. Despite the 3TB mark being available in other drives, it's no longer an issue of capacity with storage, now it's all about speed. Or is it?
As much as having your PC turn on in 30 seconds is eminently desirable, a decently capacious SSD is frighteningly expensive. The SSD on the previous pages was half the size for around 17 times the price, for example. If you're on a tight budget, and have a little patience in your soul, then the humble hard drive will still do the job.
Cooler Master Silent Pro 800W
As much as you might baulk at paying £100 for a power supply, if you drop a ton on this PSU you'll find yourself the proud owner of one of the finest power supplies on the market. At 800W it's far more than this system needs, but that's not a problem for the Cooler Master supply as it's still very efficient even at the lower end of its load spectrum.
The unit's Silent Pro moniker isn't just for show either, it's easily one of the quietest PSUs we've ever tested. Looking down at the more technical tests too it comes out tops with almost zero interference when running.
The truth about gaming hardware
Four lab rats, two very different PCs and one astonishing outcome
Setting out on this sort of feature, the outcome is always uncertain. Obviously, we had an inkling that the wallet-pillaging cost of high-end components outweighed the real world performance advantage they bequeath. But £300 versus £3,000? That was really pushing our luck, surely?
What's more, even the best possible result involves a generous helping of humble pie. After all, we've already been harsh on AMD's inability to keep with Intel's CPU performance. If it turns out that £100 worth of previous generation AMD Phenom II chippery can mix it with Intel's latest 32nm, six-core master work, well, let's just say we'll need to do a little in-house recalibration.
With all that in mind, and the results in, the outcome was even more dramatic than we could possibly have imagined. Where, then, to begin this exposition of the unbelievable but undeniable results about PC gaming hardware; this dissertation on the implausible but now proven? Let's start with a quick recap of how we set the test up and who got involved.
The basic idea is to compare the gaming performance of something affordable with something impossibly exotic in purely subjective terms. It's not about benchmark results; it's about what it feels and looks like to play the game.
In super simple terms, can you tell the difference between the preposterously pricey and the relatively parsimonious? Thus, our real-world rig has an AMD Phenom II X4 960T processor, an AMD Radeon HD 6850 graphics and a 1TB magnetic hard disk courtesy of Hitachi.
Meanwhile, the tippy top system sports Intel's Core i7 3960X, AMD's Radeon HD 7970 and an OCZ RevoDrive 3 X2 480GB solid state drive. Games-wise, Elder Scrolls V: Skyrim, Crysis 2 and DiRT 3 are our on-monitor muses.
While we're on the subject of screens, the 24-inch Philips Brilliance 241P4QPYES (yes, it's a silly name) did the display duties. While it's undoubtedly an awfully nice panel thanks to AMVA LCD technology, its most significant attribute is a 1,920 x 1,080 pixel grid. That's pretty much the default resolution for modern gaming and the setting we chose for both systems.
As for image quality settings, the idea was to max out the eye candy on the high-end rig, to set the bar as high as possible and see if the cheaper, more attainable system - the one most gamers can at least aspire to own - could live with it.
Then we plopped a quartet of gamers with a broad spectrum of experience - from the casual to the professional - in front of our systems and let rip.
No hard feelings
Splitting out the influence of the three major components - CPU, graphics and storage - isn't a trivial job, especially when we're dealing with subjective experiences, not cold, hard numbers. Nevertheless, what we can say is that the issue of storage or hard drive performance is least critical to gaming performance, both subjectively and by the numbers.
In our game-level load tests, the bulk-storage 1TB Hitachi Deskstar was a long way off the pace of OCZ's exotic RevoDrive 3 X2 in percentage terms. In both Skyrim and Crysis 2, OCZ's expensive PCI Express SSD delivers load times roughly twice as rapid as the Deskstar and its portly magnetic platters.
But when you're talking seven seconds versus 15 seconds or 10 versus 20, well, how much does a few seconds really matter? Certainly, none of our test subjects rumbled the true identities of our rigs based on storage performance.
Admittedly, that was partly because the tendency was to obsess over image quality and frame rates once the game had loaded. We also made sure that each test subject's first taste of both systems didn't involve experiencing level loads as we wanted to canvas their unbiased impressions of actual gameplay.
Once they'd experienced that, focussing attentions on level loads was a little tricky. What's more, the biggest performance difference involving storage and games was something none of them got a feel for, namely installation.
The OCZ RevoDrive absolutely tore through game installs using a backup of PC Format's Steam library. The Hitachi Deskstar was a real chore. And that despite the fact that our Steam backup is itself housed on a conventional magnetic hard disk.
It's also true that both our systems were based on fresh installations of Windows 7 and therefore optimal, non-fragmented drive performance. Six months of a routine daily disk abuse would no doubt make the gap big enough to begin to impinge on gaming joy. But then, a fresh reinstall a couple of times a year is a pretty minor inconvenience.
Go go graphics
With storage sidelined, we're left with CPUs and GPUs. Traditionally, we've always put the emphasis on graphics when it comes to gaming grunt. And this experiment has done nothing to change our belief that the surest way to bork your frame rates is a crap graphics card.
Meanwhile, we've long known that the most expensive CPUs are overkill for games. What we didn't anticipate was how little difference a high-end GPU makes. At this point we've little choice but to reveal the overall result.
Most of the time, most of our subjects literally could not tell the difference. In fact, it was worse than that. It wasn't just a case of putting hands up in the air and declaring it a dead heat. Frequently, our testers actually got it the wrong way round, declaring the low-end rig to be rendering with greater fidelity or actually running faster.
We're a courteous bunch, so we'll save blushes by not naming names. However, we will reveal that expectations played a big part in perceptions. Much more so than the background or experience of the tester.
One of our more experienced testers was left almost completely in the dark. All he was told was that we had two setups to compare. We didn't say whether the hardware was different or whether it was just settings. We just asked him to play through a level in each game on both rigs and tell us how they compared.
The result was an intriguing narrative on things that weren't there - differences in lighting, anti-aliasing and texture quality settings that didn't exist. Needless to say, when the truth of the settings and the hardware were revealed, a pair of socks went sailing across the office. That's because the image quality settings on both systems we're very similar indeed.
For Crysis 2 we had to knock a few of the advanced IQ options down a notch. For Skyrim, it was a case of winding anti-aliasing back from 8x to 4x. And for DiRT3 it was even simpler - both systems ran at maximum detail.
There are plenty of caveats to the result of our experiment, but if you take away anything from this feature, it should be that budget PCs are not the second rate gaming systems they once were. Run a budget rig at the right settings, and you'll produce a quality experience. S
pend your money wisely, and you'll reap the rewards. At the same time, don't be too disparaging of high-end rigs - they definitely have their place. We're thinking particularly when it comes to high-resolution displays.