F.E.A.R. makes its cinematic pretensions clear from the start. As soon as the credits roll, and the music starts, you are treated to the full works. The camera pans across scores of troops locked 'n' loaded and ready to hunt you down, all seemingly linked to 'Paxton Fettel', a strange kind of guy with extraordinary psychic power capable of controlling battalions of soldiers and a habit of feeding off any poor unfortunate innocents - presumably to aid his powers of concentration. It doesn’t end there, after a short briefing at F.E.A.R. HQ you are sent off to hunt down Fettel equipped with reflexes that are 'off the chart'. These reflexes are put to excellent use, with a slow motion effects like that of Max Payne, or the before mentioned Matrix. But here, it is oooohhhh so much more satisfying thanks to the outstanding environmental effects. Sparks fly everywhere, as chunks of masonry are blasted from the walls and blood splatters from your latest victim. The physics are just great, with boxes sent flying, shelves tipped over, and objects hurtling towards your head. And the explosions, well, the explosions just have to be seen, and what's so great about this is you can witness it in all its glory in slow motion.

As always with memory and CPU tests, we include a game or two.

For FEAR we opted to use a a low resolution as a high resolution would be GPU bound, meaning the graphics card would limit the results. It's pretty amazing to see that the results throughout the scope are extremely small and close to NIL. For games to show a significant difference, they need to be CPU limited. See, if you can free up that CPU processing power with faster memory bandwidth, it'll show an incremental difference.

However, here's an issue ... fact is, the CPU (Core i7 965) is so fast, that the GPU is pretty much maxed out even with older titles. And thus even at 1024x768 a GPU bottleneck kicks in due to the extremely fast processor.

We do see some difference, but granted they are marginal at best. Let's try another title.

Crysis Warhead DX10

As in last year's game, expect to encounter dense jungle environments, barren ice fields, Korean soldiers and plenty of flying aliens. There's no denying that this is more of the same, except here it's a more tightly woven experience with a little less freedom to explore.

With a top-end PC (although Warhead has supposedly benefited from an improved game engine, you'll still need a fairly beefy system) rest assured, developer Crytek has enhanced more than just the graphics engine.

Vehicles are more fun to drive, firefights are more intense and focused, and aliens do more than just float around you. More emphasis on the open-ended environments would have been welcome, but a more exciting (though shorter) campaign, a new multiplayer mode, and a whole bunch of new maps make Crysis Warhead an excellent expansion to one of last year's best shooters.

Crysis Warhead has good looks. As mentioned before, the game looks better than Crysis, and it runs better too. Our test machine that struggled a bit to run the original at high settings ran Warhead smoothly with the same settings. Yet as much as you may have heard about Crysis' technical prowess, you'll still be impressed when you feast your eyes on the swaying vegetation, surging water, and expressive animations. Outstanding graphics. Couldn't say more here.

We up the ante a little more by enabling DX10. Though we really wanted to push 4x AA here. Current day graphics cards yet again run out of memory, in the highest resolutions, somewhere where this 6GB kit would definitely help.

• Level Ambush
• Codepath DX10
• Anti-Aliasing 2x MSAA
• In game Quality mode Gamer

As stated in the FEAR test. There is a bit the problem with game / memory reviews.

Memory differences show really well in situations where you are not GPU bottlenecked. Even with the GeForce GTX 285 that's the case. So if you look at 1024x768, there's definitely something to say for faster frequency memory. But once we pass 1280x1024 .. the difference vanishes into thin air and become too small to measure.

And for this test, again triple versus dual-channel configuration. Very little difference really.