Labs

Graphics cards

As always, we use real-world games to test graphics cards rather than synthetic benchmarks, because they provide far more meaningful results that are simple to interpret.

For this Labs, we've updated our3D benchmarks accordingly, choosing two new games that are both state-of-the-art and intensive, yet scalable enough to be valid at all budget and performance levels.

No current game pushes graphics cards harder than the staggeringly beautiful first-person shooter Crysis (web ID: 145212). Held up by Microsoft during Vista's promotion as the poster boy for DirectX 10, its release proved that developer Crytek had succeeded in creating the most realistic and life-like game environment to date.

Real-time physics and lighting effects, along with a set of top-quality settings that remain a huge challenge for even the most expensive cards, make it a 3D benchmark that will remain valid for some time.

To create the test, we first applied the official 1.1 patch, which greatly improves performance, then simply recordeda realistic and accurately repeatable timedemo, using the lagoon stage from the gorgeous Island level.

Crysis has a selection of quality effects and settings that can be set to Low, Medium, High or Very High, depending on your system.

We start the budget cards on Low settings - which are basic enough to make the game look little more thana slightly updated clone of Far Cry, Crytek's previous release - anda resolution of 1,024 x 768.

Cards that achieve playable frame rates are bumped up to the far more immersive Medium settings, and then to a 1,280 x 1,024 resolution, which makes up our Medium test.

Those cards that achieve 30fps or more here are then tested on High settings, before we finally move to 1,600 x 1,200 - a combination which makes up our High test.

We're yet to see a card that can average a smooth 30fps at Very High, even at 1,280 x 1,024, so that settingis kept as an extreme test for SLI and CrossFire setups. Turn to p102 fora look at just how these technologies are progressing.

We've been using Call of Duty 2 asa 3D benchmark for some time, so the chart-topping success of Call of Duty 4 made it an easy choice for our second benchmark. It's less intensive than Crysis, but uses a range of lighting effects and textures to ensure it remains a challenge at higher settings.

We use a pre-recorded segment of the second level, using FRAPS to work out an average frame rate. Again, we use 1,024 x 768, 1,280 x 1,024 and 1,600 x 1,200 resolutions as our Low, Medium and High benchmarks. The settings aren't as simple as Crysis, but for each we use a mixture of the game's Low and High settings to most appropriately match the intensity of the chosen resolution.

To test each card's maximum performance, we build a test rig that eliminates all other bottlenecks from the system. To this end, we use an MSI X48 Platinum motherboard, which employs the latest Intel X48 chipset for the ultimate in CPU compatibility.

We fit a 3GHz Intel Core 2 Quad Extreme QX9650 processor and 4GB of Corsair DDR3 RAM. To ensure the hard disk doesn't slow things down, we use a 36GB, 10,000rpm Western Digital Raptor with a SATA interface.

How we work out the ratings

At the bottom of each review, you'll see three star ratings: one for Performance, another for Value for Money and a final Overall score.

The Performance score is calculated by summing the average frame rates achieved by each card in both games, using the Medium tests as the main indicator. The result is then converted into a score out of six, which is comparable across the whole group.

The Value for Money rating is based on the relationship between the performance and price scores to givea bang-per-buck result.

The Overall score is simply a straight average of the other two scores. Rather than give a Labs Winner award toone card this month, we've given Recommended awards to thosewe feel deserve the accolade.