Graphics Cards: New Test Procedure

We've decided to update the test procedure we use to review graphics cards. From now on, we'll be testing each model with a selection of new games and we've worked out a new way of deciding a card's overall score.

Our graphics card test procedure hadn't been updated since 2009. Some of the games we were using were starting to show their age, and our framerate readings in particular tended to plateau at a certain level. We've therefore replaced these games with four new ones, which are more power hungry—requiring more GPU processing power to run—but which are also interesting from a technical perspective. These new games are compatible with DirectX 11, which is a new field of competition in the graphics card market.

We have, however, kept four of the games from the original test procedure. This means that cards tested under the new procedure can still be compared with cards previously tested on the site. This can be a handy way of telling whether or not it's worth upgrading your graphics card.

Furthermore, we'll be getting rid of the 1280 x 1024 resolution test. From now on we'll only be using 1680 x 1050 and 1920 x 1200 resolutions, as these are sufficient for differentiating between cards.

Two Power Consumption Tests

We've decided to add a second power consumption reading to our reviews. Previously, we had just been using a graphics card stress test tool. However, AMD and NVIDIA have introduced systems that reduce the speed of their graphics cards when this type of program is detected.

We've always been able to find ways of getting around this, but from now on, we'll be double-checking things with a power reading taken while gaming in a real-life situation. As well as getting round the manufacturers' clever test-detection systems, this also increases the processor's power consumption, making it a more effective way of comparing and contrasting cards in real terms.

A New 3D Scoring System

Another thing we've changed is the way we rate 3D performance. The score is now automatically calculated in relation to the results obtained by the fastest card we've reviewed to date. This automatic system means that the 3D score for each graphics card—as well as its overall score—will be automatically recalculated when we review a new super-fast model (which in turn becomes our new reference card). A five-star product therefore won't necessarily hang onto its five stars forever!

The final score is calculated with a certain weighting: the 3D rating counts for ¾ of the overall score while the remaining third is split 50:50 between the power consumption and noise output.

Performance Tables in Each Review

The last thing we've changed is the addition of a drop-down performance index table to each graphics card review. As we give each card an average overall score, models can be easily compared and contrasted at a glance in our performance index tables. The full selection of performance index tables is available on a separate page of the site, with results for tests at different resolutions and with and without texture filters.

> Graphics Card Reviews
> Graphics Cards: Face-Off

> Graphics Cards: Performance Index Tables
> Graphics Cards: Test Procedure and Hardware

> Buyer's Guides: Our Pick of The Best Products