Quick Game Look

As a lead in to the full set of results, we grabbed results for the HD 2600 XT versus 8600 GTS in Company of Heroes under Vista x64, with the D3D10 renderer, to see how performance stacked up in a real-world title. We use the built-in performance test.

Performance is roughly one third of HD 2900 XT (and that board bests GeForce 8800 GTX, as a sneak peak at our upcoming R600 performance article), and at least 2x the performance of the GeForce 8600 GTS, under Vista x64. NVIDIA have some Vista x64 driver work to do, given the hardware is more closely matched under Vista x86 (although it's still slower).

General Performance

We managed to hit near instruction issuing peak with the same scalar shaders as we used to evaluate HD 2900 XT on launch, indicating a decent level of out-of-the-box shader assembler maturity when handling obvious cases. Data sampling rates are roughly expected on HD 2600 XT, but a little lower than the on-paper spec would suggest with HD 2400 XT.

We note that with Supreme Commander with 4x AA on, HD 2600 XT has performance around half HD 2900 XT, and significantly better than GeForce 8600 GTS again. However with S.T.A.L.K.E.R., HD 2600 XT is slower than GeForce 8600 GTS by some distance, 4x AA or not. We'll flesh out the game performance testing data when able.

Image Quality

Cursory investigation shows RV610 and RV630 to have the same achievable potential image quality as R600 when it comes to game and theoretical testing with antialiasing and surface filtering. Using the same driver and settings, identical images were obtained in our usual IQ theoreticals, hinting that performance is what truly separates AMD's current discrete graphics products.