Performance in Synthetic Benchmarks
First let’s take a look at a couple of synthetic tests. This will show us what to expect from the GeForce 7950 GX2 in games.
The GeForce 7950 GX2 delivers a much higher scene fill rate than the Radeon X1900 XTX due to the difference in their TMU count: 48 against 16. The graphics cores of the GeForce 7950 GX2 are clocked at a slightly higher frequency than the cores of the GeForce 7900 GT SLI, the fill rate they provide is higher in all cases. The new graphics card from Nvidia isn’t going to lack texturing speed even where the GeForce 7900 GTX used to fail, i.e. in applications that use large amounts of high-resolution textures or shaders with multiple texture lookups.
The GeForce 7950 GX2 also boast a tremendously high pixel shader performance, especially when it comes to executing simple shaders – it may deliver two times the performance of a Radeon X1900 XTX despite the same number of pixel processors (because the GeForce has more texture-mapping units). The gap isn’t that big when complex shaders are executed, especially those that render per-pixel lighting, yet it is large enough for the GeForce 7950 GX2 to claim superiority at processing shaders. Well, real-life games vary in the total number and complexity of shaders they use, so the overall picture of performance will be individual for each particular game.
Our Xbitmark test measures pixel shader performance, too, and its main advantage is in using various kinds of shaders, from almost purely math1ematic ones to shaders that are actively working with textures.
Here, the GeForce 7950 GX2 isn’t always ahead of the Radeon X1900 XTX: particularly, it is slower when there’s a lot of math1ematics to be done and on shaders with dynamic branches. But when high texturing speed is crucial, like in the NPR (hatch) shaders, the new solution from Nvidia has no opponents at all.
So, here’s what the GeForce 7950 GX2 can do in synthetic tests. Let’s see if it is as good in real-life games.