New RightMark3D 2.0 includes two old Direct3D 9 PS 3.0 tests, rewritten for DirectX 10, and two brand new tests. The first two tests can now enable self-shadowing and shader supersampling, which increase their load on GPUs.
These tests measure efficiency of executing looped pixel shaders with a lot of texture lookups (up to several hundreds of lookups per pixel in the heaviest mode!) and a relatively low ALU load. In other words, they measure a texture sampling rate and branching efficiency in a pixel shader.
The first pixel shader test will be the Fur test. When used with the lowest settings, it uses 15-30 texture lookups from bump maps and two lookups from the main texture. The High Effect Detail mode increases the number of lookups to 40-80. When shader supersampling is enabled - the number of lookups grows to 60-120. And the High mode with SSAA is the heaviest mode - 160-320 lookups from a bump map.
Let's see what happens in modes without supersampling - they are relatively simple, and the correlation of results in Low/High modes must be similar.
All results in the High mode are approximately 1.5 times as low as in the Low mode. Procedural fur tests with lots of texture lookups traditionally demonstrate a big advantage of NVIDIA solutions over AMD ones. None of RADEON cards can vie with both GeForce 8800 products. Theoretically, this lag shouldn't have existed, I don't understand where this difference comes from...
The RADEON HD 2600 XT is much slower than the top solutions, because is has fewer execution units. Interestingly, the RV670-based graphics cards demonstrate a tad better results versus the top R600, adjusted for the difference in clock rates. The HD 3850 performs on a par with the HD 2900 XT, and the HD 3870 is faster by 10-15%, although the frequency difference is smaller. Let's have a look at the results in this test with enabled shader supersampling, which quadruples the load. Perhaps it will change the situation:
Theoretically, supersampling quadruples the load. But its slows NVIDIA graphics cards down more than AMD R6xx cards. That's why the latter cards come closer to the cards based on G80 and G92. Only the GeForce 8800 can cope with such test complexity. All other cards demonstrate low results, and the RADEON HD 2600 XT is outperformed by the RV670/R600-based cards more than twofold. Nothing changes in the comparison of the HD 3800 and the HD 2900, the new cards are faster than the previous top solution, even though they operate at almost identical frequencies.
The second test that measures efficiency of executing complex looped pixel shaders with many texture lookups is called Steep Parallax Mapping. With low settings it uses 10-50 texture lookups from a bump map and three lookups from main textures. The heavy mode with self-shadowing doubles the number of texture lookups, and supersampling quadruples this number. The most complex test mode with supersampling and self-shadowing uses 80-400 texture lookups, that is eight times as many as in the low mode. Let's analyze simple modes without supersampling first:
This test is more interesting from the practical point of view, because parallax mapping methods have been used in games for a long time already. And heavy modifications, like our steep parallax mapping, have recently appeared in games, e.g. in Lost Planet and Crysis. Along with supersampling, this test can enable self-shadowing that doubles the GPU load (High mode.)
Although AMD solutions were previously strong in our Direct3D 9 tests of parallax mapping, they cannot cope with the updated DX10 test without supersampling faster than the GeForce 8800 GT and GTS. Besides, self-shadowing causes a bigger performance drop in AMD products, over two times versus 1.5 in NVIDIA solutions.
The HD 3850 is slightly slower in the Low mode than the HD 2900 XT, and it outperforms the latter in the High mode. The RADEON HD 3870 is already faster than these cards, even though it's still much slower than the GeForce 8800 GT. On the other hand, the previous Mid-End solution from AMD drags behind, being twice as slow as the other cards. Let's see what supersampling will change - it caused bigger performance drops on NVIDIA cards in the previous test.
Another heavy task for GPUs. What both options are enabled, the load grows eightfold resulting in serious performance drops. Performance differences between our graphics cards remain. However, when supersampling is enabled, the R6xx-based cards improve their results versus NVIDIA, just like in the previous case. Unfortunately, the RV670 only comes close to the G92, being slower. What concerns the comparison between the HD 3800 and the HD 2900, the RV670-based cards demonstrate better results than the top R600-based card as the load grows.
Direct3D 10: PS 4.0 Tests (computing)
The next couple of pixel shader tests contains very few texture lookups to minimize the effect of TMUs on performance. They use a lot of arithmetic operations, so they measure arithmetic performance of GPUs, how fast they execute arithmetic instructions in pixel shaders.
The first computing test is called Mineral. It's a complex procedural texturing test, which uses only two texture lookups and 65 sin and cos instructions.
The R600 fares well in complex arithmetic tasks, outperforming competing solutions from NVIDIA. The RADEON HD 2900 XT is apparently stronger than any other card in this test, G80/G92-based solutions are slower. Performance in this test seems to be affected much by memory bandwidth, because the HD 2900 XT is 1.5 times as fast as the HD 3870. This result cannot be explained with other reasons. The RV670 hardly executes arithmetic instructions less efficiently.
The new RV670-based cards perform almost on a par with the GeForce 8800 GT, the GeForce 8800 GTS is a tad slower. Performance of the RADEON HD 2600 XT is more than twice as low as that of the RV670-based cards, this ratio is not demonstrated for the first time.
The second shader test is called Fire, it's even harder for ALUs. It contains only a single texture lookup, while the number of sin/cos instructions is doubled to 130. Let's see what changes as the load grows:
Nothing has changed in this test for AMD with the appearance of the RADEON HD 3800 family, there is evidently a bug in AMD drivers, which hasn't been fixed yet. If it's not a bug, the GeForce 8800 GT is 50 times as fast as all competing cards from AMD. Our test is not 3DMark or Crysis, nobody will optimize drivers for it... On one hand, it's good. On the other hand, such bugs should be fixed.
Your link here
