In the new 3DMark test, the Radeon RX 6000 cards beat the GeForce RTX 30 cards. Speech on the Mesh Shaders Feature Test

A new test has appeared in the list of 3DMark benchmarks – Mesh Shaders Feature Test. This is a benchmark test using Mesh Shaders.

Mesh Shaders is a new shader model that is responsible for all work on geometry, vertices, tessellation at once. Such shaders replace vertex and geometry shaders and tessellation, and the entire usual vertex pipeline is replaced by an analogue of computational shaders for geometry, which can do everything a developer needs: transform vertices, create them or remove them, using vertex buffers for their own purposes, as you like, creating geometry directly to the GPU and sending it for rasterization. As a result, such a solution can greatly reduce the dependence on CPU power when rendering complex scenes and will allow you to create rich virtual worlds with a huge number of unique objects. This method will also allow for more efficient discarding of invisible geometry, advanced level of detail (LOD) techniques, and even procedural geometry generation.

Shaders themselves appeared several years ago, but only now, thanks to the fact that they have become part of DirectX12 Ultimate, they can gain popularity among developers. Using mesh shaders in some cases can help dramatically improve performance, and the Mesh Shaders Feature Test demonstrates how much. The benchmark simply compares the performance in a short scene without and with grid shaders, and then gives the results in absolute terms, as well as an indicator of the performance gain.

The test came out only today, so there are still few results on the Web. In addition, AMD later released the Radeon Adrenalin 2020 21.2.2 driver, which appears to greatly improve the performance of Radeon cards in this test.

Based on the results of one of the sources, on average, Nvidia GeForce RTX 30 video cards provide a 702% increase when using grid shaders, RTX 20 cards – 409%, and Turing generation GeForce GTX cards – 244%. At the same time, the Radeon RX 6000 adapters before the release of the fresh driver provided an increase of 547%, and after – over 1600%! But, of course, it’s worth looking at each model specifically, since the difference between the RTX 3090 and RTX 3060 Ti is almost twofold.

It is also worth noting that in this case it is the performance gain that is important, and not the absolute indicators before the inclusion of mesh shaders, since in the latter case the inexpensive GTX 1660 Ti bypasses the monstrous RTX 3090.