Investigation into different Polygon pipeline performance *spawn

Digidi

Regular
In the 3d Center we made a comparsion of the polygon output test. We got the following results:

Link to Benchmark, zip-file belwo:

It is intresting that 6900xt is not looking so bad in the old Primitive Pipline. Can somboy explain what this resault mean, and what it means that 6900xt is bad in MultiDraw Indirect?

I also interpolate also 2ghz i think it is linear.

Edit 2022.12.06 : i took the wrong clock for 6900xt was running 2,5 ghz sorry for my fault. corrected the graph
polygons2.jpgpolygons3.jpgpolygons4.jpg
 
Last edited:
Link to Benchmark, zip-file belwo:
It is intresting that 6900xt is not looking so bad in the old Primitive Pipline. Can somboy explain what this resault mean, and what it means that 6900xt is bad in MultiDraw Indirect?
Wow, saw this before. Those guys rock, and the data they give is precious.

The blog has detailed explanation about the numbers. Without that your charts are pretty useless.
However, we can see that compute raster wins on pretty any HW with small triangles, including mobiles.
And i have just learned that Mesh Shaders seem to be an NV only feature.

Idk exactly what MultiDrawIndirect (Edit: confused with DX12 Execute Indirect) is, because Vulkan does not have it. Seems the biggest difference between DX12 and Vulkan. Afaik it is an indirect draw but also bundles resources with it, but somebody else needs to tell.
 
Last edited:
Idk exactly what MultiDrawIndirect is, because Vulkan does not have it. Seems the biggest difference between DX12 and Vulkan. Afaik it is an indirect draw but also bundles resources with it, but somebody else needs to tell.
Thank you. But Vulkan Must have MultiDrawIndirect this Benchmark was running on Vulcan api.

AMD has Mesh Shader under DX12 but it is not running with this benchmark on Windows 11

Here they talk about indices. AMD is much faster without indices. For what are indices used and what does this mean for AMD and Nvidia?
 
Last edited:
Thank you. But Vulkan Must have MultiDrawIndirect this Benchmark was running on Vulcan api
Oops, i've confused it with DX12 'Execute Indirect'.
But i'm still a gfx pipeline noob. However, Multi Draw Indirect then surely is a way to generate draw calls on GPU. Useful for GPU driven rendering.
 
Thank you for splitting! Does anybody now which pipline we use today and what are the benefits and trade offs of the different pipelines?
 
Can nobody explain what are the differences and pro and cons of the different piplines? Would be realy nice to know-
 
Here they talk about indices. AMD is much faster without indices. For what are indices used and what does this mean for AMD and Nvidia?
Say you are rendering a rectangle. That's 2 triangles. So how many vertices do you use? 4 or 6? With indices you use 4 vertices + 6 indices and without indices you have 6 vertices.
Thank you for splitting! Does anybody now which pipline we use today and what are the benefits and trade offs of the different pipelines?
Depends. Multi draw indirect is also something quite powerfull in combination with compute shaders. Mesh shaders on the other hand keep all the data on the GPU.
 
Here are Values for 7900XTX it is so weared. Thanks for Maestrox from the 3dcenter forum!

polygons6.jpg
polygons5.jpg
 
Last edited:
Back
Top