Does Passmark test GPUs also?
First, does Passmark have a testing routine for GPUs?
Did anyone ever tried to do that?
I have Passmark 6.1 (1016) 64-bit and a GPU dedicated for computing purposes (number crunching, no graphics) but I am not able to run Passmark and test it.
Might it be related to the software version? would an update help me?
The 3D tests run on the GPU (and CPU).
There is shader code that executes on the GPU as part of the 3D test in the sense that you are referring to. It isn't a pure GPU test, but faster GPUs tend to be better on the 3D test.
What exactly do you mean by "can't run it"?
Thanks for your reply. With "I can't run it" I mean that the 3D graphics suite only tests for my graphic card (a Quadro) versus I am interested in knowing the performances of my GPU (Tesla, still from Nvidia) which is not used for graphics but for calculation purposes. So, bottom line, I am not able to find an option in Passmark (also in the 3D Graphics test) which allows me to test my GPU. Is there any solution? any workaround? you said there is a shader code but when I run the test it only refers to "Quadro".
You seem to be implying that the graphic's card and the GPU are totally different devices. And that a test of graphics card test will not also test the GPU. This is not true. Any 3D graphics using DirectX requires and uses the GPU. The GPU (plus some RAM, and a PCI interface) is the graphic's card.
What is true is there is different code that can be executed on the GPU. It is possible to write algorithms in programming languages, like C, that execute on the GPU (just like a CPU). Shaders are one example of this. For example the water effect, smoke effect, etc.. in the PerformanceTest 3D tests runs on the GPU. And the faster it does these calculations, the higher the frame rate will be.
But what is also true is that you might be wanting to use your GPU for other purposes. e.g. Bitcoin generation, password cracking, etc..
The 3D test results in PerformanceTest will give a rough indication of what video card might be good at doing this, but the best would be to benchmark the actual GPU application.
Thanks again for clarifying my doubts. I was just confused because in the Passmark results I only saw the name of (what I called) my actual graphic card (Nvidia Quadro) versus I didn't see any reference to the GPU that I am using for number crunching (Nvdia Tesla). That's why I was not sure Passmark would have taken in consideration also Tesla during the 3D test.