Hi Guys,
I’m doing some 3D modeling from time to time with Blender. I was able to get my hands on some GRID/Tesla cards, and thought I would do some benchmark testing.
The GPU’s I have:
GTX 580 (3GB)
GRID K2 (4GB)
Tesla K20Xm (5GB)
CPU:
2x Xeon 2695
I’ve installed the standard Blender benchmark, and had it run in the default configuration. I did run the test in single GPU mode, and finally all together
So what are the render times:
| GPU/CPU | Render Time | |
|---|---|---|
| GTX 580 (3GB) | Nvidia GTX580 | 1:29.35 seconds |
| GRID K2 (4GB) | GRID K2 | 2:01.03 seconds |
| Tesla K20Xm (5GB) | Tesla K20Xm | 3:02.74 seconds |
| All GPU's together | 0:36.43 seconds | |
| 2x Xeon 2695 | Intel Xeon 2690v2 | 1:03.98 seconds |
Conclusion:
As you can see, the later K20Xm is not living up to my expectations, which was quite a surprise to me. As the GRID K2 has two GPU’s on the card, and I happened to have two of them, it gives me the most performance.
On the other hand I’m looking at getting the “Titan X” which holds 12GB of RAM and has the latest GPU. Although I read that the CUDA performance in Blender is not so great (80s rendering time for the same scene). So let’s see what the market is bringing new 🙂
Cheers
Roman
