GPU no OC tile 256x256 39:28.48
GPU no OC tile 200x360 about 1 min less (i forgot to paste it and don’t want to wait 38 minutes more :P)
GPU OC GPU 1050MHz RAM 1500MHz tile 200x360 34:39.59 (no overvolt)
GPU OC GPU 1150MHz RAM 1600MHz tile 200x360 32:04.14 (little overvolt, figured i can go to 1100/1550 without ov)
As you see, it’s disappointingly slow compared, for example, to a GTX 560. I hope this can be better.
If you take a quick look to all results you’ll see a GTX750 between 4xGTX590 and 3x780ti, just were your “OpenGL Engine” is found.
Or some GTX560ti faster than GTX980 or 780ti.
This benchmark is a good idea but the results are full of incoherence. I think criticism is necessary here.
I originally ran this benchmark a few weeks ago but I didn’t post because I got roughly average times and didn’t think it would add any helpful info. However, yesterday I was playing around in the blender preferences and realized that I was only rendering at half speed. Blender and windows sees the Titian Z as two separate titan cards and I needed to select the option for “Geforce GTX Titan Z (2x)” from the drop down. Now the GPU renders 2 tiles at a time and renders take half the time.
This is probably just my blender novice mistake, but hopefully this could be a helpful watch-out for anyone else running a Titan Z card.
Windows 7 64bit
Blender 2.73
Geforce GTX Titan Z (2x)
No optimizations made to any of the files, just opened and press F12.
file 1: 04:34.31
file 2: 03:03.68
are #49 and #50 the result of the gpu strength or hitting the 2 gig vram cap? would more vram help? the 650 taking almost twice as long as the 460 makes me think it cant be just gpu speed. or maybe it was an opencl not cuda render?
So it is not only that here and there some chips or VRAM settings are different:
For example the 590 were or are two cards in one - thats why it is so expensive.
Windows 7 (64)
GTX 580 (PNY XLR8 OCed and watercooled)
GTX 580 (Stock)
Rendertime:
Sponza: 5:26.78
2.72 file: 5:41.04
Additional note for people interested in the performance difference of the cards, here’s the test run on each card individually:
GPU 1 9:42.93
GPU 2 11:00.56
The problem with running with dual-GPUs is that you only get a small performance improvement from upgrading a card to a 780ti or a 980. I really want to upgrade but… so little reason to. 3 GPUs is clearly the answer here.
