Testing 980 TI's + 970
Posted: Fri May 20, 2016 2:42 pm
I ran some tests this morning that I thought I'd go ahead and post here.
I had heard that my one 970 might be slowing down my four 980 Ti's, so I rendered out a bunch of tests to check it out.
CONCLUSIONS:
1) 980 TI is 2.5 times faster than the 970 in these tests. That's surprising, because published benchmarks say (from memory something like) the 980 TI benchmarks at 125 and the 970 at 80. So I thought it would be more like 1.5 times.
2) The 970 does NOT slow down the faster cards here (when VRAM isn't loaded up). But it helps out so little that it's not even worth it. Here I'm not using a lot of VRAM memory, if I was, the 970's 4gig would limit the 980's to only using 4gig of their 6gig.
3) 970's are essentially useless for my particular rendering setup. A 970 makes a difference next to a single 980 Ti, but next to four 980 Ti's it isn't worth having. My 970s are useful if I need to drive a monitor with them. Unfortunately, they'll take up a slot that I could have a faster card in.
4) 980 TI's scale up roughly linearly, in this test of 1 to 4 of them.
RENDER TIMES:
4x 980 TI + 1x 970 = 2:21
4x 980 Ti = 2:32
3x 980 Ti = 3:23
2x 980 Ti = 5:06
1x 980 Ti = 9:58
1x 970 = 24:42 (Snoooooozzzze.......)
1x 980 Ti + 1x 970 = 7:08
TEST CONDITIONS:
The test scene had a lot of specular, so you could say this was a Octane specular material test (No diffuse or glossy in the scene)0. It was a C4D human figure cloned 72 times (not render instances) with a simple specular material with colored transmission. Cranked specular depth up to an unnecessary 25. Samples at 512. One Octane area light, and an Octane environment image. I'm on Windows 10, Cinema 4D 17, Octane 2.25 R3, Intel i7. I've got the 970 on the motherboard, and the 4x 980 Ti's on a splitter externally, plenty of PSU power for them.
THANKS FOR WATCHING
Anyway, I thought that was interesting. I make no claims about what anyone's best setup might be, I'm just trying to get the most out of this rig.
Cheers,
M
I had heard that my one 970 might be slowing down my four 980 Ti's, so I rendered out a bunch of tests to check it out.
CONCLUSIONS:
1) 980 TI is 2.5 times faster than the 970 in these tests. That's surprising, because published benchmarks say (from memory something like) the 980 TI benchmarks at 125 and the 970 at 80. So I thought it would be more like 1.5 times.
2) The 970 does NOT slow down the faster cards here (when VRAM isn't loaded up). But it helps out so little that it's not even worth it. Here I'm not using a lot of VRAM memory, if I was, the 970's 4gig would limit the 980's to only using 4gig of their 6gig.
3) 970's are essentially useless for my particular rendering setup. A 970 makes a difference next to a single 980 Ti, but next to four 980 Ti's it isn't worth having. My 970s are useful if I need to drive a monitor with them. Unfortunately, they'll take up a slot that I could have a faster card in.
4) 980 TI's scale up roughly linearly, in this test of 1 to 4 of them.
RENDER TIMES:
4x 980 TI + 1x 970 = 2:21
4x 980 Ti = 2:32
3x 980 Ti = 3:23
2x 980 Ti = 5:06
1x 980 Ti = 9:58
1x 970 = 24:42 (Snoooooozzzze.......)
1x 980 Ti + 1x 970 = 7:08
TEST CONDITIONS:
The test scene had a lot of specular, so you could say this was a Octane specular material test (No diffuse or glossy in the scene)0. It was a C4D human figure cloned 72 times (not render instances) with a simple specular material with colored transmission. Cranked specular depth up to an unnecessary 25. Samples at 512. One Octane area light, and an Octane environment image. I'm on Windows 10, Cinema 4D 17, Octane 2.25 R3, Intel i7. I've got the 970 on the motherboard, and the 4x 980 Ti's on a splitter externally, plenty of PSU power for them.
THANKS FOR WATCHING
Anyway, I thought that was interesting. I make no claims about what anyone's best setup might be, I'm just trying to get the most out of this rig.
Cheers,
M
lol....