3 ms·
I'd be curious to hear who is using servers with the proprietary form factor GPUs. The prices for datacenter vs. gamer versions of the same GPUs have gotten re
by cameldrv 9y ago
I'd be curious to hear who is using servers with the proprietary form factor GPUs. The prices for datacenter vs. gamer versions of the same GPUs have gotten really out of whack, where a 1080ti is $700, and a P100 is $7000, and at least on FP32, they have the same performance.
From what I hear, most all of the major DL research labs are using 4U servers with 8 of either the 1080ti or the Titan Xp rather than the SXM2 Teslas. On the other hand, all of the cloud providers seem to be sticking with the Tesla products, and pricing GPU instances commensurately. I really wish there were a cloud provider that offered instances with the gamer cards in them so that companies in the DL field weren't effectively forced to buy and manage their own hardware.
- sacheendra 9y agoYou can use the AWS g2 instances.
- verall 9y agoA P100 has a gp100 with nvlink, hbm2, and a 610mm die, while a 1080ti or titan Xp has a gp102 with gddr5x and 471mm die. It is not like quadro vs geforce where they both run on a gp102 with different drivers, the gp100 is considerably physically larger and has a much smaller userbase. I would be very surprised if they have the same fp32 performance in real world situations.
- tanderson92 9y agoRight, well, people who need FP64 aren't going to be satisfied with the 1080ti so that is the answer. See this comment of mine: https://news.ycombinator.com/item?id=14597486 https://news.ycombinator.com/item?id=14597486
- matt4077 9y agoI actually think FP16 performance is more of a factor for those choosing the Tesla platform.
- tanderson92 9y agoYou mean for the Volta architecture, surely? I'm talking about currently available hardware.
- jabl 9y agoThe Pascal Teslas (P100), at least, support FP16.