From: Aron Broom (broomsday_at_gmail.com)
Date: Thu Sep 13 2012 - 18:01:14 CDT
Thanks for posting these, I've been very curious about the 600 series
It looks like at the moment the performance is roughly on par with a GTX
570 or M2070 card.
I would suspect that with some optimization specific for these cards future
releases might be much faster? Or is your sense that this is where they
will be staying?
I guess one thing to note is that based on a rapid google-search, the H77
chipset would be a PCI 2.0, so the CUDA performance may be suffering from
not being able to use the full card on account of bandwidth limitations,
which could be alleviated with a PCI 3.0. Maybe this idea is supported by
the low power consumption. I wonder if the power consumption is higher
with AMBER or OpenMM/GROMACS? I noticed when doing things on the previous
line of cards that NAMD produced far less GPU heat than other packages.
Thanks again for the post!
On Thu, Sep 13, 2012 at 3:50 PM, Thomas Albers <talbers_at_binghamton.edu>wrote:
> Since someone was asking a few days back regarding performance of Kepler
> GPUs, here are some benchmarking results. This is with the
> Linux-x86_64-multicore and Linux-x86_64-multicore-CUDA builds of NAMD
> 2.9 as downloaded from the NAMD website, on a system with i5-3570 CPU
> and Intel H77 chipset and a GTX-660 TI graphics card. Power consumption
> with GPU acceleration is about 200 W.
> The last three are from the Daresbury benchmark suite, download here:
> Timing results in s/step.
> NAMD NAMD-CUDA
> ApoA1 (90K atoms) 0.187 0.038
> F1ATPase (330K atoms) 0.537 0.136
> STMV (1M atoms) 2.351 0.472
> GlnBP (61K atoms) 0.140 0.030
> EGFR-flat (465K atoms) 1.136 0.267
> EGFR-tall (1.4 M atoms) 3.421 0.755
-- Aron Broom M.Sc PhD Student Department of Chemistry University of Waterloo
This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:22:04 CST