Re: [AMBER] pmemd.cuda DPDP performance on GTX480 vs C2050?

From: Scott Le Grand <>
Date: Fri, 11 Jun 2010 09:41:35 -0700

C2050 will smash a GTX480 in DPDP mode...

1.5x to 2x faster the last time I tried it... But since my GTX480 committed suicide last month (emanating a delicious but likely amazingly toxic curry-like scent in the process) I haven't been able to test lately.

That said, it's 20-30% better than C2050 at SPDP but likely won't be as good at multi-GPU (untested hypothesis there) and prone to instability without slavish attention to power and cooling, something which the Folding.Home people bemoan whenever their hand-me-down SLI gaming computer is turned into a mini folding farm.

-----Original Message-----
From: Ross Walker []
Sent: Friday, June 11, 2010 07:50
To: 'AMBER Mailing List'
Subject: Re: [AMBER] pmemd.cuda DPDP performance on GTX480 vs C2050?

Hi Sahsa

> has anyone compared performance of pmemd.cuda in DPDP mode on GTX480
> and
> C2050?
> Tesla cards are supposed to have a better double precision performance,
> but how much better in the event of an all-double pmemd code?
> Thanks for any info

I don't have a GTX480 so have not been able to compare the differences,
however, here are the numbers for C2050, I'd be interested to see the


NVE 128x64x64 FFT
 Typical Production MD NVE with
 GOOD energy conservation.
   ntx=5, irest=1,
   ntc=2, ntf=2, tol=0.000001,
   ntpr=1000, ntwx=1000,
   dt=0.002, cut=8.,
   ntt=0, ntb=1, ntp=0,

                SPSP SPDP (Default) DPDP
Specs Time (s) NS/day Time (s) NS/day Time (s)
8 proc N/A N/A 1034 1.671179884
GTX295 395.46 4.369594902 525.6 3.287671233 2469.18 0.699827473
C1060 366.25 4.718088737 492.84 3.506208912 2354.73 0.733842097
C2050 311.34 5.550202351 333.04 5.188565938 706.98 2.44419927

Myoglobin = 2492 atoms
  ntpr=1000, ntwx=1000, ntwr=50000,

  cut=9999.0, rgbmax=15.0,
                SPSP SPDP (Default) DPDP
Specs Time (s) NS/day Time (s) NS/day Time (s)
8 proc N/A N/A 395.73 4.3666136
GTX295 39.62 43.61433619 64.99 26.58870595 3.0 Compiler bug
C1060 37.79 45.72638264 61.94 27.89796577 3.0 Compiler bug
C2050 31.39 55.04937878 34.6 49.94219653 155.7 11.0982659

All the best

AMBER mailing list
This email message is for the sole use of the intended recipient(s) and may contain
confidential information. Any unauthorized review, use, disclosure or distribution
is prohibited. If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message.

AMBER mailing list
Received on Fri Jun 11 2010 - 10:00:03 PDT
Custom Search