Ashutosh,
I tested the prmtop/coord files you sent me (offlist). I tried
pmemd.cuda.MPI on two GPUs and pmemd.cuda on a single GPU. I did not see
any errors for 50000 steps.
A few questions/comments:
- How long until you see the error? Is always immediately after you start
the simulation or does it take a while?
- What CUDA driver version are you using?
- FYI, the input file you included is constant pressure not constant volume
(ie, ntp=1)
- The GPU code in AMBER does not (yet) give good scaling across multiple
GPUs for systems of this size. On the GTX 680's I was using, I got 21
ns/day on one GPU and 26 ns/day on two GPUs. (I used larger values for
ntpr,ntwx,ntwr)
--Niel
On Wed, Nov 20, 2013 at 9:19 AM, Ashutosh Shandilya
<izerokelvin.gmail.com>wrote:
> While attaching topology file it is asking for approval from amber site. I
> am resending the same mail again.
>
> Sorry for duplication if any.
>
>
> To remove the extra point issue I made the topology using TIP3P instead of
> TIP4PEW. I tried a test run using pmemd.cuda on individual machines and its
> running fine without any error. I checked that with cpu serial and multiple
> processors there are no errors and are running fine there as well.
>
> One question as said earlier that amber 12 does not really gain much using
> multiple GPUs. Is it the problem with AMBER 12 or AMBER 11 as well.
> Technical specifications for my computers are
> *CPU/Processor *
> 2x 8-core Intel Xeon® E5-2670 with 20MB Cache . 2.6Ghz
> *RAM *
> 64GB ECC DDR 3,1600MHz equal distribution of memory on DIMMs
> *GPU *
> NVIDIA K20
> *CPU:GPU Ratio*
> 2:2 only on Compute nodes with GPU Cards
> *GPU Memory*
> 6GB per GPU
> *Connectivitybetween CPU andGPU*
> Dedicated 16xPCI Express bus (Gen 3.0)
>
> Input files are with the attachment for the test run
> Heating the system with 25 kcal/mol restraints on protein,at constant
> volume
> &cntrl
> imin = 0, ntx=1, ntb=2, ntp = 1, taup=2.0, cut = 10,
> ntc = 2, ntf = 2, tempi = 0.0, temp0 = 300.0, ntt = 3,
> gamma_ln = 1.0, nstlim = 500, dt = 0.002, ntpr =50, ntwx = 50, ntwr = 50
> /
> for checking purposes, I have removed the restraints for now.
> Topology and coordinate files are with the attachment.
>
> Thanks
>
> Ashutosh
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Nov 20 2013 - 10:30:03 PST
