Hi Marek,
I have a GPU machine with the motherboard you were interested in (ASUS P6T7 WS SuperComputer, single LGA1366 socket with the Intel X58 chipset). It has an Intel Xeon W3520 Nehalem 2.66 GHz quad core CPU and 4x NVIDIA Tesla C2070s. Below are some numbers (ns/day) for this machine running a subset of the official AMBER benchmark suite so you can get an idea of the speedup we get when going to all 4 GPUs for a single job.
Benchmark 1xC2070 2xC2070 3xC2070 4xC2070
GB/myoglobin 63.03 77.14 90.69 102.38
GB/nucleosome 1.10 1.34 1.68 1.97
GB/TRPCage 354.40 334.06 330.76 330.75
PME/Cellulose_production_NPT 1.97 2.76 3.29 3.56
PME/Cellulose_production_NVE 2.19 3.06 3.67 3.96
As for the issue of cooling, this system is housed in a Lian Li chassis (25x24.9x8.6) with three fans in the front, one in the back, and one on top. I had also noted that the GPUs were getting extremely hot and contacted Microway, the company who assembled the machine for us. The assured me that they'd had experience with 4x Teslas in that chassis, and that cooling was sufficient.
I will warn you, however, that we recently experienced a socket burn with this motherboard, where the 24-pin ATX power connection from the PSU to the motherboard had the plastic melt off of some of the pins. Microway/ASUS are replacing it for us under warranty, so hopefully we just had a lemon.
Hope this helps,
Jodi Hadden
University of Georgia
On Sep 13, 2011, at 1:17 PM, Marek Maly wrote:
OK,
thanks again ! If anyone has experience with Amber calculations on
a single socket machine with a 4 core CPU equipped with 4 GPUs,
please comment.
Best wishes,
Marek
Dne Tue, 13 Sep 2011 19:25:20 +0200 Ross Walker <ross.rosswalker.co.uk<mailto:ross.rosswalker.co.uk>>
napsal/-a:
first of all thanks a lot for your complex answer !
In fact I assume mainly independent single GPU jobs. So if I understood
well, in such case there should not be problem considering below
mentioned
motherboard/CPU/4xGPU. Am I right ?
For single GPU runs (i.e. 4 independent jobs) then things should be fine
assuming the I/O can keep up etc. The caveat with that is that I have not
actually tried 4 GPUs in a single socket machine with a 4 core GPU so am
speaking from a theoretical standpoint here given how the AMBER GPU code
works. Someone else who is running such a system might want to chime in
with
some specific performance numbers if they have them.
All the best
Ross
/\
\/
|\oss Walker
---------------------------------------------------------
| Assistant Research Professor |
| San Diego Supercomputer Center |
| Adjunct Assistant Professor |
| Dept. of Chemistry and Biochemistry |
| University of California San Diego |
| NVIDIA Fellow |
|
http://www.rosswalker.co.uk |
http://www.wmd-lab.org/ |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk<mailto:ross.rosswalker.co.uk> |
---------------------------------------------------------
Note: Electronic Mail is not secure, has no guarantee of delivery, may
not
be read every day, and should not be used for urgent or sensitive issues.
_______________________________________________
AMBER mailing list
AMBER.ambermd.org<mailto:AMBER.ambermd.org>
http://lists.ambermd.org/mailman/listinfo/amber
__________ Informace od ESET NOD32 Antivirus, verze databaze 6459
(20110913) __________
Tuto zpravu proveril ESET NOD32 Antivirus.
http://www.eset.cz
--
Tato zpráva byla vytvořena převratným poštovním klientem Opery:
http://www.opera.com/mail/
_______________________________________________
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 Sep 14 2011 - 09:00:02 PDT
