From: John Stone (johns_at_ks.uiuc.edu)
Date: Thu Mar 20 2008 - 19:29:58 CDT

Michael,
  Wellllllll if you work for a hardware vendor.... I'd love to see
someone make machines with with four x16 PCIe 2.0 slots,
on-board infiniband, several CPU sockets, and large memory capacity!
We could definitely have a use a server box with that sort of config.
If your server vendor employer is looking for input on such things,
let us know. :-)

Cheers,
  John Stone
  vmd_at_ks.uiuc.edu

On Thu, Mar 20, 2008 at 07:42:50PM -0400, Michael Di Domenico wrote:
> John, thanks. I understand, by day I work for a server manufacturer... :)
>
> On Thu, Mar 20, 2008 at 4:34 PM, John Stone <johns_at_ks.uiuc.edu> wrote:
>
> >
> > Michael,
> > If you're building a new machine for CUDA/GPU work, I'd suggest trying
> > to get a motherboard that's based on PCI express 2.0, and I'd make sure
> > you
> > get a power supply with the 8-pin PCIe power connectors, since all of the
> > new GPU boards have already begun to require them...
> > The PCIe 2.0 motherboards can transfer over 5 gigabytes/sec to and
> > from the GPU, the best of the PCIe 1.x motherboards can only push
> > about 3.4 gigabytes/sec. The Asus motherboard in our test machines
> > are in-between, as they have a "LinkBoost" feature that overclocks
> > the PCIe bus, but the two numbers I gave above are more standard.
> >
> > Cheers,
> > John Stone
> > vmd_at_ks.uiuc.edu
> >
> > On Thu, Mar 20, 2008 at 04:29:41PM -0400, Michael Di Domenico wrote:
> > > Thanks, that's pretty much what i was looking for. I'm looking for a
> > test
> > > machine to learn cuda with and was curious what you had used for
> > > MB/Processor.
> > >
> > > Thanks
> > > - Michael
> > >
> > > On Thu, Mar 20, 2008 at 12:25 PM, John Stone <johns_at_ks.uiuc.edu> wrote:
> > >
> > > >
> > > > Hi,
> > > > Sure, what do you want to know? We ordered the basic boxes from
> > > > a local vendor and I put the rest of these system together myself.
> > > > The motherboard is an Asus P5N32-E SLI, with an Intel QX6700 CPU,
> > > > 8GB of memory, running RHEL4u5 with the 64-bit 2.6.9-55 SMP kernel.
> > > > One of the test systems has the stock 800 watt PSU that came from
> > > > our local vendor, I installed a 1200 watt Thermaltake PSU in
> > > > the other one.
> > > >
> > > > Both PSUs work fine, but the 1200 watt PSU is much
> > > > more convenient because it has six of the PCIe GPU power connectors,
> > > > and three of them are the new 8-pin variety making it easy to test
> > > > newer GPU cards. The one with the stock PSU runs fine, but I had
> > > > to use a bunch of molex-to-PCIe power converter cables. Both test
> > > > machines originally contained three GeForce 8800GTX GPUs, but most
> > > > recently I'm now testing one of them with a mix of the the Tesla
> > > > D870 (two GPUs) deskside unit, and I hope to test some of the newly
> > > > announced cards sometime in the near future.
> > > >
> > > > At present these two test machines are just running CUDA tests and we
> > > > aren't doing much visualization on them, but we have previously had
> > > > them setup for running some very demanding VMD visualizations.
> > > >
> > > > If you have more questions, let me know.
> > > >
> > > > Cheers,
> > > > John Stone
> > > > vmd_at_ks.uiuc.edu
> > > >
> > > > On Thu, Mar 20, 2008 at 12:04:59PM -0400, Michael Di Domenico wrote:
> > > > > Does anyone know if the full spec sheet for the Triple GPU test
> > machine,
> > > > > thats shown on the VMD webpages, exists?
> > > > >
> > > > > Thanks
> > > > > - Michael
> > > >
> > > > --
> > > > NIH Resource for Macromolecular Modeling and Bioinformatics
> > > > Beckman Institute for Advanced Science and Technology
> > > > University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
> > > > Email: johns_at_ks.uiuc.edu Phone: 217-244-3349
> > > > WWW: http://www.ks.uiuc.edu/~johns/><
> >
http://www.ks.uiuc.edu/%7Ejohns/>
> > > > Fax: 217-244-6078
> > > >
> >
> > --
> > NIH Resource for Macromolecular Modeling and Bioinformatics
> > Beckman Institute for Advanced Science and Technology
> > University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
> > Email: johns_at_ks.uiuc.edu Phone: 217-244-3349
> > WWW: http://www.ks.uiuc.edu/~johns/ <http://www.ks.uiuc.edu/%7Ejohns/>
> > Fax: 217-244-6078
> >

-- 
NIH Resource for Macromolecular Modeling and Bioinformatics
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
Email: johns_at_ks.uiuc.edu                 Phone: 217-244-3349
  WWW: http://www.ks.uiuc.edu/~johns/      Fax: 217-244-6078