[Beowulf] Nvidia Tesla GPU clusters?

[Vincent Diepeveen]

hi Jon,

for each system one card is needed to serve for video.
the ones that are not serving for video can be used for normal
calculations. Like my laptop has a build in nvidia 8600M GT card,
but that card is needed for me to see the display.

Please realize the huge wattage those cards eat from the psu.

A quadcore in itself eats like 172 watt that's INCLUDING the loss a PSU
causes.

Including loss of PSU a graphics card eats far over 200 watt and PS3 even
can go up to 380 watt.

So in short you're not going to put in graphics cards in clusters: "just
to use them now and then".

For software that's not requiring huge amount of ram and which is parallel
embarrassingly and has DSP habits, it's quite possible to use those GPU's
fast. On paper i calculated that its practical performance should be
about factor 3 faster for prime95 for the ATI 2900 card. Tesla probably a
bit slower than that for prime95.

Achieving that speed is not going to be easy though, if it is possible at
all.

Even then, it's quite impressive what potential those cards have.

Vincent

On Wed, 18 Jul 2007, Jon Forrest wrote:

> Toon Knapen wrote:
> >> http://www.nvidia.com/object/tesla_computing_solutions.html
> >
> > Anyone can point me to more information about the 'thread execution
> > manager' and how threads can enable getting optimal performance out of
> > this hardware ?
>
> This is a good question. When word first came out about using GPUs
> for regular computation I sent a message to comp.arch (which is pretty
> much a wasteland these days) asking how jobs were going to be
> scheduled on a GPU. Nobody knew. I would think this would be
> especially important if the same GPU were going to be used
> for graphics display and HPC computations. Even if it would only
> be used for HPC computations its resources will have to be scheduled
> one day. Maybe it could be scheduled as a asymetric MP, which
> certain tasks, e.g. the graphics and HPC tasks, having affinity to
> the GPU.
>
> --
> Jon Forrest
> Unix Computing Support
> College of Chemistry
> 173 Tan Hall
> University of California Berkeley
> Berkeley, CA
> 94720-1460
> 510-643-1032
> jlforrest at berkeley.edu
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
>