[ExI] Life @ Playstation

Stefano Vaj stefano.vaj at gmail.com
Sun Nov 4 17:42:34 UTC 2012


On 4 November 2012 18:04, Tomasz Rola <rtomek at ceti.pl> wrote:

> You mean this table?
>
> http://fah-web.stanford.edu/cgi-bin/main.py?qtype=osstats
>

Yes.

It is interesting indeed. What it tells me, however, is that only 10%
> Windows machines use GPU for computing. Together (ATI+nVidia) those GPUs
> deliver four times total PS3 power in x86-TFlops, while the number of
> units in both groups roughly matches.
>

What is strange is that client v7.x makes concurrent CPU and GPU folding
very easy (alas, only under Windows), so it is not very clear why so many
more Windows people still do CPU-only.

Moreover, Nvidia users used to have a significant lead over ATI users, then
their number dropped quite significantly. I suspect that Nvidia itself was
running Folding at Home on its PCs and that they stopped at a point in time...

The project leaders insist however that a CPU Teraflop and a GPU Teraflop
are not exactly the same thing, given that the latter is only useful to run
rather specialised projects, often sacrifying accuracy and realism for
speed (eg, explicit or implicit solvent models).

I don't contribute to any of those efforts at the moment. There were times
> when I thought about it, but misc "events" in life prevented me from
> making a step. Besides, usually I need cpu too.


Once upon a time, and on the PS3,  it was a screen saver, so such
considerations did not apply. The existing client is supposed to run at all
time, but is very discreet, meaning that it has the lowest possible
priority, and if you have anything computationally intensive it does not
interfere in the least, it really makes use only of *spare* cycles, which
are wasted anyway. The only tradeoff is that you burn a little more
electricity.

And if one is paranoid, he could run it of course in a sandbox or in a
virtual machine.

-- 
Stefano Vaj
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20121104/8f7c76ca/attachment.html>


More information about the extropy-chat mailing list