Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sat, 13 Jul 2013 07:47:48 +0200
From: marcus.desto <marcus.desto@...pl>
To: john-dev@...ts.openwall.com
Subject: Re: new dev box wishes

Hi Alexander


Dnia 13 lipca 2013 3:15 Solar Designer <solar@...nwall.com> napisaƂ(a):

> 
> For now, our new dev box has been setup with GPUs, but without Xeon Phi.
> I am running some Bitcoin mining as a stress-test on its 3 GPUs at the
> moment.  7990's first GPU got to 93C at first, but then I forced 7990's
> fans speed to 100% (somehow it didn't go above 74% on its own) and the
> temperatures dropped to under 90C.
> 

I recommend liquid cooling for highend gpus devices:

http://www.ekwb.com/shop/ek-fc7990-se-acetal-nickel.html

Advantages are obvious, besides low noise. Esspecially live time is an issue.
I heard a lot of times peoples high end gpus (esspecially the ati 4000, 5000 series) died,
because of overheating after fan failure.

I am wondering why vendors still ship high end gpu devices with air fans knowing the gpus produce enormous heat.

There are very less OEM licuid cooling configurations on the market, in general.
Mosty, they are 30-45% more than regular price. Producing a lot of liquid cooled gpu devices would lower producing costs, so the price, too.

Maybe there is an option for Xeon Phi to go liquid cooling, but it is patch work - which actually sucks.

I am wondering whether it is really necessary to use high end gpu devices in high end PCIe slots (i.e. AMD 7990 in PCIe v3 16x) when using the device for GPGPU purposes. I mean gpus memory acts as buffer. The higher the performance of the gpu the less the traffic on the buff, because of the buffer, right? Respecting some lower bounds on need for transfering volume of data.

Regards,
Marcus

Powered by blists - more mailing lists

Your e-mail address:

Powered by Openwall GNU/*/Linux - Powered by OpenVZ