Cool GPU based supercomputer

[-Sn1PeR-]

http://www.dvhardware.net/article27538.html

Ah, the future of folding.

 

[digital_exhaust]

Um.....wow

That is really interesting, and now I find myself feeling like I need one, but that would severely cut into the beer fundage...big time.

This should be interesting to follow, but I do think they need to work on some sort of cooling solution for that thing.... those four cards crammed in there look as if they may turn into a molten blob of plastic goop before too long....

Cool stuff.... thanks for the link dude...

 

[Shady240]

HOLY SH*T!!! do i see liquid in the future for that uber rig?

 

[APOLLO]

Apparently, they keep the case open at all times. I would place at least one 120 or 140mm fan over the cards.

 

[alan2308]

So basically a homemade Tesla on the cheap. Seriously though, that many cards packed in that tight and the load that I'm sure that they are keeping this thing under, some more cooling is in order.

 

[APOLLO]

Speaking of Tesla, anyone know what potential there is for the GPU2 client on one of those? They're way out of my purchase league but I'm curious just the same.

 

[Tigerbiten]

There's only 4 CPU cores driving the 8 GPU cores.
If the nVidia client is anything like the ATI client, then 1x CPU core will be needed to dive 1x GPU core.
At a pure guess of 2,000-2,500 PpD per client, then its only 8,000-10,000 PpD.
Not that much for the power used.

I'm hopeing my new box will beat that for a lot less power used.

Luck ................

 

[BillR]

There's only 4 CPU cores driving the 8 GPU cores.
If the nVidia client is anything like the ATI client, then 1x CPU core will be needed to dive 1x GPU core.
At a pure guess of 2,000-2,500 PpD per client, then its only 8,000-10,000 PpD.
Not that much for the power used.

I'm hopeing my new box will beat that for a lot less power used.

Luck ................

They have a few different units now, interesting stuff for sure:

http://www.nvidia.com/object/tesla_d870.html

http://www.nvidia.com/object/tesla_computing_solutions.html

http://www.nvidia.com/object/tesla_compatible_platforms.html

http://www.nvidia.com/object/cuda_home.html

Maybe we can get the [H] deal on a few dozen

 

[Diablo2K]

Do they even have waterblocks for any of the x2 video cards? I would imagine that it would not be easy to make one for them, but i have not seen the inside of one either.

 

[Lord von Schtupp]

Danger Den has one - 209 bucks

http://www.dangerden.com/store/product.php?productid=336&cat=0&page=

 

[Lord von Schtupp]

There's only 4 CPU cores driving the 8 GPU cores.
If the nVidia client is anything like the ATI client, then 1x CPU core will be needed to dive 1x GPU core.
At a pure guess of 2,000-2,500 PpD per client, then its only 8,000-10,000 PpD.
Not that much for the power used.

I'm hopeing my new box will beat that for a lot less power used.

Luck ................

A 3780 will get 2100+ ppd if the core can feed it 100% ( no cpu bottlenecking) and IF this info is to be believed;

http://www.nordichardware.com/news,7809.html

then a GX 2 should net ~5500 ppd if other factors are constant. Of course this is just speculation based on what I have seen...

 

[alan2308]

Speaking of Tesla, anyone know what potential there is for the GPU2 client on one of those? They're way out of my purchase league but I'm curious just the same.

According to Beyond3d:

The basic unit of the current Tesla line, the Tesla C870, should be very familiar to anyone who's seen the GeForce 8800. It's essentially an 8800 GTX--a 575MHz core clock and 128 SPs at 1.35GHz--with 1.5GiB of GDDR3 RAM. Of course, it's not quite an 8800 GTX--there are no display outputs at all on the card, even though it has a new version of the NVIO chip.

Other models differ in the number of GPUs they have onboard.

 

[leSLIe]

yeah but does it run Crysis @60 FPS with 16AF and 16AA ? i don't think so!

 

[Crazy Larry]

yeah but does it run Crysis @60 FPS with 16AF and 16AA ? i don't think so!

That's the next GPU Nvidia comes out with - The GeForce OMGWTFLOLBBQ

 

[Kendrak]

That's the next GPU Nvidia comes out with - The GeForce OMGWTFLOLBBQ

BBQ is right