Fudzilla
May 24, 2013, 03:06:32 AM *
Welcome, Guest. Please login or register.

Login with username, password and session length
News: Fudzilla Forum is open
 
   Home   Help Search Login Register  
Pages: [1]   Go Down
  Print  
Author Topic: four 9800GX2 cards: will it work?  (Read 2867 times)
kjoost
Newbie
*
Posts: 2


View Profile
« on: March 18, 2008, 10:55:25 PM »

I work as a Computer Scientist at an academic image processing department. In my research, I regularly work with nvidia Cuda http://www.nvidia.com/object/cuda_learn.html for high performance computing on the graphics card (GPGPU). These are general computations that have nothing to do with rendering 3D scenes. For GPGPU, SLI is not required. Therefore, it seems possible to me to put four 9800GX2 cards on a single motherboard, for example the MSI K9A2 Platinum V2 (http://global.msi.com.tw/index.php?func=proddesc&prod_no=1395&maincat_no=1) that has four (physical) PCI-Express X16 slots with double spacing between the slots.

So far, I have not found any information from others who have tried this before. Cooling and power are obvious problems that can probably be dealt with. Any ideas on problems that I can expect on the software side? For example, will the (non-SLI) driver see the four 9800GX2 cards as eight independent graphics cards? Or is there a maximum limit built-in?

Any comments would be appreciated!
Logged
Jon
Global Moderator
Full Member
*****
Posts: 234


swallow your pride.

audiofreak39@gmail.com AuDioFreaK39
View Profile WWW Email
« Reply #1 on: March 19, 2008, 01:33:06 AM »

Nvidia has no plans to support four physical 9800GX2 cards in any driver releases.  In my opinion, yes, it is possible to physically have 4 cards running at the same time on a Quad-PCI Express motherboard, however you cannot link them in SLi as the 9800GX2 bridge only allows for a maximum of Tri-SLi, or three cards.  Therefore, each card would be seen as an individual non-interlinked unit and drivers would have to be self-created.  Basically, there would be no point to this for CUDA development because I don't see any efficient means of calculating general purpose computations in parallel if the cards can't be interlinked.
« Last Edit: March 19, 2008, 01:36:00 AM by AuDioFreaK39 » Logged

Core i7 Extreme 965 {} EVGA X58 SLI Classified Limited Edition{} 6GB Mushkin XP Series DDR3 1600MHz 7-8-7-20 {} EVGA GeForce GTX 295 Plus [Quad-SLI] {} Thermalright TRUE Copper
kjoost
Newbie
*
Posts: 2


View Profile
« Reply #2 on: March 19, 2008, 08:07:42 AM »

Thanks! Cuda does not support SLI at the moment, so it cannot be used anyway. All communication between the cards must go from one card, through the CPU, back to the other card. However, for my applications, the computations can be efficiently spread among several cards without any communication between the cards.

Could you clarify what you mean by "self-created" drivers? Any idea on the maximum number of individual (non-communicating, non-SLI) GPU's that will be seen by the standard driver in WinXP?
Logged
Jon
Global Moderator
Full Member
*****
Posts: 234


swallow your pride.

audiofreak39@gmail.com AuDioFreaK39
View Profile WWW Email
« Reply #3 on: March 19, 2008, 06:33:03 PM »

Thanks! Cuda does not support SLI at the moment, so it cannot be used anyway. All communication between the cards must go from one card, through the CPU, back to the other card. However, for my applications, the computations can be efficiently spread among several cards without any communication between the cards.

Could you clarify what you mean by "self-created" drivers? Any idea on the maximum number of individual (non-communicating, non-SLI) GPU's that will be seen by the standard driver in WinXP?

By "self-created" I mean drivers that you make yourself, on your own, modified, unauthorized, uncertified, etc. lol.  I'm not exactly sure what the limit is for how many GPUs can be seen in the standard Windows XP VGA driver, but I'm assuming that it would have to be modified either way to allow for octal-GPU support.
Logged

Core i7 Extreme 965 {} EVGA X58 SLI Classified Limited Edition{} 6GB Mushkin XP Series DDR3 1600MHz 7-8-7-20 {} EVGA GeForce GTX 295 Plus [Quad-SLI] {} Thermalright TRUE Copper
aznstriker92
Full Member
***
Posts: 151



View Profile Email
« Reply #4 on: March 20, 2008, 12:18:13 AM »

Nvidia has no plans to support four physical 9800GX2 cards in any driver releases.  In my opinion, yes, it is possible to physically have 4 cards running at the same time on a Quad-PCI Express motherboard, however you cannot link them in SLi as the 9800GX2 bridge only allows for a maximum of Tri-SLi, or three cards.  Therefore, each card would be seen as an individual non-interlinked unit and drivers would have to be self-created.  Basically, there would be no point to this for CUDA development because I don't see any efficient means of calculating general purpose computations in parallel if the cards can't be interlinked.
I thought the 9800GX2 can't do tri sli but quad sli with 2 9800GX2s.
Logged

Running a X600 128mb, In need of a new graphics card badly......
Bristolbulldog
Newbie
*
Posts: 9


View Profile Email
« Reply #5 on: April 24, 2008, 12:38:43 PM »

I really don't see you having an issue here.

You don't need sli support.

All you need is for a pc to recognise four dual gpu cards, which it will do, as its hardware, and your asking it to assign a driver, irq's and what not.

Is there any way you can borrow these cards to physically plug them in.

I see your scenario similar to can i have two gx2's, and two hd3870x2's in a system.

Answer should be yes no problem, but with driver limitations of not being able to enable support for functions you do not even need (sli/ crossfire).

Logged
Pages: [1]   Go Up
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.11 | SMF © 2006-2009, Simple Machines LLC Valid XHTML 1.0! Valid CSS!