I am having significant stability problems running my CUDA app on my current HP laptop (8400M). I believe the problem to be related to the fact that I am using a display attached GPU as my CUDA processor. Does anyone know if the SLI mode on the Alienware dual GPU laptops can be disabled? Would this allow me to have two CUDA capable devices?
I am pretty sure that my problems are related to this issue since I do not have the same problems running the apps on a workstation w/ multiple non display attached GPUs. External Image
Do you see only 1 device when you run devicequery?
Oh, I don’t have a dual GPU laptop right now… I am considering buying one. The HP laptop that I am currently working on is a single GPU.
Oh ok. As far as I rememeber CUDA sees two devices when they are in SLI mode, so no problem.
Actually, isn’t it the other way around? With SLI on, CUDA only sees one device?
If SLI is enabled in the driver I thought. As far as I remember I saw 2 devices when the SLI-bridge was still on my 2 8800’s. Hmm, maybe someone who has more recent experience can comment. I am already happy at this time when I know what I did last week :blink:
Oh I see, you’re talking about the hardware connector, not the driver. Never mind then. You’re probably right. :)
Yes, I thought that was what moulik was afraid of. Although I have to warn: I am using Linux, I have no idea if you can switch off SLI in the windows drivers
To use multiple GPUs with CUDA, SLI must be disabled in the driver. The connector does not have to be removed.
Paulius