Is GeForce 9400M cuda-enabled?

As the topic title.
Thanks.

Certainly should be, although I haven’t been able to try one yet. (I have to say, I really want one of the new MacBook Airs…)

certainly looks like it, check out 9100m nvidia motherboard

from that it appears then that the macbook has either the 9200m or the 9300m graphics chip…

More questions from my side:

How much of the Macbook Pro#s 256MB respectively 512MB GDDR3 memory for the new nVidia Geforce 9600M GT chips will be available to CUDA?

Can CUDA be used simultaneously on the onboard Geforce 9400M and the dedicated 9600M GT chip of a Macbook Pro? If not, how is the switching managed? Does Mac OS X perform this switching automatically or is there a control panel applet for switching the graphics over?

Will Linux recognize both of graphics chips and can they be used simultaneously?

Anyone got answers yet?

Christian

Official answer: Yes, the chips support CUDA, but no, the driver does not support CUDA out of the box. We will support it ASAP, though.

When do you think you will have support of it?

Thanks.

I would hope that it’s available by 2.1 final, but no promises.

ideally a Macbook Pro would be able to use it’s 9400M for graphics display while performing CUDA operations on the “unattached” 9600M.

any idea if this is something that will be possible with the new driver?

thanks! i’m getting excited to do some CUDA on Mac.

Haven’t the foggiest idea. Once I have one to try, I’ll let you know.

Now the next question, can CUDA use BOTH the 9400M and 9600M GT at once? It sounds like no, it’s something where one GPU is shut down completely.
http://www.ubergizmo.com/15/archives/2008/…ike_pcs_do.html

But of course later software may change that… it’d be nice to get the extra crunch power.

i’m not sure you would want to use the 9400M with the 9600M, as i assume one would remain “connected” to a display, limiting the execution time on that GPU. that’s why i asked about running CUDA on the non-connected 9600M - seems ideally situated to use it for CUDA if not driving a display.

I have a new macbook and have cuda running on it … but it not that fast/ pretty slow … the CPU is even faster than the GPU on scanLargeArray example…

How did you get it running? I have a new macbook pro with the 9400M and 9600M GT and I installed the toolkit and the SDK but no luck so far. I tried both GPUs and I only get “Device Emulation” when doing a deviceQuery… what is the trick? The only thing I haven’t tried is using boot camp to test under windows.

Try installing the toolkit again. Make sure you click “Customize” and select the “kext” thingy. It’s unselected by default … and its description is … “It allows you to run CUDA”, NVIDIA … :-P

Salivian - is it a MB or MBP?

still wondering about running on the unconnected GPU, e.g. can you specify the other device?

Ah thanks! Any idea why this essential component is not selected by default?

I got both the 9400M and the 9600M GT running on CUDA under Mac OS X. When I’m in “Better battery life” mode I can only see the 9400M but if switch to “High performance” and restart then I can see both when I do a device query. What is odd though is that I get low numbers for the clock rate when in high performance mode (0.34GHz for the 9600M GT and 0.25GHz for the 9400M). I don’t remember exactly what was the clock rate when I only had the 9400M active but it was certainly higher.

I also tested under Windows Vista and, although I could only see the 9600M GT there, I got much better numbers… the clock rate was much higher, and the memory bandwidth also. In Windows my memory bandwidth in pinned mode was over 3000 MB/s, while in Mac OS X it is only around 1300 MB/s.

Maybe someone from NVIDIA can clarify if these performance figures will be more consistent in future driver releases.

how do these numbers compare to the previous Macbook Pr0, with the 8600M?