CUDA 2.2 still fails to enumerate devices on 9200/9400 Hybrid SLI system. "could not open the de

My system is a Dell XPS Studio 13 which features an MCP79 chipset with both a GeForce 9200M GS and GeForce 9400M G.

I have installed driver 185.18.08 and CUDA 2.2, but the deviceQuery SDK example does not detect any CUDA devices:

toojays@yoshi:~/NVIDIA_CUDA_SDK$ bin/linux/release/deviceQuery

CUDA Device Query (Runtime API) version (CUDART static linking)

NVIDIA: could not open the device file /dev/nvidia0 (Input/output error).

There is no device supporting CUDA.

Device 0: "Device Emulation (CPU)"

  CUDA Capability Major revision number:		 9999

  CUDA Capability Minor revision number:		 9999

  Total amount of global memory:				 4294967295 bytes

  Number of multiprocessors:					 16

  Number of cores:							   128

  Total amount of constant memory:			   65536 bytes

  Total amount of shared memory per block:	   16384 bytes

  Total number of registers available per block: 8192

  Warp size:									 1

  Maximum number of threads per block:		   512

  Maximum sizes of each dimension of a block:	512 x 512 x 64

  Maximum sizes of each dimension of a grid:	 65535 x 65535 x 1

  Maximum memory pitch:						  262144 bytes

  Texture alignment:							 256 bytes

  Clock rate:									1.35 GHz

  Concurrent copy and execution:				 No

  Run time limit on kernels:					 No

  Integrated:									Yes

  Support host page-locked memory mapping:	   Yes

  Compute mode:								  Default (multiple host threads can use this device simultaneously)

Test PASSED

Press ENTER to exit...

When I run this, the kernel logs the following:

[   56.301808] NVRM: RmInitAdapter failed! (0x31:0xffffffff:1022)

[   56.301815] NVRM: rm_init_adapter(0) failed

Anything I can do to make this work?

nvidia-bug-report.log.gz is attached.
nvidia_bug_report.log.gz (38.1 KB)

A workaround is to run:

sudo rm /dev/nvidia0 && sudo ln -s /dev/nvidia1 /dev/nvidia0

Now I get:

toojays@yoshi:~/NVIDIA_CUDA_SDK$ bin/linux/release/deviceQuery

CUDA Device Query (Runtime API) version (CUDART static linking)

There is 1 device supporting CUDA

Device 0: "GeForce 9400M G"

  CUDA Capability Major revision number:		 1

  CUDA Capability Minor revision number:		 1

  Total amount of global memory:				 265617408 bytes

  Number of multiprocessors:					 2

  Number of cores:							   16

  Total amount of constant memory:			   65536 bytes

  Total amount of shared memory per block:	   16384 bytes

  Total number of registers available per block: 8192

  Warp size:									 32

  Maximum number of threads per block:		   512

  Maximum sizes of each dimension of a block:	512 x 512 x 64

  Maximum sizes of each dimension of a grid:	 65535 x 65535 x 1

  Maximum memory pitch:						  262144 bytes

  Texture alignment:							 256 bytes

  Clock rate:									1.10 GHz

  Concurrent copy and execution:				 No

  Run time limit on kernels:					 Yes

  Integrated:									Yes

  Support host page-locked memory mapping:	   Yes

  Compute mode:								  Default (multiple host threads can use this device simultaneously)

Test PASSED

Press ENTER to exit...

Can anyone explain why CUDA reports 256M of global memory while my Xorg log reports 512M?

I’m in the same boat as you, I followed your suggestion and have the same result. Right now we only have the integrated chip doing our heavy work. The next step is how to get the dedicated (9200) to pick up some slack, even if not in hybrid mode, but just on its own.

I have the same problem, seems that the driver fails to enumerate two CUDA-capable cards.
Have anyone got any progress in that?

Offtopic: just curious - have NVIDIA guys ever tested their driver for CUDA? :) Two video cards is a quite typical setup…

uh, yes, we do test it. I don’t know the status of hybrid SLI, I’ll ask around.

Thank you. Btw I did not mean to offense anyone.

I tried to upgrade to 2.6.30, but did not succeed. Because old 180.22 does not compile on this kernel, and new 185.18.14 compile but fails on cuInit() with two different cards. Actually my setup is not hybrid sli, just two different nvidia cards (9200 and GTX260), first one for desktop and second for CUDA purposes.

that certainly shouldn’t fail… what motherboard are you using?

Asus P5Q Premium

lspci

01:00.0 VGA compatible controller: nVidia Corporation G86 [GeForce 8500 GT] (rev a1)
02:00.0 VGA compatible controller: nVidia Corporation GT200 [GeForce GTX 260] (rev a1)

oops, actually 8500 but not 9200. Anyway…

Maybe the problem is a relatively new kernel 2.6.30? But the 185.18.14 compiled without any problem, and the desktop (KDE4.2) worked as usual.

I used to assume this could be my dirty hands building 2.6.30, so I tried to reproduce the bug in the common environment.

Kernel 2.6.29-5, both drivers 180.22 and 185.18.14 compiles, both show desktop well, still CUDA SDK works in 180.22, and does not work in 185.18.14 (throwing out on cuInit(), saying ‘NVIDIA: could not open the device file /dev/nvidia1 (Input/output error)’ - just like reported above in the hybrid sli problem).

So I think it’s really your bug, guys.
Looking forwards for fixed driver for CUDA 2.2 :-)

Yes, me too, going to try out CUDA 2.3/ driver 190.18 tomorrow (stay away from 190.15 with this laptop guys! - fails to boot gnome, just hangs with a black screen, i’ve read that its the mobile graphics chips that are having the problems)

After being a reader of these forums for some time, now I feel like writing something… and it’s going to be a question:

Are there any updates regarding CUDA with Hybrid SLI on Dell Studio XPS 1340? My OS is Vista Ultimate 64 bit, the graphics is 9500M (9400MG+9200MGS). CUDA 2.3 works, but does not recognize both processors even when the system claims to have switched to the hybrid mode. Moreover, the OS itself doesn’t seem to make use of both even in the hybrid mode after installation of the beta driver from CUDA 2.3 (although both do appear in the device list).

Since it is not Linux in my case, I’m afraid it is some “general” driver problem… or the hardware problem? :)

(BTW, for CUDA 2.2, the display driver installer did not recognize the graphics hardware; brute force attempts failed too.)

Some time passed, is there any status on this?

The problem is:
a) new CUDA drivers are required for kernels 2.6.30+ (old ones do not compile with new kernel headers)
b) these new drivers fail to init second video card in a heterogenous two-card setup.

My current setup is: 2.6.31, 190.18. Compiled successfully, compositing works on video card #1, while CUDA fails on #2.

tmurray, could you help on this?

I should have been more attentive…
syslog showed “vmap allocation for size 16781312 failed: use vmalloc= to increase size.”

Setting boot parameter vmalloc=256M solved the issue.
For the reference, [url=“Chapter 9. Known Issues”]http://http.download.nvidia.com/XFree86/Li...chapter-09.html[/url]