CUDA error at bandwidthTest

Dear All,

I got installed a 2 NVIDIA tesla M60 and I got the right PASSED test from deviceQuery, see logs below,
still when I launch the bandwidthTest i got the follwoing error:

[CUDA Bandwidth Test] - Starting…
Running on…

Device 0: Tesla M60
Quick Mode

CUDA error at bandwidthTest.cu:730 code=46(cudaErrorDevicesUnavailable) “cudaEventCreate(&start)”

any idea hot to solve this ?

Thanks,

Antonio

Logs:
Linux s-jrciprhpc103p 3.10.0-957.5.1.el7.x86_64 #1 SMP Fri Feb 1 14:54:57 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux

cat /proc/driver/nvidia/version
NVRM version: NVIDIA UNIX x86_64 Kernel Module 418.39 Sat Feb 9 19:19:37 CST 2019
GCC version: gcc version 4.8.5 20150623 (Red Hat 4.8.5-36) (GCC)

./deviceQuery
./deviceQuery Starting…

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

Detected 4 CUDA Capable device(s)

Device 0: “Tesla M60”
CUDA Driver Version / Runtime Version 10.1 / 10.1
CUDA Capability Major/Minor version number: 5.2
Total amount of global memory: 8129 MBytes (8524136448 bytes)
(16) Multiprocessors, (128) CUDA Cores/MP: 2048 CUDA Cores
GPU Max Clock rate: 1178 MHz (1.18 GHz)
Memory Clock rate: 2505 Mhz
Memory Bus Width: 256-bit
L2 Cache Size: 2097152 bytes
Maximum Texture Dimension Size (x,y,z) 1D=(65536), 2D=(65536, 65536), 3D=(4096, 4096, 4096)
Maximum Layered 1D Texture Size, (num) layers 1D=(16384), 2048 layers
Maximum Layered 2D Texture Size, (num) layers 2D=(16384, 16384), 2048 layers
Total amount of constant memory: 65536 bytes
Total amount of shared memory per block: 49152 bytes
Total number of registers available per block: 65536
Warp size: 32
Maximum number of threads per multiprocessor: 2048
Maximum number of threads per block: 1024
Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
Max dimension size of a grid size (x,y,z): (2147483647, 65535, 65535)
Maximum memory pitch: 2147483647 bytes
Texture alignment: 512 bytes
Concurrent copy and kernel execution: Yes with 2 copy engine(s)
Run time limit on kernels: No
Integrated GPU sharing Host Memory: No
Support host page-locked memory mapping: Yes
Alignment requirement for Surfaces: Yes
Device has ECC support: Disabled
Device supports Unified Addressing (UVA): Yes
Device supports Compute Preemption: No
Supports Cooperative Kernel Launch: No
Supports MultiDevice Co-op Kernel Launch: No
Device PCI Domain ID / Bus ID / location ID: 0 / 131 / 0
Compute Mode:
< Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

Device 1: “Tesla M60”
CUDA Driver Version / Runtime Version 10.1 / 10.1
CUDA Capability Major/Minor version number: 5.2
Total amount of global memory: 8129 MBytes (8524136448 bytes)
(16) Multiprocessors, (128) CUDA Cores/MP: 2048 CUDA Cores
GPU Max Clock rate: 1178 MHz (1.18 GHz)
Memory Clock rate: 2505 Mhz
Memory Bus Width: 256-bit
L2 Cache Size: 2097152 bytes
Maximum Texture Dimension Size (x,y,z) 1D=(65536), 2D=(65536, 65536), 3D=(4096, 4096, 4096)
Maximum Layered 1D Texture Size, (num) layers 1D=(16384), 2048 layers
Maximum Layered 2D Texture Size, (num) layers 2D=(16384, 16384), 2048 layers
Total amount of constant memory: 65536 bytes
Total amount of shared memory per block: 49152 bytes
Total number of registers available per block: 65536
Warp size: 32
Maximum number of threads per multiprocessor: 2048
Maximum number of threads per block: 1024
Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
Max dimension size of a grid size (x,y,z): (2147483647, 65535, 65535)
Maximum memory pitch: 2147483647 bytes
Texture alignment: 512 bytes
Concurrent copy and kernel execution: Yes with 2 copy engine(s)
Run time limit on kernels: No
Integrated GPU sharing Host Memory: No
Support host page-locked memory mapping: Yes
Alignment requirement for Surfaces: Yes
Device has ECC support: Disabled
Device supports Unified Addressing (UVA): Yes
Device supports Compute Preemption: No
Supports Cooperative Kernel Launch: No
Supports MultiDevice Co-op Kernel Launch: No
Device PCI Domain ID / Bus ID / location ID: 0 / 132 / 0
Compute Mode:
< Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

Device 2: “Tesla M60”
CUDA Driver Version / Runtime Version 10.1 / 10.1
CUDA Capability Major/Minor version number: 5.2
Total amount of global memory: 8129 MBytes (8524136448 bytes)
(16) Multiprocessors, (128) CUDA Cores/MP: 2048 CUDA Cores
GPU Max Clock rate: 1178 MHz (1.18 GHz)
Memory Clock rate: 2505 Mhz
Memory Bus Width: 256-bit
L2 Cache Size: 2097152 bytes
Maximum Texture Dimension Size (x,y,z) 1D=(65536), 2D=(65536, 65536), 3D=(4096, 4096, 4096)
Maximum Layered 1D Texture Size, (num) layers 1D=(16384), 2048 layers
Maximum Layered 2D Texture Size, (num) layers 2D=(16384, 16384), 2048 layers
Total amount of constant memory: 65536 bytes
Total amount of shared memory per block: 49152 bytes
Total number of registers available per block: 65536
Warp size: 32
Maximum number of threads per multiprocessor: 2048
Maximum number of threads per block: 1024
Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
Max dimension size of a grid size (x,y,z): (2147483647, 65535, 65535)
Maximum memory pitch: 2147483647 bytes
Texture alignment: 512 bytes
Concurrent copy and kernel execution: Yes with 2 copy engine(s)
Run time limit on kernels: No
Integrated GPU sharing Host Memory: No
Support host page-locked memory mapping: Yes
Alignment requirement for Surfaces: Yes
Device has ECC support: Disabled
Device supports Unified Addressing (UVA): Yes
Device supports Compute Preemption: No
Supports Cooperative Kernel Launch: No
Supports MultiDevice Co-op Kernel Launch: No
Device PCI Domain ID / Bus ID / location ID: 0 / 195 / 0
Compute Mode:
< Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

Device 3: “Tesla M60”
CUDA Driver Version / Runtime Version 10.1 / 10.1
CUDA Capability Major/Minor version number: 5.2
Total amount of global memory: 8129 MBytes (8524136448 bytes)
(16) Multiprocessors, (128) CUDA Cores/MP: 2048 CUDA Cores
GPU Max Clock rate: 1178 MHz (1.18 GHz)
Memory Clock rate: 2505 Mhz
Memory Bus Width: 256-bit
L2 Cache Size: 2097152 bytes
Maximum Texture Dimension Size (x,y,z) 1D=(65536), 2D=(65536, 65536), 3D=(4096, 4096, 4096)
Maximum Layered 1D Texture Size, (num) layers 1D=(16384), 2048 layers
Maximum Layered 2D Texture Size, (num) layers 2D=(16384, 16384), 2048 layers
Total amount of constant memory: 65536 bytes
Total amount of shared memory per block: 49152 bytes
Total number of registers available per block: 65536
Warp size: 32
Maximum number of threads per multiprocessor: 2048
Maximum number of threads per block: 1024
Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
Max dimension size of a grid size (x,y,z): (2147483647, 65535, 65535)
Maximum memory pitch: 2147483647 bytes
Texture alignment: 512 bytes
Concurrent copy and kernel execution: Yes with 2 copy engine(s)
Run time limit on kernels: No
Integrated GPU sharing Host Memory: No
Support host page-locked memory mapping: Yes
Alignment requirement for Surfaces: Yes
Device has ECC support: Disabled
Device supports Unified Addressing (UVA): Yes
Device supports Compute Preemption: No
Supports Cooperative Kernel Launch: No
Supports MultiDevice Co-op Kernel Launch: No
Device PCI Domain ID / Bus ID / location ID: 0 / 196 / 0
Compute Mode:
< Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

Peer access from Tesla M60 (GPU0) -> Tesla M60 (GPU1) : Yes
Peer access from Tesla M60 (GPU0) -> Tesla M60 (GPU2) : No
Peer access from Tesla M60 (GPU0) -> Tesla M60 (GPU3) : No
Peer access from Tesla M60 (GPU1) -> Tesla M60 (GPU0) : Yes
Peer access from Tesla M60 (GPU1) -> Tesla M60 (GPU2) : No
Peer access from Tesla M60 (GPU1) -> Tesla M60 (GPU3) : No
Peer access from Tesla M60 (GPU2) -> Tesla M60 (GPU0) : No
Peer access from Tesla M60 (GPU2) -> Tesla M60 (GPU1) : No
Peer access from Tesla M60 (GPU2) -> Tesla M60 (GPU3) : Yes
Peer access from Tesla M60 (GPU3) -> Tesla M60 (GPU0) : No
Peer access from Tesla M60 (GPU3) -> Tesla M60 (GPU1) : No
Peer access from Tesla M60 (GPU3) -> Tesla M60 (GPU2) : Yes

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 10.1, CUDA Runtime Version = 10.1, NumDevs = 4
Result = PASS

Is nvidia-persistenced running? If not, does starting it help?

Many Thanks for the help !!!

It was not running.
I started with following commands:
sudo systemctl enable nvidia-persistenced
sudo /usr/bin/nvidia-persistenced --verbose

I launched the bandwidthTest

I got same error:

./bandwidthTest
[CUDA Bandwidth Test] - Starting…
Running on…

Device 0: Tesla M60
Quick Mode

CUDA error at bandwidthTest.cu:730 code=46(cudaErrorDevicesUnavailable) “cudaEventCreate(&start)”

any other things I shoud do or I missed.

Antonio

nvidia-bug-report.log.gz (1.91 MB)

Please run nvidia-bug-report.sh as root and attach the resulting .gz file to your post. Hovering the mouse over an existing post of yours will reveal a paperclip icon.
https://devtalk.nvidia.com/default/topic/1043347/announcements/attaching-files-to-forum-topics-posts/

Dear All,

Many thanks again. File posted as requested.

Antonio

The Teslas are in graphics mode, please switch them to compute mode:
https://docs.nvidia.com/grid/latest/grid-gpumodeswitch-user-guide/index.html

I can not donwload the software from the Enterprise Portal.
Could you send me only the packet gpumodeswitch ?

No, sorry, looks like it’s only included in the vGPU software package. Might also be a red herring. Does bandwidthtest work if you just run it on one gpu, e.g using options --device=0 --dtoh

Thanks for help me. Much appreciated !!!
Nope. The bandwidthtest it doesn’t work.

But I purchased the cards for computing, in principle i should be able to change it. Is there any way to reset the cards to defaults parameters…I need only to have this 2 cards working for computing and I don’t need this vGPU stuff.

Logs:

./bandwidthTest --device=0 --dtoh
[CUDA Bandwidth Test] - Starting…
Running on…

Device 0: Tesla M60
Quick Mode

CUDA error at bandwidthTest.cu:618 code=46(cudaErrorDevicesUnavailable) “cudaEventCreate(&start)”

Just to give you 2 more hints about the issue…

I compiled and run a simple matrixmul and it doesn’t work either:

./matrixMul
[Matrix Multiply Using CUDA] - Starting…
GPU Device 0: “Tesla M60” with compute capability 5.2

MatrixA(320,320), MatrixB(640,320)
CUDA error at matrixMul.cu:152 code=46(cudaErrorDevicesUnavailable) “cudaMalloc(reinterpret_cast<void **>(&d_A), mem_size_A)”

and also Add a vector:

./vectorAdd
[Vector addition of 50000 elements]
Failed to allocate device vector A (error code all CUDA-capable devices are busy or unavailable)!

Even the bandwidthtest sample alone is not that complex to easily fail. Maybe this is just a regression in cuda 10.1 or the driver, did you try with an earlier toolkit, 10.0 or 9.2? In any case you should mail the bug-report.log together with a description of the problem to linux-bugs[at]nvidia.com . Maybe that brings up something new.

Again many thanks for your help. I sent the email to linux-bugs@nvidia.com.

Hi, Antonio, is now the problem solved? I met the same error.

I got the same error with Tesla P4.
I’ve tried with CUDA toolkit 7.5, 8.0 and 9.0.
Driver version: 410.107
OS: Rhel 7.4
Kernel: 3.10
The server is created on a virtualization platform like Xenserver, does it need additional configuration compared to that created on physical machine?
Any help would be appreatiated.