Re: [AMBER] cudaGetDeviceCount failed unknown error

From: Ercig Bogac (BIOCH) <"Ercig>
Date: Thu, 23 Nov 2017 22:29:35 +0000

Dear Taisung,

Thank you very much or your suggestion. I see the "arch = sm_20" in my output as below

Fatbin elf code:
arch = sm_20
code version = [1,7]
producer = cuda
host = linux
compile_size = 64bit

Should I give a try on the solution 1 anyhow ? The sm version for Tesla C2075 that I am using, I see sm_20 version.

Thanks again your help.

Best regards

From: Taisung Lee <>
Sent: 23 November 2017 23:18
To: 'AMBER Mailing List'
Subject: Re: [AMBER] cudaGetDeviceCount failed unknown error

Hi Ericg,

        My guess is that your nvcc8.0-compiled pmemd.cuda cannot run on your
C2075 GPU due to the fact that nvcc8.0 will not produce sm_20 code, which is
required for your C2075, an sm_20 GPU.

        You can check by doing:

"cuobjdump $AMBERHOME/bin/pmemd.cuda"

        If you do not see " arch = sm_20" in your output, then likely this
is the reason.....

        Solutions: 1. Use nvcc 7.5 to compile AMBER; or 2. Run on a newer
GPU (w/ sm>30). You can go to to
check the sm version of your GPU.


-----Original Message-----
From: Ercig Bogac (BIOCH) []
Sent: Thursday, November 23, 2017 8:30 AM
Subject: [AMBER] cudaGetDeviceCount failed unknown error

Dear all,

I am trying to install Amber16 to my computer with the CUDA support. All
steps of the installation went without any errors.

I have used ./configure -cuda gnu

Then make install

But when I try the make test , I get the following error.

cudaGetDeviceCount failed unknown error

0 file comparisons passed
0 file comparisons failed
333 tests experienced errors

I cant find any solutions on internet. I have seen one entry on the mail
list, but there was no solution for this case. I have updated my driver and
reinstalled cuda but it did not changed anything.

Here is what I obtain from nvidia-smi

Thu Nov 23 14:27:12 2017
| NVIDIA-SMI 384.81 Driver Version: 384.81
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr.
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute
M. |
| 0 Tesla C2075 Off | 00000000:01:00.0 On |
Off |
| 30% 57C P12 34W / N/A | 470MiB / 6066MiB | 9%
Default |

| Processes: GPU
Memory |
| GPU PID Type Process name Usage
| No running processes found

And the nvcc -v

nvcc warning : The 'compute_20', 'sm_20', and 'sm_21' architectures are
deprecated, and may be removed in a future release (Use
-Wno-deprecated-gpu-targets to suppress warning).
nvcc fatal : No input files specified; use option --help for more

I am looking forward to hear your help

Best regards

Bogac Ercig
AMBER mailing list

AMBER mailing list

AMBER mailing list
Received on Thu Nov 23 2017 - 14:30:03 PST
Custom Search