From: Renfro, Michael (
Date: Mon Feb 01 2021 - 16:17:09 CST

For 2.14, the last paragraph of [1] indicates the binary was built against CUDA 10.1. If so, the compatibility matrix below indicates you'll need driver 418.39 for the current release of 10.1. The driver search at [2] indicates a P5000 can support up through driver 460.39.


From: <> on behalf of Prince, Chandler <>
Date: Monday, February 1, 2021 at 3:40 PM
To: <>, Seibold, Steve Allan <>
Subject: namd-l: Re: NAMD CUDA

External Email Warning

This email originated from outside the university. Please use caution when opening attachments, clicking links, or responding to requests.

On the NVIDIA site, they have a handy table for CUDA Compatibility. The trouble is knowing what version of CUDA NAMD was compiled with. If someone tells me or you how to determine that, knowing what Driver is needed becomes easy.;!!DZ3fjg!s-0deLC7V9kmcG2aue_yJSo4jAN9KWDFPxRhHyr37DuFrhKn0ZKgHh2uMl6SUJUcWA$ <*3A*2F**2Fv3*2F__https*3A*2F**2Fdeploy*2Fcuda-compatibility*2Findex.html__*3B!!DZ3fjg!pSWhg-vZZtlXZJJSqAxxcvIjJrs0HWSNsDQoi2ji74SVHykHrFOZMi4mvozSB70tiQ*24&data=04*7C01*7CRenfro**7C21da45b47ed746ebbc1808d8c6f9ee1a*7C66fecaf83dc04d2cb8b8eff0ddea46f0*7C1*7C0*7C637478124042615540*7CUnknown*7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0*3D*7C2000&sdata=bJxrgY3WJ0*2FWU8e6XLdGs*2BPzQUQmkIoG6ANVvd74I5g*3D&reserved=0__;JSUlJSUlJSUlJSUlJSUlJSUlJSUlJSUlJSUlJQ!!DZ3fjg!s-0deLC7V9kmcG2aue_yJSo4jAN9KWDFPxRhHyr37DuFrhKn0ZKgHh2uMl7wvF-qpA$ >

See Table 1. CUDA Toolkit and Compatible Driver Versions

CUDA 10.0 (10.0.130)
>= 410.48
From: <> on behalf of Seibold, Steve Allan <>
Sent: Monday, February 1, 2021 1:50 PM
To: <>
Subject: namd-l: NAMD CUDA


I have down loaded the binary cuda version of NAMD, NAMD_2.14_x86_64-multicore-CUDA onto my Ubuntu (16.04) machine. Using the “nvidia-smi”, I found my GPU card Quadro P5000, NIVIDIA-SMI 415.27, Driver Version: 415.27 and the CUDA vesion 10.0.

When I attempt to run namd2 get the following: “CUDA driver version is insufficient for CUDA runtime version”…I’m a novice at this, but, is this a barrier that cannot be overcome?

Thanks, Steve

This archive was generated by hypermail 2.1.6 : Fri Dec 31 2021 - 23:17:10 CST