Re: CUDA limitations

From: Ajasja Ljubetič (
Date: Fri Jun 24 2011 - 07:37:13 CDT


I wanted to put this info on the NAMD wiki, but it kept telling me to go

"Unknown IP address. Please go away"

But rigth below it says:
*"YOU* can edit, extend, or even create any page in this site by clicking on
the edit button."

So now I'm a bit confused as to what I'm doing wrong.

Best regards,

On Thu, May 12, 2011 at 19:45, Jim Phillips <> wrote:

> Here are quick answers with some explanation:
> ABF, meta-dynamics - yes (doesn't alter nonbonded interactions)
> useFlexibleCell, useConstantRatio, useConstantArea
> - yes since NAMD 2.7 (pressure tensor is now calculated on GPU)
> accelerated MD - no (requires energy, which is not accelerated) but
> should work when energy evaluation on the GPU is implemented
> Martini force field - martiniSwitching is broken in both CPU and GPU,
> but yes when the bug is fixed, but coarse grained systems won't
> see much of a performance boost from the GPU because there just
> aren't that many particles. You could benchmark it as-is, though.
> implicit solvent - no (requires special nonbonded interactions) but
> certainly a candidate for the GPU in a future release
> -Jim
> On Thu, 12 May 2011, Ajasja Ljubeti~M wrote:
> Dear all,
>> I'm wondering if there is a list of NAMD features not supported by CUDA
>> builds. Currently the manual just states
>> "Some features are unavailable in CUDA builds, including alchemical free
>> energy perturbation."
>> I think that the implicit solvent is also not currently supported.
>> I am most interested in ABF and meta-dynamics in membranes.
>> Does ABF work? What about useFlexibleCell and useConstantRatio or
>> useConstantArea?
>> What about accelerated
>> MD<**Research/namd/2.8b1/ug/node63.**html>
>> >?
>> And the new Martrini force field?
>> Thank you & best regards,
>> Ajasja LjubetiĨ

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:20:29 CST