Yes indeed !
I am really curious here. Maybe cuFFT in CUDA 4.2.9 is somehow
"more tolerant"/"more compatible" with TITANs ...
Another hypothesis might be that the particular TITANs (ASUS) do not
have problems at all.
Tru if your TITANs pass all the tests without any problems
and with 100% reproducibility, might be possible for you to
repeat them with CUDA 5.0 or CUDA 5.5 and eventually
the actual driver (319.23) ?
If not or it is problematic for you, I will then try to downgrade
to CUDA 4.2.9 if I find proper install file.
BTW which is the actual driver version you use ?
(you can obtain it e.g. using nvidia-smi command).
Best,
Marek
Dne Wed, 03 Jul 2013 19:42:52 +0200 Ross Walker <ross.rosswalker.co.uk>
napsal/-a:
>
>> On Wed, Jul 03, 2013 at 10:13:45AM -0700, Scott Le Grand wrote:
>>> Did the final energies match? Were the final restarts bitwise
>>> identical?
>> yes to both -> I have diff'ed the mdout and md5sum the mdcrd/restrt
>> files.
>>
>> I am still using CUDA/4.2.9 for this run
>
> That might be an interesting data point. I don't think anyone has tried a
> titan with anything other than 5.0 or 5.5. That said the proof will be in
> the PME runs.
>
> All the best
> Ross
>
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
> __________ Informace od ESET NOD32 Antivirus, verze databaze 8521
> (20130703) __________
>
> Tuto zpravu proveril ESET NOD32 Antivirus.
>
> http://www.eset.cz
>
>
>
--
Tato zpráva byla vytvořena převratným poštovním klientem Opery:
http://www.opera.com/mail/
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jul 03 2013 - 11:00:04 PDT
