Hi Josmar,
> I have built Amber 10 pmemd.cuda using a Nvidia GTS250 and everything
> went fine. However, when I run pmemd.cuda tests I get the following
> message:
>
> GPU SM revision must be 1.3 or better.
> ./Run.igb1_ntc1_min: Program error
> make: ** [test.pmemd.cuda] Erro 1
>
> Does that mean my card is not supported? This "SM" seems to mean Shader
> Model, and according to the video card specification it supports Shader
> Model 4.0. I would appreciate if somone could help me understand what
> that mean. Thanks in advance.
Yes this unfortunate but the GTS250 is an economy part that doesn't support double precision. The way to check this is to run the deviceQuery command from the CUDA_SDK. This will give you the hardware major and minor revision numbers which need to be 1.3 or 2.0 and greater. Most of the cards prior to the C1060 line are 1.2 or 1.1 which does not include double precision support. The annoying thing is, and I think the GTS250 falls under this, is that there is a hardware revision termed 1.5 which actually supports LESS than 1.3 since it does not include double precision. This is somewhat annoying marketing but alas there is nothing that can be done here. I am told that from Fermi (which is 2.0) and onwards the hardware revision info will be a lot clearer.
In the meantime I will update the webpage to specifically list the GTS250 as unsupported.
All the best
Ross
/\
\/
|\oss Walker
| Assistant Research Professor |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
|
http://www.rosswalker.co.uk |
http://www.wmd-lab.org/ |
Note: Electronic Mail is not secure, has no guarantee of delivery, may not be read every day, and should not be used for urgent or sensitive issues.
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu May 20 2010 - 15:00:03 PDT
