Cellulose NVE on C2050 is currently 1.16 ns/day
Cellulose NVE on C1060 is currently 0.72 ns/day
1.16/0.72~=1.6x
The patch isn't checked in yet as it's wrapped up in the multi-gpu work. It will be checked in this week...
-----Original Message-----
From: Marek Maly [mailto:marek.maly.ujep.cz]
Sent: Monday, July 12, 2010 09:50
To: AMBER Mailing List
Cc: andrea.danani.supsi.ch; massimo maiolo
Subject: Re: [AMBER] The system size limitations for Tesla C2050 ?
Dear Scott,
thank you very much for both your reactions. Especially that information
that also with C2050 is possible to calculate systems around 400K
(the cellulose benchmark) after some patch application is very valuable !
Could you please send some more information (the proper web link is
enough) about the patch after which single
C2050 was able calculate 400K atom system (i.e. the cellulose benchmark ) ?
You are rigt the speedup is (speaking about the explicit solvent Amber
calc.) from cca 40 to 100% according
to the relevant benchmark:
http://ambermd.org/gpus/benchmarks.htm
From that benchmark is evident that speedup is strongly dependent on
system size (with higher size the speedup is decreasing).
Thanks again !
Best wishes,
Marek
Dne Mon, 12 Jul 2010 18:13:05 +0200 Scott Le Grand <SLeGrand.nvidia.com>
napsal/-a:
> Please don't mix C1060 and C2050 in the same run. It'll work but it's a
> load-balancing nightmare... Where are you only seeing a 30% speed
> boost? It should be at least 1.5x to 1.7x...
>
>
>
> -----Original Message-----
> From: Marek Maly [mailto:marek.maly.ujep.cz]
> Sent: Monday, July 12, 2010 06:59
> To: AMBER Mailing List
> Subject: Re: [AMBER] The system size limitations for Tesla C2050 ?
>
> Dear Dairsie,
>
> thanks for your information. I know C2070 will have 6GB of memory which
> will be a nice improvement comparing C2050 which has just 3GB but will be
> also a
> little bit more expensive and moreover will be available in Europe cca
> from October.
>
> Maybe that for this moment is still better to buy C1060 (with 4GB memory)
> than C2050 however
> C2050 seems to be cca 30%-40% faster (explicit solvent) or to consider
> some hybrid solution
> if one wants to buy let say 4 GPU system (2xC1060 + 2x2050). Of course
> this is probably not
> optimal for eventual GPU paralelisation higher than 2 GPUs.
>
> Anyway thanks again !
>
> Best wishes,
>
> Marek
>
>
>
>
> Dne Mon, 12 Jul 2010 11:04:12 +0200 Dairsie Latimer
> <dairsie.petapath.com>
> napsal/-a:
>
>> Marek,
>>
>> I don't know the exact number but it is linked to the board memory
>> size.
>> The C1060 has 4GB and can run the Cellulose NVE test case so my best
>> guess
>> at this point would be that the C2050 will be in the 275K to 300K region
>> (3GB without ECC turned on). The C2070 is going to provided a good
>> increase
>> here and going multi-card will also probably allow the simulation of
>> larger
>> systems.
>>
>> Dairsie
>>
>> On 12 July 2010 09:37, Marek Maly <marek.maly.ujep.cz> wrote:
>>
>>> Dear all,
>>>
>>> does anybody know the system size limitation (number of atoms)
>>> for Amber calculations in explicit solvent on Tesla C2050 GPU ?
>>>
>>>
>>> From this Amber benchmark
>>>
>>> http://ambermd.org/gpus/benchmarks.htm
>>>
>>> is only evident that 100k atoms is OK but 400k not but
>>> there is a big gap between this numbers so the real "border"
>>> is not clear.
>>>
>>>
>>> Thanks in advance for any comments/experiences regarding to this
>>> GPU.
>>>
>>> Best wishes,
>>>
>>> Marek
>>>
>>>
>>> _______________________________________________
>>> AMBER mailing list
>>> AMBER.ambermd.org
>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>> __________ Informace od ESET NOD32 Antivirus, verze databaze 5270
>> (20100711) __________
>>
>> 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
-----------------------------------------------------------------------------------
This email message is for the sole use of the intended recipient(s) and may contain
confidential information. Any unauthorized review, use, disclosure or distribution
is prohibited. If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message.
-----------------------------------------------------------------------------------
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Jul 12 2010 - 10:30:05 PDT
