Oh and of course accept that using GeForce means that NVIDIA consider you to be a criminal, accept it, move on and sleep more soundly knowing how much more wisely you spent precious taxpayer supported research funding. :-)
> On May 18, 2017, at 18:14, Scott Le Grand <varelse2005.gmail.com> wrote:
>
> Agreed, stick with the Founders Edition. Don't give NVIDIA any basis for
> their emerging narrative that GeForce GPUs can't be used for HPC because
> with deterministic code, they can...
>
> Admitting that most people (Google included) are utterly incapable of
> writing efficient deterministic code, but that's a different conversation.
>
> On Thu, May 18, 2017 at 2:50 PM, Ross Walker <ross.rosswalker.co.uk> wrote:
>
>> Short answer, In my experience: NO! Stick with the Founders edition.
>>
>> The custom cards tend to have fancy looking heatsinks which do not work
>> well in multi-GPU configurations, have many more failure points and are
>> often engineered for looks ahead of function and longevity. The exception
>> to this would be water cooled systems.
>>
>> The custom cards also tend to be overclocked - especially the memory - and
>> this leads to incorrect results, unstable behavior and/or early failure.
>>
>> Now if you want to play Crysis at 4K in stereo at 120FPS then by all means
>> go for the luminous blue overclocked card.
>>
>> All the best
>> Ross
>>
>>> On May 18, 2017, at 6:00 AM, Hadházi Ádám <hadadam.gmail.com> wrote:
>>>
>>> Dear All,
>>>
>>>
>>> We are about to build a new MD workstation with 3 x GTX 1080 TI GPUs.
>>>
>>>
>>> - Do you think it is worth to buy more expensive (custom) cards than the
>>> Founders Edition for greater reliability/better cooling/longer lifespan?
>>> (E.g. Asus ROG STRIX/TURBO series)
>>>
>>>
>>> Our machine would be running 24/7 (~ 90-95% load)
>>>
>>>
>>> Thanks,
>>> Adam
>>> _______________________________________________
>>> 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
>>
> _______________________________________________
> 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
Received on Thu May 18 2017 - 17:30:03 PDT
