Hi Divi,
regarding 1)
I am really not an expert here but in my opinion it should be possible
to down clock Titan to 732 MHz (after GPU bios editing using e.g. tools
like Kepler BIOS Tweaker, NVflash). But before you do the bios editing,
I would recommend to try check how Titan works under this ferquency
under Windows where you can change GPU frequency also without direct
bios editing (using some special software which sets the desaired
parameters
on the level of driver e.g. to use "MSI Afterburner" OR "EVGA precision"
...).
Again I am here not an expert but try to google a bit "overclocking Titan"
... (see some info e.g here
http://www.overclock.net/ ) to get
any additional info.
One more thing:
It seems that any card which have some base and boost frequency defined
is running under linux automatically under the boost frequency (tested on
my
Titans and GTX 680) so at least for linux work perhaps is important to set
the boost frequency to the desired value which you would like for your
linux work. You may check this e.g. by checking frequency info in mdout
Amber files.
Anyway just let's notice that playing with GPU bios may void your
warranty.
regarding 2)
Can you please do Amber benchmarks (at least 100K steps long, the same
random seed and repeat twice for each GPU) with your Titan cards to
check i) if they are able to successfully finish the tests
ii) if the results are prefectly reproducible i.e. exactly the same
results from
both runs on each GPU ?
Thanks in advance !
M.
Dne Mon, 03 Jun 2013 02:38:00 +0200 Divi/GMAIL <dvenkatlu.gmail.com>
napsal/-a:
> Hello Titanites:
>
> I have two questions as I follow this discussion thread..
>
> 1. Is it possible to down clock Titan to 732 MHz (as that is the
> speed of
> K20X) ?
>
> 2. Since I purchased two TITANS soon after they were out in March,
> lot of
> research work has been done in my lab using them. Most of my work
> involves
> MD refinement of protein-protein complexes with system sizes of 300-400K
> atoms (including waters, ions etc..). I have not seen anything weird or
> untrustworthy regarding the data analysis goes. I have not critically
> benchmarked the simulations and data analysis. While some differences in
> the
> energies appear when compared between CPU and GPU codes, they are still
> within acceptable limits (< 1-2%??).
>
> Can somebody throw some comments on research data integrity of using
> TITANS at this point, with all the discrepancies being reported?
>
> Apart from TITANS, I have GTX 650 Ti card (~1300 cores). Test
> calculations
> show that this card reproduces accurately every repeat of all benchmark
> sets. This card gives me 18 ns/day on FIX/NVE benchmark and it costs me
> $225
> on sale at newegg!
>
> Best
> Divi
>
>
>
> -----Original Message-----
> From: Scott Le Grand
> Sent: Sunday, June 02, 2013 3:49 PM
> To: ET
> Cc: AMBER Mailing List ; marek.maly.ujep.cz
> Subject: Re: [AMBER] experiences with EVGA GTX TITAN Superclocked -
> memtestG80 - UNDERclocking in Linux ?
>
> Definitely remove one of the GPUs and see what effect that has...
> That's a
> great datapoint.
>
>
> On Sun, Jun 2, 2013 at 11:02 AM, ET <sketchfoot.gmail.com> wrote:
>
>> Hi,
>>
>> Got the next bunch of perps coming into to stand in the lineup! :P
>>
>> Ran the long GB sims and they finished with no probs. Perfectly
>> replicability across all steps on both GPUs
>>
>> sdiff Files attached.
>>
>> I'm going to pull one of the cards out in half an hour and run GPU0 in
>> solo configuration to see if it makes a difference to the PME run. If
>> you
>> want me to run anything in the dual GPU config, let me know. :)
>>
>> br,
>> g
>>
>>
>> On 2 June 2013 18:15, Scott Le Grand <varelse2005.gmail.com> wrote:
>>
>>>
>>>
>>>
>>> On Sun, Jun 2, 2013 at 9:44 AM, Marek Maly <marek.maly.ujep.cz> wrote:
>>>
>>>
> _______________________________________________
> 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 8404
> (20130602) __________
>
> 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 Mon Jun 03 2013 - 05:00:03 PDT
