Re: [AMBER] T4 vs RTX2080Ti

From: David Cerutti <dscerutti.gmail.com>
Date: Fri, 12 Apr 2019 11:21:38 -0400

The T4 is going to be slow but run cool. It only takes 70W as opposed to
RTX-2080Ti's 250-300W envelope. So you could power four of those at the
same load as a single RTX-2080Ti. Each will probably give performance on
the order of 50-60% of an RTX-2080Ti, I'm guessing, perhaps even a little
more for a small system. The power consumption is no small thing,
either--it helps save the planet, but it also helps ensure that the load on
whatever server they're installed in does not get too high. Powering
multiple RTX cards in a server has been a problem. However, the T4's are
going to be $2K a pop--you're buying Tesla.

The new 1660Ti cards are another interesting option, but I haven't tested
them and the price per ns of MD, even the power envelop, looks like it will
merely keep pace with RTX-2080Ti, which is where the best deal for most of
us in a single card remains.

Dave


On Thu, Apr 11, 2019 at 11:03 AM Shenglong Wang <sw77.nyu.edu> wrote:

> Dell has benchmarking for T4
>
>
> https://www.dell.com/support/article/us/en/19/sln316570/hpc-applications-performance-with-turing?lang=en
> <
> https://www.dell.com/support/article/us/en/19/sln316570/hpc-applications-performance-with-turing?lang=en
> >
>
> Best,
> Shenglong
>
> > On Apr 11, 2019, at 11:00 AM, Giovanni Grazioso <
> giovanni.grazioso.unimi.it> wrote:
> >
> > Dear Amber Users,
> >
> > I'd like to buy a new GPU card to run calculations by pmemd.cuda.
> >
> > The new NVIDIA T4 Turing GPUs are interesting but I have no idea of
> > ​​their performance in comparison with the RTX 2080Ti.
> >
> > Has anyone tried the T4 GPU for MD simulations by pmemd.cuda?
> >
> > Thanks in advance for your hints.
> >
> > Giovanni
> >
> >
> > --
> >
> >
> >
> >
> >
> > *****************************************************************
> > Giovanni GRAZIOSO, PhD
> > Università degli Studi di Milano (Italy)
> > E-mail: giovanni.grazioso.unimi.it
> > *****************************************************************
> >
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> >
> https://urldefense.proofpoint.com/v2/url?u=http-3A__lists.ambermd.org_mailman_listinfo_amber&d=DwIGaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=gI7YPDDMIksUN29Y65_UCg&m=mec-c_baY84eoHiB3kLfmKhBrRIq2O4s8gh-8Q6uHKw&s=3Q-jadUviqnAAGgUAg2PheOagWwZ10ZOkJbh8jyCI1A&e=
>
> _______________________________________________
> 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 Fri Apr 12 2019 - 08:30:02 PDT
Custom Search