[gmx-users] Videocard selection

[Szilárd Páll]

TL;DR if cost is no issue Turing is better.

The "core" count is not the only than matters, #cores x frequncy (~
instruction rate) is the metric that does have some predicting power within
a single GPU generation (but even then, the clocks a consumer GPU runs at
can be wildly different from the clocks in the specs -- especially if you
cool them well).

However, often you can't compare across generations: case in point, the
Volta and Turing GPUs got a lot faster due to microarchitectural
improvements and therefore the 20xx series are quite a good option.

--
Szilárd


On Tue, Mar 12, 2019 at 3:54 PM <pbuscemi at q.com> wrote:

> 3500 cores vs 4500 cores and
>
> See
> http://hwbench.com/vgas/geforce-rtx-2080-ti-vs-geforce-gtx-titan-xp
>
> And
>
>
> https://gpu.userbenchmark.com/Compare/Nvidia-Titan-Xp-vs-Nvidia-RTX-2080-Ti/m265423vs4027
>
> However, if you must have support, then it must be titan
>
> -----Original Message-----
> From: gromacs.org_gmx-users-bounces at maillist.sys.kth.se <
> gromacs.org_gmx-users-bounces at maillist.sys.kth.se> On Behalf Of Benson
> Muite
> Sent: Tuesday, March 12, 2019 8:57 AM
> To: gromacs.org_gmx-users at maillist.sys.kth.se
> Subject: Re: [gmx-users] Videocard selection
>
> Hi!
>
> For most applications single precision performance is most important - may
> want to check if this will be fine or your workflow.
>
> An older study is at:
>
> https://arxiv.org/pdf/1507.00898.pdf
>
> Regards,
>
> Benson
>
> On 3/12/19 2:18 PM, Никита Шалин wrote:
> > Dear Gromacs users,
> >
> > I would like to buy a videocard for calculation on GPU. I choose between
> RTX 2080Ti and Titan Xp. Please, tell me which videocard to choose? And
> what characteristics of the videocard are important for calculating?
> >
> > I'm doing modeling of copolymer system with an applied electric field.
> >
> > Thank you for advance!
> >
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.