Yep, during the last generations (k20 - k80, p100, v100) we got
computation-wise always a speedup one would expect, proportional to the
theoretical flop/s of the different models.

Regarding your off-list question on Hypnos K20: they only have 5 GByte
and prior to P100 one also needs to take into account that activated ECC
RAM costs you around 12%.

Hypnos has 64 K20 cards and 36 K80 cards (72 "GPUs").
Both queues have four cards per node, respectively (16 nodes in the
"k20" queue and 9 nodes in the "k80" queue).

   https://www.hzdr.de/db/Cms?pOid=29813&pNid=852

Credit: one node of the nine k80 nodes we are hosting with and for our
collaborators at HI Jena / GSI. In return, they can also use the other
k80 nodes.
I can imagine similar concepts for future queues, e.g. we are building a
new cluster that has P100 nodes and adding additional nodes from
collaborators for shared usage would be cool to allow larger simulations.


Best,
Axel

On 9/3/18 12:45 PM, Andrei Berceanu berceanu@runbox.com wrote:
> Thank you René.
> So are the 64 K80 boards on Hypnos 64 cards (with 2 GPUs each), or 64
> GPUs (32 actual cards)?
>
> Best,
> Andrei
>
> On Mon, Sep 3, 2018 at 11:31 AM René Widera r.widera@hzdr.de
> <mailto:r.widera@hzdr.de> <picongpu-users@hzdr.de
> <mailto:picongpu-users@hzdr.de>> wrote:
>
>     Hi Andrei,
>
>     I have currently no performance data for the P100 available but if I
>     remember correct one P100 is two times faster than one k80 (with on K80
>     I mean one card with two sockets).
>     For V100 I have currently no numbers in my head but due to improvements
>     of the shared memory atomics a V100 should be faster than a P100.
>
>     best regards,
>
>     René
>
>     On 08/31/2018 09:52 AM, Andrei Berceanu berceanu@runbox.com
>     <mailto:berceanu@runbox.com> wrote:
>     > Dear all,
>     >
>     > In you experience, how does PIConGPU performance vary across
>     different
>     > Tesla cards, such as the K80, P100 and V100?
>     >
>     > Is it fair to say, for instance, that one P100 is equivalent to
>     two K80s?
>     >
>     > Best regards,
>     > Andrei
>
>     --
>     René Widera
>     Phone: +49 351 260 3543
>     Institute of Radiation Physics
>     http://www.hzdr.de/crp
>     Helmholtz-Zentrum Dresden - Rossendorf (HZDR)
>     Bautzner Landstr. 400 | 01328 Dresden | Germany
>     Board of Directors: Prof. Dr. Dr. h. c. Roland Sauerbrey, Dr. Ulrich
>     Breuer
>     Company Registration Number VR 1693, Amtsgericht Dresden
>
>     #############################################################
>     This message is sent to you because you are subscribed to
>       the mailing list <picongpu-users@hzdr.de
>     <mailto:picongpu-users@hzdr.de>>.
>     To unsubscribe, E-mail to: <picongpu-users-off@hzdr.de
>     <mailto:picongpu-users-off@hzdr.de>>
>     To switch to the DIGEST mode, E-mail to
>     <picongpu-users-digest@hzdr.de <mailto:picongpu-users-digest@hzdr.de>>
>     To switch to the INDEX mode, E-mail to <picongpu-users-index@hzdr.de
>     <mailto:picongpu-users-index@hzdr.de>>
>     Send administrative queries to  <picongpu-users-request@hzdr.de
>     <mailto:picongpu-users-request@hzdr.de>>
>
>

--

Axel Huebl
Phone: +49 351 260 3582
Institute of Radiation Physics
http://www.hzdr.de/crp
Helmholtz-Zentrum Dresden - Rossendorf (HZDR)
Bautzner Landstr. 400 | 01328 Dresden | Germany
Board of Directors:
Prof. Dr. Dr. h. c. Roland Sauerbrey, Dr. Ulrich Breuer
Company Registration Number VR 1693, Amtsgericht Dresden