[OpenSPIM] computer configuration

[Thomas Pujol]

Thanks guys,

so I already have a pipeline working on the cluster but our problem is that
the priority rules for the jobs have been establish many years ago without
taking into account the needs of image analyst (lot of ram for a short
time) which lead to unexpected and unpleasant delays in the processing
(wait 2 weeks for a 2h job is a bit excessive). Two solutions make a
smarter pipeline or buy a big computer, I would rather go for option number
two.

I the end I will push for a workstation more than a private node on the
cluster because you can do the maintenance yourself  and even thought we
don't use GPU calculation at the moment it would be nice to keep this door
open.

I will ask for a quote to HP for the configuration below if someone is
interested.

Small question, linux kind of have a repulsive effect on brave biologist
exploring biophysics, is there any downside to use windows 8.1 instead of
linux ?

thanks

thomas


*OS *: Windows 8.1 Pro 64 or Ubuntu Linux 64

*Processor :* Intel® Xeon® E5-2680 v3 (2.5 GHz, 30 MB cache, 12 cores,
Intel® vPro™)

*Graphic card :* NVIDIA® Quadro® K6000 (12 GB) High Performance GPU
Computing: NVIDIA® Tesla® K40

*RAM :* 128 GB DDR4-2133 ECC

*Main storage :* 1 TB SATA SSHD11 (may be less)

*Storage :* mucho To at 7200 rpm or more



2016-01-27 17:28 GMT+01:00 Ulrik Günther <guenther at mpi-cbg.de>:

> Hey Thomas,
>
> in addition to Michael’s remarks, I’d like to add that the graphics cards
> in your listing are quite dated, but anyway still expensive — I found them
> for around 1000EUR online.
>
> If you plan on running things like Multiview Deconvolution (
> http://fiji.sc/Multi-View_Deconvolution) on the GPU, you’d probably be
> better off getting a more recent card, like a GeForce Titan X with 12 GB
> RAM, instead. That card also costs around 1000EUR.
>
> The main difference (apart from the Titan X having a much more recent
> architecture) between the two is the double precision calculation speed:
> The consumer (read non-Quadro) cards are artificially capped at 1/8th of
> the single precision processing speed. For most image processing tasks
> however, double precision is not necessary, but more RAM on the GPU helps
> tremendously.
>
> Hope that helped :)
>
> cheers!
>
> ulrik
>
> —
> Ulrik Günther, Dipl.-Phys.
> PhD Student
> MOSAIC Group & Tomancak Lab, Center of Systems Biology Dresden
> Max Planck Institute of Molecular Cell Biology and Genetics
> Pfotenhauerstr. 108, D-01307 Dresden, Germany
>
> E-Mail: guenther at mpi-cbg.de
>
> On Jan 26, 2016, at 6:04 PM, Thomas Pujol <thomas.pujol at crg.eu> wrote:
>
> Hello dear light sheet enthousiaste.
>
> An email to fish advices regarding computer configuration.
>
> I convince my boss that we need a super duper computer to process the spim
> data that we are producing and I want to know if the configuration advice
> by the open spim website is up to date(description below) or if I
> should/could aim for even bigger machine. Online, I also found the HP Z840
> that can accomodate a tremendous amount of ram, any recommandation?
>
> We have to choose between buying a work station with massive a amount ram
> or our private node on the institute cluster, does anyone has advice or
> feedback on previous expererience ?
>
> thank you and may the light sheet be with you
>
> cheers
>
> thomas pujol
>
> Processor : *Two Intel Xeon Processor E5-2630* (*Six Core, 2.30GHz Turbo*, 15MB, 7.2 GT/s)
> Memory : *128GB* (16x8GB) 1600MHz DDR3 ECC RDIMM
> Hard Drive : *4x2TB* 3.5inch Serial ATA (7.200 Rpm) Hard Drive
> HDD Controller : PERC H310 SATA/SAS Controller for Dell Precision
> HDD Configuration : C1 SATA 3.5inch, 1-4 Hard Drives
> Graphics : *Dual 2 GB NVIDIA Quadro 4000* (2cards w/ 2DP & 1DVI-I each) (2DP-DVI & 2DVI-VGA adapter) (MRGA17H)
> Network : Intel X520-T2 Dual Port *10GbE Network Interface Card*
>
> _______________________________________________
> OpenSPIM mailing list
> OpenSPIM at openspim.org
> http://openspim.org/mailman/listinfo/openspim
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openspim.org/pipermail/openspim/attachments/20160128/e8eaf387/attachment-0002.html>