[gmx-users] Workstation configuration advices?

[Magnus Lundborg]

Hi,

I'd recommend going for a GTX1080Ti GPU instead. Consumer cards work 
fine for GROMACS. The 1080Ti is cheaper and has higher performance than 
the Quadro P4000. It consumes more power, though.

Regards,

Magnus

On 2017-11-10 11:53, Javier E wrote:
> Dear users and developers,
>
>
> I am about to acquire a Workstation to mainly run GROMACS. In choosing a
> good configuration that fits the budget, I came out with this basic arrange
> (from Dell - Precision 7820 Tower):
>
> - Processor: Intel Xeon Gold 6130 2.1 GHz 3.7 GHz Turbo, 16Core 10.4 GT/s 2
> UPI, 22MB cache, HT (125W) DDR-2666. Or any other similar of the Intel
> Xenon (R) type.
>
> - Graphic card: NVIDIA Quadro P4000, 8GB, 4DP (7x20TB): even if I would
> like a Tesla P100, anyone has used Quadro cards or knows about its
> acceleration in GROMACS? Also, are Raedon graphic cards compatible/ tested
> for GROMACS (I didn't check if they support CUDA).
>
> - Memory 32 GB (2x16 GB or 4x8GB?) It wont be better 16GB of RAM and maybe
> 18Cores, 20Cores? The price rises very easy when adding only two-four
> cores! The same for the graphic card!
>
> - Primary Hard disk: Solid state (SSD) Class 50 of 512 GB. About the
> primary hard disk, should it be NVMe or SATA SSDs? I saw that NVMe has much
> better I/O communication but since its relative new tech, I don't know if
> GROMACS can fully run here, I believe it does. Also, NVMe it doesn't
> support RAID (is RAID a must for GROMACS?), and I'm afraid to encounter
> basic problems such as not being able to properly install the OS and the
> right dependencies/ libraries/ related tools-software.
> http://info.grcooling.com/nvme-vs-sata-ssd
> - Secondary disk: SATA 3.5'' 1TB 7200 rpm.
> - Optional: PCoIP Remote Workstation Card. I saw that it can enhance the
> lag. My personal laptop is quite old, is this useful?
> http://www.tomshardware.co.uk/answers/id-2792150/dell-tera-remote-access-host-card-works.html
>
> I'm intended to simulate polyelectrolyte hydrocarbon-polymers, its water
> uptake and ion-diffusion, in simulation boxes of a maximum about 10x10x10
> nm, but most of the work would be in much smaller dimensions.
>
>
> Any comments are highly appreciated.
> Best regards
>
> ________________________________________
>
> *Javier Luque Di Salvo*
>
> Dipartamento di Ingegneria Chimica
>
> Universitá Degli Studi di Palermo
> *Viale delle Scienze, Ed. 6*
> *90128 PALERMO (PA)*
> *+39.09123867503*