> I was wondering has anyone used AWS for Amber18 simulations and can
> guide me to some resources wherein I can see the compute costs and
> details about Amazon cloud services? At the lab we currently have 20
> Nvidia RTX 2080 TI machines, but because of an influx of new rotation
> students/postdocs we were thinking whether we could transition to AWS...
> Any ideas on expensive or not-expensive it will be ?
It will be prohibitively expensive unless you can get free cycles. For
parallel CPU runs, we estimate 5-11x more expensive than the total cost of
operations (including power, data center, staff) than our on-prem
clusters. 5x more if you prepay for AWS for 3 years. That is assuming
5-years of operation, but actually we run longer than 5 years on our
hardware, i.e. I still have K80's and C2070's running.
For GPUs, the costs would be even greater since the cloud providers do not
have the commodity cards and the prices are high for A100, V100, etc.
--tec3
\-/ Professor Thomas E. Cheatham, III
-/- Department of Medicinal Chemistry, College of Pharmacy
/-\ Director, Research Computing and CHPC, UIT, U of Utah
\-/
-/- tec3.utah.edu
http://www.chpc.utah.edu/~cheatham
/-\ SKH-4914 (801) 587-9652; FAX: (801) 585-6208
\-/ INSCC-410 (801) 585-6318; FAX: (801) 585-5366
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Aug 12 2020 - 16:00:02 PDT
