Hi Everyone, and more specifically developers of new code,

I am considering upgrading a few nodes whose Tesla K40 GPUs limit our capabilities. 

The chassis only allows for Tesla type cards and the motherboard is only supported for up to K80 cards. We’re going to test if P100s run on them but in case they do, 

what should I go with?

- K80s that I can find used for  ~$1000 but are limited to CUDA compute 3.7

            or

-P100s that cost ~$4000 but are a bit faster and have CUDA compute ability 6.

K80s are a lot cheaper, but are you developers soon going to take advantage of compute ability higher than 3.7 and make them immediately outdated?

Any other viewpoint is welcome. 

Best,

Amedee

---

To unsubscribe from the CCPEM list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1