"Short run" item using ACEMD 9.05 (CUDA 8.0)

Message boards : Number crunching : "Short run" item using ACEMD 9.05 (CUDA 8.0)

Author	Message
Dave Peachey Send message Joined: 16 May 09 Posts: 11 Credit: 131,226,034 RAC: 0 Level Scientific publications	Message 44870 - Posted: 27 Oct 2016 \| 16:00:43 UTC Last modified: 27 Oct 2016 \| 16:22:26 UTC
	All, Notwithstanding this "short run" item is slated to run for 15h:30m (hardly "short", but that could be because it has no precedent), has anyone else encountered one of these yet and, if so, do we know whether these will work on the NVidia GTX10xx series cards? I've found out, when I read the relevant thread in the "Graphics Cards" section, that a new version of the application is being worked on (using CUDA 8) and wondered whether I am the lucky recipient of an early example (which, if I am, I'll try out on my GTX1060 rather than on my GTX750ti). <edit>Discovered the "PASCAL App Testing" post in the "News" section so that answers my question; now let's see if it works!</edit> Cheers Dave
	ID: 44870 \| Rating: 0 \| rate: / Reply Quote

Beyond Send message Joined: 23 Nov 08 Posts: 1112 Credit: 6,162,416,256 RAC: 0 Level Scientific publications	Message 44976 - Posted: 28 Oct 2016 \| 17:19:19 UTC - in response to Message 44870.
	I've found out, when I read the relevant thread in the "Graphics Cards" section, that a new version of the application is being worked on (using CUDA 8) and wondered whether I am the lucky recipient of an early example (which, if I am, I'll try out on my GTX1060 rather than on my GTX750ti). They're working on your 1060 but so far the run times are really ugly. Hopefully they'll improve.
	ID: 44976 \| Rating: 0 \| rate: / Reply Quote

skgiven Volunteer moderator Volunteer tester Send message Joined: 23 Apr 09 Posts: 3968 Credit: 1,995,359,260 RAC: 0 Level Scientific publications	Message 44983 - Posted: 28 Oct 2016 \| 18:51:21 UTC - in response to Message 44976.
	It's far too early to judge the GTX1000 range. The Short Run times are all over the place on my GTX970's (even for all 5ns tasks) never mind the Pascals (which are in Beta testing) - The present short WU's are variable, so best not to draw conclusions based on their return times. The tasks also replicate small size molecular interactions (so they don't scale). Anyway the Pascal's performance/Watt was the headline grabber rather than their out and out performance, which is expected to be similar to previous generations (just roughly equating to one GPU down; GTX980Ti≈GTX1080, GTX1070≈GTX980...). ____________ FAQ's HOW TO: - Opt out of Beta Tests - Ask for Help
	ID: 44983 \| Rating: 0 \| rate: / Reply Quote

Message boards : Number crunching : "Short run" item using ACEMD 9.05 (CUDA 8.0)

//