Microsoft surface book question
4 Nov 2017 6:02:09 UTC
Topic 210698
(moderation:
Hello!
Currently acquired a surface book with a core i5-6200U and Nvidia GPU in the dock. I want to say it closely resembles a 940M from what I have read.
I also own a laptop with a Nvidia gtx-670M. How will both of these compare?
Right now I'm running WCG on the CPU. Should I select only run GPU on this project since, I'm assuming anyway, the GPU will be much faster and let WCG crunch away and use all available CPU, rather than trying to use 4 threads for 2 consecutive projects? Will the graphics card use one thread of the processor too?
Lastly, what sort of runtimes per WU should I expect? Are there different projects that will vary in their runtimes or are most things similar? (Or am I thinking of this wrong - on WCG there is SCC, MCM, HST etc...)
Thanks for any help! Looking forward to crunching!
Will likely set this up later today.
Short answer: definitely keep
)
Short answer: definitely keep the CPU at WCG. And generally I wouldn't crunch 24/7 with any latop if a failure would hurt me financially. However, if you want you can crunch with both of those GPUs at Einstein. They won't score records, but the WUs are short enough for it to make sense (in contrast to GPU-Grid).
MrS
Scanning for our furry friends since Jan 2002
It won't hurt me too bad - I
)
It won't hurt me too bad - I got this very cheap. The GPU tends to sit right around 61 to 65 c while running folding at home, at any rate. The CPU sits 80 c or so, which is warm enough but not enough to concern me.
It won't be crunching 24/7, though and when it is, it's in a basement room in Canada and it's below freezing outside already.
I don't have anything powerful enough for gpugrid from what I've seen. Really sucks they don't have anything shorter for the little guys.
Tyler_40 schrieb:Really sucks
)
Kind of. There's method to the madness: their WUs are generated based on the results of the previous one. They can have e.g. 1000 parameter variations running in parallel, but at some point they need to get results back. If that's not quick (e.g. 2 weeks as allowed for Einstein) they'd wait 1-2 years until they get the 1st scientific results - at which point they might notice that all other 999 parameter variations were useless. In fact, getting results back quickly is so important to them that they give a credit bonus for quick returns.
MrS
Scanning for our furry friends since Jan 2002