Contributions are appreciated
Analysis of 367 verified O2MD GPU work units, v2.02/v2.03
GPU: Nvidia GTX 1050, single task
CPU: i5-4460, one extra core reserved
OS: Linux, Debian testing
Average runtime: 1046 s
Min. runtime: 231 s
Max. runtime: 1364 s
Standard deviation: 153.6 s
Copyright © 2024 Einstein@Home. All rights reserved.
Such comparisons don't/won't
)
Such comparisons don't/won't make much sense, since run-times are very dependant on many "other factors".
So, I take it, this is just for plain fun or out of "curiosity".
Here you are (just 84 WUs on 2.02) -- nothing special or fancy:
Run-time: 52 WUs took 545 - 659 s on Win10 CPU Xeon 6144 using Titan X(pascal) GPU.
Average 604 secs.
Run-time: 32 WUs took 705 - 791 s on Win7 CPU Xeon e5-2643 using Titan V GPU.
Average 719 secs.
GPU WUs seem to generally run faster on Win10 than on Win7 -- I have no idea why.
CPUs and GPUs are not OV.
CPUs HT off.
I will not declare the above values as representative of or for anything.
Run times today... averages
)
Run times today... averages of "remaining (estimate)" for the tasks in queue. Based on a quick manual observation, times reflect pretty well the reality at the moment:
* These run times show "how much it is for 1 task", not completion times for the cards running 2x
AMD R9 390, running 1x ... 820 s / task ... (2 hosts, WIndows 7 & 8,1, can't run tasks in parallel or the second task would enter "sloth mode" and the first task would slow down too, propably a GPU driver thing)
AMD RX 580, running 2x ... 680 s / task ... (2 hosts, WIndows 10)
AMD RX 570, running 2x ... 969 s / task ... (WIndows 10) - seems to be running some hefty tasks right now, has been quicker occasionally
Nvidia GTX 960, running 2x ... 1175 s / task ... (2 hosts, WIndows 10)
Nvidia GTX 960, running 2x ... 1320 s / task ... (WIndows 8.1)
** All hosts above run with Intel Xeon X56xx @ 3.8 GHz
Nvidia GTX 1060, running 1x ... 870 s / task ... (Windows 7, Intel Q9550 @ 3.6 GHz)
Thanks for the replies. Not a
)
Thanks for the replies. Not a big spread in runtimes across the different cards, it seems. Titan X is not even twice as fast as my GTX 1050?