You are viewing a single comment's thread from:
RE: My quest to use an NVIDIA TITAN V for BOINC (TITAN V & 1080 Ti BOINC Benchmarks)
Man, such a waste, dont use it on windows, you need to use it on linux so you can run it on cli only mode, the perfomance difference is usually around 15%! Or at least dont use it as display, but in that case you could still gain a few % since these apps are built for linux and then ported.
I don't run my display off the TITAN, I use the 1080 Ti for that because I game on the 1080 Ti. I'm also pretty unfamiliar with Linux in command line only mode. I also don't know if I'd be able to overclock my GPUs on Linux CLI-only. So yeah, I'll be sticking with Windows even though I might be losing some performance.
O, you could just set it up in a graphic enviroment and then swap out to a tty pressing ctrl+alt+f(1-7)
Is there any data backing up the claim that Linux is faster for BOINC than Windows? I've heard it repeated, but I've never actually seen any data.
Yes.
There is the simple fact that linux has been much more optimized for scientific tasks than windows, its free, you can modify it, and it runs in 99-98% of all supercomputers. Of course there is going to be more work towards that.
then there is the fact that windows has much more tasks to do when idle (not really that important for gpu computing).
Last but not least, there are some applications such as TN-grid that just dont run natively on windows or run very slow and have to use mingw, which brings the linux family of oses way of interacting with programs to windows, and its slower than doing it natively on linux, obviously.
If you have any numbers to share of actual performance comparisons I would be interested.