Message boards :
Number crunching :
Linux vs. Windows credit?
Message board moderation
Author | Message |
---|---|
![]() Send message Joined: 16 May 15 Posts: 10 Credit: 15,994,451 RAC: 0 ![]() ![]() |
Recently started crunching NumberFields again on a few hosts and have noticed a credit oddity: On a 10+ year old Celeron900 Win 7 64bit laptop: workunit 17395431--- 13661.69sec CPU--- 300.88 credits (79/hr) 17395394--- 14357.49sec CPU--- 312.60 credits (78/hr) Core i3 2120 desktop Linux Mint 18.1 64bit (HT on): workunit 17370121--- 9421.11sec CPU--- 140.45 credits (54/hr) 17370149--- 9481.86sec CPU--- 141.44 credits (54/hr) Never before have I seen the old laptop outperform the i3 on a credit per hour basis. Had these (cr/hr) numbers been reversed, I would have thought nothing of it if I had even noticed. Is this credit difference between the 2 operating systems normal? ![]() ![]() |
![]() Send message Joined: 8 Jul 11 Posts: 1371 Credit: 645,292,942 RAC: 852,243 ![]() ![]() ![]() |
I think the root cause is a difference in benchmarks. Look at the "Measured floating point speed" for the 2 computers: https://numberfields.asu.edu/NumberFields/show_host_detail.php?hostid=23634 https://numberfields.asu.edu/NumberFields/show_host_detail.php?hostid=37707 For some reason BOINC thinks the old Celeron is faster with floating point operations than the newer core i3. The i3 has faster integer benchmarks, but I'm guessing those are not used when determining credits per hour. This may not be the whole story though, since the ratio of FLOPS (2133/2626) does not match your credit ratio (54/79). Maybe BOINC is using some additional information to determine credits per hour? |
![]() Send message Joined: 16 May 15 Posts: 10 Credit: 15,994,451 RAC: 0 ![]() ![]() |
Thanks for the reply, Eric. I think I'll manually request BOINC to run the CPU benchmarks again and see if anything changes and report back. |
![]() Send message Joined: 16 May 15 Posts: 10 Credit: 15,994,451 RAC: 0 ![]() ![]() |
Re-ran the benchmark and floating point value went up by over 50%! Last time I rebooted the i3 was for a kernel update and may explain the low value if benchmarks were running at the same time. Thanks for pointing that out. |
![]() Send message Joined: 8 Jul 11 Posts: 1371 Credit: 645,292,942 RAC: 852,243 ![]() ![]() ![]() |
Yeah, I've gotten really low benchmarks in the past when I was running something in the background that used all my cores. Best to do benchmarks when nothing else is running (if you care about the credit). |
![]() Send message Joined: 16 May 15 Posts: 10 Credit: 15,994,451 RAC: 0 ![]() ![]() |
I recall reading that close to 10 years ago when I first started BOINC. The problem is, BOINC likes to run the benchmarks right at reboot when it's been awhile. I'll have to remember to run them again when rebooting for Windows updates etc. Thanks again for the reminder! I like badges, and for those I need credit ;) ![]() ![]() |