Message boards :
Number crunching :
Best OS for seti?
Message board moderation
Author | Message |
---|---|
sweede Send message Joined: 5 Mar 01 Posts: 1 Credit: 358,015 RAC: 0 |
Hi! Which OS is the best for calculating WUs as fast and good as posible? According to statistics on Boincstats OS page it seams to be Microsoft Windows Server 2003 "R2". Anyone that got a different awnser to my question or just can confirm that I have read the list as it should be read? //sweede |
ML1 Send message Joined: 25 Nov 01 Posts: 20147 Credit: 7,508,002 RAC: 20 |
Which OS is the best for calculating WUs as fast and good as posible? A quick look for the top computers for s@h shows "Darwin 8.10.1" (Mac) to be favourite. The Windows may well be fine for performance if you are NOT running any antivirus. My strong preference is for a Linux or BSD based system. If you are really looking for maximum performance, then you should be looking at a minimal Linux system, or even a minimal Gentoo system. But then again, any good OS takes so little percentage of the CPU that the OS overhead should be negligible. The latest (tickless) Linux kernels should offer the ultimate in minimising OS overhead. And then again, what is most important is whatever is most reliable for you. Happy crunchin', Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
ML1 Send message Joined: 25 Nov 01 Posts: 20147 Credit: 7,508,002 RAC: 20 |
Wow! I never realised that Linux was so far ahead of Vista!! Aside: Note that those stats will be heavily skewed by what hardware the various OSes are running on. Older OSes may well be on older (slower) hardware. Server OSes will likely be on high performance hardware. And Linux...? That can be on anything from your washing machine and mobile phone all the way up to the world's fastest supercomputers. Happy crunchin', Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
Sutaru Tsureku Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 |
To my knowledge.. Windows 32 Bit is faster than 32 Bit Linux and.. Windows 64 Bit is faster than 64 Bit Linux and.. Linux 64 Bit is faster than 32 bit Windows.. ..all opt. apps.. |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
LOL, from those Boincstats OS numbers it looks like I'm not the only one sticking with 32 bit Win XP for a while. :D "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
TeamDGC Send message Joined: 27 Oct 99 Posts: 19 Credit: 7,091,042 RAC: 0 |
I suspect this would be quite fast! http://blog.schreiter.info/boincpe-livecd-for-boinc/ W2003 Server is most likley the fastes couse it's mostly run by fast Xeon CPUs! All time #1 M.U.R.C. Cruncher! |
zombie67 [MM] Send message Joined: 22 Apr 04 Posts: 758 Credit: 27,771,894 RAC: 0 |
|
ML1 Send message Joined: 25 Nov 01 Posts: 20147 Credit: 7,508,002 RAC: 20 |
To my knowledge.. Is that for the s@h science applications? As for the actual OSes, I don't know that there are any reliable numbers for either OS being faster than the other, anyway around. Happy crunchin', Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
Sutaru Tsureku Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 |
There are a lot of other threads that handles this issue more seriously. Please could you tell us/me this threads? |
Sutaru Tsureku Send message Joined: 6 Apr 07 Posts: 7105 Credit: 147,663,825 RAC: 5 |
To my knowledge.. This infos are from the optimizer-COOP.. OOppss.. I forgot.. I thought only to the 'normal' PCs.. ;-) I think Alex's opt. apps for MACs are the fastest apps around.. |
Gecko Send message Joined: 17 Nov 99 Posts: 454 Credit: 6,946,910 RAC: 47 |
There has never been a definitive conclusion to this question that has been asked several times and discussed in several threads the past few years. The reason simply is that conditions don't easily allow for an apples-apples test. Per Zombie's comment: Yes, OSX + Alex's V8 is the fastest current combo, as V8's predecessors have been for the past couple of years. Of course as most know, this is primarily because of Alex's ap code, not OSX, and not always dependent on the CPU arch. It wasn't long ago when PPC G5s at 2.5GHZ running Alex's aps would still handily out-perform the fastest Netburst-based chips....all this w/o Intel compilers, large L2s, or 3GHZ+ clock freq. GCC, FFTW, & VMX (Altivec) optimizing performed quite admirably until the new sheriff, Core2 came on scene. Reminiscing aside, a logical way to truly answer the question would be to compare an identical application build that was ported to identical hardware running in same 32bit or 64bit mode on common public OS distros as XP or Vista, OSX, and comparable Linux i.e. RedHat, Suse, Mandriva, Ubuntu, Fedora etc...pick your flavor. This could provide an answer from a "level" playing field perspective, though one might be able to un-level the field w/ Linux to an extent and completely build their own "optimized OS" from the ground-up to compliment a specific CPU, custom BOINC client and S@H application optimized to the CPU. http://www.linuxfromscratch.org/ It's possible that one might claim they had the "best" possible combo this way, IF...Linux, OSX and Windows "X" are comparatively close in basic performance to start. It appears like ALOT of work to reach a credible answer. |
Tklop Send message Joined: 11 May 03 Posts: 175 Credit: 613,952 RAC: 0 |
Though this is probably not much use to anyone here (with the majority of posters using modern fast crunching equipment) I can testify to a definite improvement in RAC when I switched some older machines (500MHz P3, 450MHz Celeron, etc.) from Windows XP Pro to Linux Debian. Maybe if I'd been running Windows 98 on those old boxes, it might not have made much difference, I don't know one way or another. Still, it seems to me that those old and slow boxes really weren't up to the overhead that XP Pro required... But I'm no expert at these things--this is only my rather non-technical observation... Keep on crunching, all... SETI@Home Forever! ___Tklop (Step-Founder, U.S. Air Force team) |
ML1 Send message Joined: 25 Nov 01 Posts: 20147 Credit: 7,508,002 RAC: 20 |
... But then again, any good OS takes so little percentage of the CPU that the OS overhead should be negligible. An interesting (non-technical) spin on the "tickless kernel" innovation in Linux is described in: Linux coders tackle power efficiency Maybe you'd be better off if you didn't spend so much time looking at your watch. That, loosely speaking, is the rationale behind a significant change at the heart of Linux that programmers hope will make the open-source operating system more efficient. New versions of the operating system are being endowed with a "tickless" kernel that forsakes traditional computer time-keeping in an effort to keep the processor in a somnolent, low-power state. The other aspect of that is that there is less overhead by not maintaining the kernel clock ticking every thousandth of a second. Rather nice! But then again... As mentioned, the OS overhead should be a long way below 1% of the CPU overheads. Other applications, especially those that do a lot of disk accesses or 'busy waits', are much more of a problem. Happy crunchin', Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
DJStarfox Send message Joined: 23 May 01 Posts: 1066 Credit: 1,226,053 RAC: 2 |
An interesting (non-technical) spin on the "tickless kernel" innovation in Linux is described in: For some reason, I can't use tickless on dual Opterons x86_64, so I just recomplied for 300Hz timer and no kernel preemption. So far, no noticable performance degredation. Maybe that will help squeeze a few more FLOPS out of SETI. |
ohiomike Send message Joined: 14 Mar 04 Posts: 357 Credit: 650,069 RAC: 0 |
I've had best luck compiling with a 100 Hz timer and Low-Latency (preemptive). Note: Arch Linux or PCLinuxOS run better than most (with Red Hat Fedora I need to turn off 20 or 30 services I don't want (( sendmail, cups, etc- it is a bad as Windoze)) or need). Boinc Button Abuser In Training >My Shrubbers< |
Toby Send message Joined: 26 Oct 00 Posts: 1005 Credit: 6,366,949 RAC: 0 |
Yeah I was hoping to play with the tickless kernel on my laptop when I upgraded to Ubuntu Gutsy Gibbon but the tickless kernel isn't stable on x86_64 (aka amd64) yet. Guess I'll have to wait for Hardy Heron. Or live on the edge and compile my own I guess :) As for "which OS is better" here is another URL to consider: Comparison of my two 4400+s. These two computers have the same CPU and motherboard. But it really still isn't a good comparison... As you can see by the pie charts, they run different projects and different projects tend to give different credits/hour so... yeah it is virtually impossible to do a good comparison. Plus, as has been mentioned, the compiler and libraries that are used when building the project application have a much bigger effect on the processing speed than the OS itself. A member of The Knights Who Say NI! For rankings, history graphs and more, check out: My BOINC stats site |
ML1 Send message Joined: 25 Nov 01 Posts: 20147 Credit: 7,508,002 RAC: 20 |
Yeah I was hoping to play with the tickless kernel on my laptop when I upgraded to Ubuntu Gutsy Gibbon but the tickless kernel isn't stable on x86_64 (aka amd64) yet. Guess I'll have to wait for Hardy Heron. Or live on the edge and compile my own I guess :) Thanks for that comment. I'm likely to do an upgrade to try that out later this weekend... Looks like I'll just have to try it out but also have the previous (ticking) kernel to fall back on. Running on the bleeding edge can be fun. Then again, keeping the latest 'stable' kernel to hand can be very boring but also very useful for the reliability. ... Plus, as has been mentioned, the compiler and libraries that are used when building the project application have a much bigger effect on the processing speed than the OS itself. Agreed, hence the choice is more of what is most useful/reliable for a particular user. Here's still hoping for Who? What? Where? Skulltrail? Vendetta? Whatever? coming up with some radical goodies despite the longest Viral Marketing Hype I've known anyone to perpetrate. Hopefully, it might overcome the 'gazillion CPU hours' of performance I've needlessly lost on my systems due to the "Naughty Intel" silliness. Full marks and all good fun but don't rely on any of it! Happy crunchin', Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) |
peristalsis Send message Joined: 23 Jul 99 Posts: 154 Credit: 28,610,163 RAC: 51 |
As for "which OS is better" here is another URL to consider: Comparison of my two 4400+s. These two computers have the same CPU and motherboard. But it really still isn't a good comparison... As you can see by the pie charts, they run different projects and different projects tend to give different credits/hour so... yeah it is virtually impossible to do a good comparison. I'm also running two 4400's. WinXP machine rac is 954. Linux (PcLinuxOS) rac is 824. Win machine is my general "do everything" (surfing/burning/email/usenet/etc.) machine and runs a p2p program constantly. Its just about 100% seti. The linux box just runs seti and occasional einstein when seti goes down. Only hardware difference between the two machines is that the win box has two gigs of memory and the linux box one gig. Noticed the same disparity between your two machines ref the memory difference. Maybe that extra gig of memory is making that (not inconsiderable) difference in rac?..john |
Pilot Send message Joined: 18 May 99 Posts: 534 Credit: 5,475,482 RAC: 0 |
To my knowledge.. The numbers below do not filter for speed of hardware so it not an absolute proof of OS efficiency. It is probably reasonably good indication though IMHO for the top 5 or 10. Pos. Name of Operating System Average credit per OS 1 Microsoft Windows Server 2003 "R2" 188.73 2 Microsoft Windows Server "Longhorn" 181.58 3 Microsoft Windows XP Professional x64 Edition 126.55 4 Microsoft Windows Server 2003 125.08 5 Microsoft Windows Vista 86.30 6 Darwin 37.21 7 FreeBSD 26.14 8 Linux 21.25 9 IRIX64 19.46 10 HP-UX 18.98 11 Microsoft Windows XP 18.35 12 OpenBSD 17.93 13 Microsoft Windows Longhorn 16.86 14 OS/2 15.24 15 Microsoft Windows 2003 14.25 16 SunOS 12.67 17 NetBSD 10.13 18 Microsoft Windows 2000 10.01 19 Microsoft Windows NT 4.37 20 Microsoft Windows 98 2.48 21 Microsoft Windows Millennium 1.97 22 Microsoft Windows 95 0.63 23 IRIX 0.44 24 0.05 25 AIX 0.04 When we finally figure it all out, all the rules will change and we can start all over again. |
Fred W Send message Joined: 13 Jun 99 Posts: 2524 Credit: 11,954,210 RAC: 0 |
The numbers below do not filter for speed of hardware so it not an absolute proof of OS efficiency. It is probably reasonably good indication though IMHO for the top 5 or 10. Have you really thought about these numbers? I don't believe that Server 2003 is really 10 times faster at crunching than XP!! Is it not far more likely that the Server OS's are running on multi-core hosts which invalidates all of the above? I am sure that optimised / stock App has far more effect on crunch time than OS too. F. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.