Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
![]() |
World Community Grid Forums
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 18
|
![]() |
Author |
|
Hypernova
Master Cruncher Audaces Fortuna Juvat ! Vaud - Switzerland Joined: Dec 16, 2008 Post Count: 1908 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
In the old times (seven years ago) Windows based platforms were less efficient than Linux based platforms with the same Intel CPU.
----------------------------------------This is something that has never been very clear to me why, but the number of results or number of calculations performed were much higher with Linux OS than with Windows OS. CPU being equal I can admit little differences one or two percent but here the differences were in the order of 20% or more. Calculations are made by the hardware and not the OS so this is still a mystery to me. I am thinking about building a new dedicated rig this time probably with AMD based CPU as for once AMD seems to have surpassed Intel in terms of performance. My question considering AMD hardware is: has still Windows 10 Pro or Enterprise much less performance than a Linux based solution or differences are now marginal? ![]() |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The last of the 'Autodock-Vina' sciences is about to finish (VINA based Smash Childhood Cancer may come back). With that the differences of the remaining sciences are nominal, observational data is absent for the new AFrican Rainfall Project as yet.
----------------------------------------As for the big difference, it's the compiler libs that determined the 'discrepancy', not the hardware. The Linux Integer parts are much better optimized. You can see it just by running the BOINC Benchmark. Dhrystone shows Double/Triple the result on Linux/Mac. [Edit 2 times, last edit by Former Member at Oct 31, 2019 9:54:47 AM] |
||
|
hchc
Veteran Cruncher USA Joined: Aug 15, 2006 Post Count: 806 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
For ARP1 (WRF application), my Ivy Bridge CPU (3rd Gen Core) on Linux is significantly faster than my Coffee Lake CPU (8th Gen Core) on Win10 Pro. I'd install Linux on your new dedicated cruncher, Hypernova, not just for the overall performance improvement but for saving the money on a Windows license as well as the reduced attack surface security-wise and much less system management (e.g. updates) and bloat.
----------------------------------------Regarding your question that "Calculations are made by the hardware and not the OS so this is still a mystery to me," the OS still handles memory management, and Linux apparently does it much better.
[Edit 2 times, last edit by hchc at Oct 31, 2019 12:11:19 PM] |
||
|
Hypernova
Master Cruncher Audaces Fortuna Juvat ! Vaud - Switzerland Joined: Dec 16, 2008 Post Count: 1908 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Thanks for the replies
----------------------------------------![]() |
||
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1673 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
The main problem is not only a compiler optimisation on Linux side but, worst, a very inefficient memory management on Windows side (since decades).
----------------------------------------Usually Windows uses much more memory for the same project than Linux (often the difference is at least factor 2). If more memory is used, more transfers must occur and less efficient is the overall system. It is a real pity if you consider how many Windows machines are world wide installed and in use, and that this memory issue impact every applications running on the machines, not only WCG resp. Boinc. Cheers, Yves |
||
|
fuzzydice555
Advanced Cruncher Joined: Mar 25, 2015 Post Count: 89 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
I tested the same Xeon 2628v4 machine with Ubuntu Server vs Windows 10. This is points/hour for one thread.
----------------------------------------Project - Points linux - Points windows SCC - 29,4 - 16,46 MCM - 17,22 - 16,69 FAH2 - 16,29 - 18,24 ZIKA - 31,16 - 16,79 MIP - 23,86 - 16,98 On FAH2 & MCM they're almost equal, all other projects favor linux heavily. I can't qoute you exact numbers, but my AMD Ryzen rig gets much more points on linux as well. Not to mention no windows update and no blue screen, you start it and 3 years later it's still going strong ![]() ![]() [Edit 1 times, last edit by fuzzydice555 at Nov 2, 2019 11:53:57 PM] |
||
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7687 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
I tested the same Xeon 2628v4 machine with Ubuntu Server vs Windows 10. This is points/hour for one thread. Project - Points linux - Points windows SCC - 29,4 - 16,46 MCM - 17,22 - 16,69 FAH2 - 16,29 - 18,24 ZIKA - 31,16 - 16,79 MIP - 23,86 - 16,98 On FAH2 & MCM they're almost equal, all other projects favor linux heavily. I can't qoute you exact numbers, but my AMD Ryzen rig gets much more points on linux as well. Not to mention no windows update and no blue screen, you start it and 3 years later it's still going strong ![]() It has been a well established fact that any of the VINA related jobs run much quicker on Linux than Windows (and,hence, more points.) Evidently MIP shares this attribute also. Thanks for the info. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I tested the same Xeon 2628v4 machine with Ubuntu Server vs Windows 10. This is points/hour for one thread. Project - Points linux - Points windows SCC - 29,4 - 16,46 MCM - 17,22 - 16,69 FAH2 - 16,29 - 18,24 ZIKA - 31,16 - 16,79 MIP - 23,86 - 16,98 On FAH2 & MCM they're almost equal, all other projects favor linux heavily. I can't qoute you exact numbers, but my AMD Ryzen rig gets much more points on linux as well. Not to mention no windows update and no blue screen, you start it and 3 years later it's still going strong ![]() The best part is spliced in kernel updates on a hot machine... no more stopping, booting, most of the times. |
||
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Good to see you around again Hypernova. JMHO but I would avoid Win10 like the plague and here is one of many reasons why.
----------------------------------------https://betanews.com/2016/02/06/windows-10-ph...g-and-telemetry-disabled/
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.
----------------------------------------![]() ![]() [Edit 1 times, last edit by nanoprobe at Nov 5, 2019 12:59:10 PM] |
||
|
DrMason
Senior Cruncher Joined: Mar 16, 2007 Post Count: 153 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
If you are going with AMD (and I agree with you about workstation core, thread and value superiority at the moment), Linux generally should put out more work. The Windows scheduler traditionally has had problems assigning work to many cores on AMD chips, and continues to struggle with NUMA memory controllers, which is currently a key part of the Ryzen/Threadripper/Epyc designs.
----------------------------------------![]() |
||
|
|
![]() |