| 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: 10
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I have a NUC (Next Unit of Computing) modell NUC5PPYH with Ubuntu server 14.04.
Was pleaced how it crunch and only drain 12 watts mecured on wall socket at 100% load on all four cores. So i bought a nother one and installed same brand and size of RAM and SSD and for fun and compairation i put Windows 10 on it and downloaded client from WCGs homepage. Raned it for a week and compaired results with it's twin with Ubuntu. What a fail Windows client is! It take wooping one and a half (1 1/2) hour longer to complete mapping cancer works of same batch compaired to Linux client. What a waste of time and energy! Even worse it becomes then you know what over 95% of computers runs windows. Huge waste of resources in the Grid! Why performe Windows client so poorley? My Windows 10 NUC has now got Ubuntu Server too and crunch same speed as it's twin. |
||
|
|
Falconet
Master Cruncher Portugal Joined: Mar 9, 2009 Post Count: 3315 Status: Offline Project Badges:
|
I can't comment on MCM but yes, Linux is faster for some projects.
----------------------------------------Try running OET1 on Linux and then on Windows 10. OET (and every other VINA project, such as the upcoming MDDS project currently in BETA) is some 50% or so faster on Linux 64-bit. I believe FAAH2 is also slightly faster on Linux. And probably HST as well ( see: https://secure.worldcommunitygrid.org/forums/wcg/viewthread_thread,39028 ) ![]() - AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W - AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W - AMD Ryzen 7 7730U 8C/16T 3.0 GHz |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Your post Falconet put more water under the bridge for my statement that WCG for Windows sucks and the grid dont run efficent as it could due to it must have most Windows machine.
My NUCs just run OET, UGM and MCM projects and they mostly get MCM works. Even if MCM stood out as slowest project under my Windows test week Ubuntu was ahead or close even, but never slower in those three projects. I dont do all project on the NUCs because 33+ hours works is lackluster. My i3 and i5 does all projects but both of them is Windows 10 machines so have nothing to compair with in the same way as identical hardware as with the NUCs. Whats why I did the test week with them (beside that the NUC is barebone and I had too give the new one an OS anyway). |
||
|
|
Falconet
Master Cruncher Portugal Joined: Mar 9, 2009 Post Count: 3315 Status: Offline Project Badges:
|
All I know is that it has to do with the Linux libraries/compilers, something like that.
----------------------------------------Use Linux 64-bit wherever you can ;) According to BOINCstats ( http://boincstats.com/en/stats/15/project/detail/host ): Close to 13% of the WCG's hosts run Linux. 78% run on Windows. Only 8% of total hosts are active. ![]() - AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W - AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W - AMD Ryzen 7 7730U 8C/16T 3.0 GHz [Edit 1 times, last edit by Falconet at May 8, 2016 10:43:24 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
"Only 8% of total hosts are active"
What is even bigger waste then run slow on Windows. But I guess what previous installation on devices not longer in use counts in too. Like I have 10 installations but only 4 is in use now. Rest of the 6 is phones what had died and been replaced or computers what been replaced too. But i can see the problem with non activity too. For exampel my team has nine members only 2 are active at precent. 4 of them have never done any work at all just signed up. |
||
|
|
Falconet
Master Cruncher Portugal Joined: Mar 9, 2009 Post Count: 3315 Status: Offline Project Badges:
|
"Only 8% of total hosts are active" What is even bigger waste then run slow on Windows. But I guess what previous installation on devices not longer in use counts in too. Like I have 10 installations but only 4 is in use now. Rest of the 6 is phones what had died and been replaced or computers what been replaced too. But i can see the problem with non activity too. For exampel my team has nine members only 2 are active at precent. 4 of them have never done any work at all just signed up. Exactly. Many installations of the same host over time. Still, it is a very low number. And many, like you said, sign up but don't crunch at all. ![]() - AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W - AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W - AMD Ryzen 7 7730U 8C/16T 3.0 GHz |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges:
|
WCG slower on Windows than on Linux. The reason for this huge difference should be found within the operating system architecture. Since decades, Windows design is poorly efficient comparing to *nix. Working in the industry since close to 30 years, I always fight - as far as possible - for relying on Unix/Linux systems for time critical and performance critical applications instead of Windows systems. You can easily notice that the memory management is far away to be efficient by Windows systems. Regardless of the OS memory foot print (which is anyway higher by Windows XP/7/... than by Linux), you can observe how much memory is required per WU (respectively per boinc thread) by Windows and by Linux. For OET1, 8 threads including OS take < 0.9 GB RAM on Linux Mint 17.3 x64, but require about 4 GB RAM on Windows 7 Pro SP1 x64. The reason for Linux success by Internet infrastructure (service providers, hosters, etc.), should be found at first by resources effectiveness, secondly by system management efficiency and finally by limited license costs. Over the years, I migrated all my WCG hosts to Linux (Ubuntu and Linux Mint), I have to keep only one Windows system alive for other reasons. I am currently following the same approach for office environment moving everything to Linux and keeping Windows available using VM for supporting specific application. Happy crunching on ... Linux, Yves |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
The all-knowing and the dont knowing might want to read back in the forums [such as posts by the late lawrencehardin]... the differential is kind of purely the integers [the BOINC benchmark Dhrystone has been expressing this since whenever BOINC came out]. There's hardly difference on the floats front, the Whetstone test.
----------------------------------------[Edit 1 times, last edit by SekeRob* at May 8, 2016 5:21:16 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Oh so it is nothing to do too make it better i guess then. just stay away fron Windows if care for performance.
About memory use I have also notice that Windows is bad to release unused memory so it takes up more and more over time. Example i ran with 4 days uptime and started with 20% use after 4 days it was up to 25% without I started other programs and with same processes as at start on. Ubuntu dont do that. |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges:
|
I don't notice release problems of unused memory on Windows.
----------------------------------------My Windows host is running for months without any reboot. The memory trouble you observed is very probably not caused by boinc WUs. In all cases maybe, it is related to memory allocated for cache purpose. Cheers, Yves |
||
|
|
|