Benchmarking Linux vs Windows

Having problems installing that new stick of memory? Found some great software or having issues with something? Or maybe want to chat about your PlayStation, X-Box, Nintendo, Sega, even your old Spectrum 48k....! Or maybe something you want to sell or acquire (computing related of course!). Let us know here...
Post Reply
chriscambridge
Active UBT Contributor 1+ yr
Posts: 2178
Joined: Mon Aug 08, 2016 1:56 pm
Location: UK

Benchmarking Linux vs Windows

Post by chriscambridge »

So I was told by a few BOINC people that considerable more output could be achieved on Linux over Windows (upto 50%), so I decided to put it to the test on a pair of Thread Rippers (TR) running on WCG. Interestly this is the same thing that was mentioned by many people also for crypto mining.

Please note the following:

1) The test has only been running for a couple of days.
2) The Window setup is inside a Server rack, where as the Linux setup is 'out of the box'; I intend to move both inside a rack very shortly.
3) Both setups have exactly the same hardware, and have been running the same sub-project on WCG, the only difference is the OS; given point #2.
4) The Windows setup has been running about a week or two longer than the Linux setup, so technically should achieve a larger credit output as it will have less pending validations (as new hosts have less trust than older ones).

* The Linux distro is Linux Mint, which has been created for Windows users; Currently it is installed, booting, and running from a 7GB thumb drive! The Windows install (for this Linux host) is still sitting on the HD not being used.

https://linuxmint.com/
https://www.zdnet.com/article/how-to-in ... indows-pc/


The figure is daily credit.

AMD Ryzen Threadripper 1950X 16-Core Processor
32
Windows 10 Professional x64 Edition, (10.00.17134.00)
25,359

AMD Ryzen Threadripper 1950X 16-Core Processor
32
Linux
45,856

https://boincstats.com/en/stats/15/host ... /0/1016004

So as you can see that is nearly 50% more credit just for running Linux over Windows.
Woodles
UBT Contributor
Posts: 11757
Joined: Thu Dec 20, 2007 12:00 am
Location: Cambridgeshire

Re: Benchmarking Linux vs Windows

Post by Woodles »

Hi Chris,

Two things:

Any improvement depends on the project involved. Some projects perform better on Linux, some better on Windows by varying degrees. Linux is usually better as it has fewer overheads than Windows.

Windows 7 is a lot less bloated and more efficient than Windows 10.

Unfortunately, the only host I have ran Windows 10 on recently was running Collatz which is almost entirely GPU biased but:

Code: Select all

OS		Total time	CPU time
Windows 7	370.97		0.68
Windows 10	375.94		1.27
Same host, dual boot setup. Average of the last twenty tasks.

So even with a project that is heavily GPU biased, Windows 10 takes roughly twice as long to do it's work than Window 7 and slightly longer overall.

My Linux hosts have only been running Wanless which doesn't have a Windows alternative.

I'll try and run some tasks on various projects this weekend under Linux, Windows 10 and windows 7 (I might even throw in some Windows XP! :o ) to get a better overall view.

Mark
chriscambridge
Active UBT Contributor 1+ yr
Posts: 2178
Joined: Mon Aug 08, 2016 1:56 pm
Location: UK

Re: Benchmarking Linux vs Windows

Post by chriscambridge »

It will be interesting to see if Linux has a similar credit output to Windows 7.

Personally I hate Win7 so I am going to move all hosts to Linux Mint now as its so simple and looks and works just like Windows (pretty much).

Although I'll probably keep all the hosts on a dual Win/Linux boot, just in case.
chriscambridge
Active UBT Contributor 1+ yr
Posts: 2178
Joined: Mon Aug 08, 2016 1:56 pm
Location: UK

Re: Benchmarking Linux vs Windows

Post by chriscambridge »

I have now moved the Linux install to the HD, and have also test-setup Linux on the V4 Xeon and I7-490 for further benchmarking.

Yesterday achieved the same results on the Thread Rippers for Linux over Windows 10 as the previous day. 45K vs 25K
Post Reply