Linux vs Windows performance.

Gray Handcock
Gray Handcock
Joined: 11 Mar 05
Posts: 211
Credit: 135,567
RAC: 0

Hello Um, no offence to

Hello

Um, no offence to Mandrake - it's a good system, but not exactly noted for its speed - it is designed to do mostly everything XP can do, as an efficient workstation, minus the virii, licensing and spyware hassles. The thing about Linux is that not all distros are the same, unlike Windows XP, where while there are 2 versions (for normal users at least) they are similar on the performance front.

The suggestion I made earlier in this thread is the idea of using a distro like Gentoo and doing a GUI-less install which can be optimised for a particular box, as a BOINC cruncher - unfair advantage: yup but hey, if you are not going to use the box for anything else, why not. My other option of using Damn Small Linux doesn't require compilation, is a 50 megs download, and has a basic GUI - and you don't even have to install it - just run in RAM (just don't ever power down...). Both of the above would knock the socks off any box on similar hardware running Windows.

Hope this is of some use

Gray

Jordan Wilberding
Jordan Wilberding
Joined: 19 Feb 05
Posts: 162
Credit: 715,454
RAC: 0

RE: Hello Um, no offence

Message 24874 in response to message 24873

Quote:

Hello

Um, no offence to Mandrake - it's a good system, but not exactly noted for its speed - it is designed to do mostly everything XP can do, as an efficient workstation, minus the virii, licensing and spyware hassles. The thing about Linux is that not all distros are the same, unlike Windows XP, where while there are 2 versions (for normal users at least) they are similar on the performance front.

The suggestion I made earlier in this thread is the idea of using a distro like Gentoo and doing a GUI-less install which can be optimised for a particular box, as a BOINC cruncher - unfair advantage: yup but hey, if you are not going to use the box for anything else, why not. My other option of using Damn Small Linux doesn't require compilation, is a 50 megs download, and has a basic GUI - and you don't even have to install it - just run in RAM (just don't ever power down...). Both of the above would knock the socks off any box on similar hardware running Windows.

Hope this is of some use

Gray

I don't think it would "knock of the socks off of the box", but probably would be a little better.

Also, I don't think anything is an "unfair advantage" :) We are all here to do some good, and maybe be a little competitive while we are at it, but the more credits the merrier!

such things just should not be writ so please destroy this if you wish to live 'tis better in ignorance to dwell than to go screaming into the abyss worse than hell

Gray Handcock
Gray Handcock
Joined: 11 Mar 05
Posts: 211
Credit: 135,567
RAC: 0

Hello Let me put it this

Hello

Let me put it this way - my Windows XP half of a dual-boot would take 8 hours 15 minutes, give or take the odd minute, to do one of the old Einstein WUs. Damn Small Linux did the same in just over 7 hours. I don't have a spare un-used partition to try a non-gui version of Gentoo, but having compared compiling speeds between Gentoo sans GUI and Gentoo with GUI, I would expect some serious speeds.

Gray

BTW Gentoo has still masked the 2 source code versions of BOINC that are available through the automatic installer, Portage, but one could always grab a completed version and install it through the sh command into a user directory.

FalconFly
FalconFly
Joined: 16 Feb 05
Posts: 191
Credit: 15,650,710
RAC: 0

I didn't mention one detail

Message 24876 in response to message 24875

I didn't mention one detail :

All my Linux boxes are stripped, text-only Terminals with minimal Services running only. No GUI, no fancy stuff, only pure SSH-connected dedicated crunching Clients :)

So regardless of Linux Version (at least those several that I tried), the performance is always very close to its theoretical maximum.
(configured that way, even a SuSE 8.1 installation runs very fast *g*)

BOINCview usually indicates CPU efficiency between 99.99 and 100%.

Gray Handcock
Gray Handcock
Joined: 11 Mar 05
Posts: 211
Credit: 135,567
RAC: 0

Hello So you reckon a

Hello

So you reckon a stripped linux box is running BOINC slightly slower than a normal Windows XP ? Assuming identical hardware I, quite frankly am amazed ! I admit that the KDE GUI in Linux affects performance quite a bit, at least in the earlier versions thereof (apparently the KDE 3.51 is pretty zappy) but the Explorer GUI for XP is not known for it's speed either...

I do know that there is a difference in the way the benchmarks are achieved, as I discovered in my dual-boot of WinXP and Damn Small Linux. According to the benchmarks an old Einstein WU should have taken around 12 hours or thereabouts in Linux. Somehow the benchmarks in Linux seem to differ from the same benchmarks in WinXP on identical hardware, yet the time taken to do a WU is reduced in Damn Small Linux for me.

The thought occurs to me that one of the reason for XP seeming slower to me is the fact that I run a firewall (even behind the router) as well as an antivirus.... but then my XP has been well tweaked, far more than the Gentoo side which I now run. This is where I wish I had a spare machine to play with !

We are going through a series of power outages here in Cape Town. I think once they are over I'll use the Damn Small Linux live CD to test times again - BOINC is very different nowadays. I also want to compile BOINC in Gentoo once it comes out of the "masked" category and see how that works, compared to a binary that is sh into place.
Gray

FalconFly
FalconFly
Joined: 16 Feb 05
Posts: 191
Credit: 15,650,710
RAC: 0

RE: Hello So you reckon a

Message 24878 in response to message 24877

Quote:

Hello

So you reckon a stripped linux box is running BOINC slightly slower than a normal Windows XP ?

Yes, but the reason for that is not the Linux Installation. It is the EAH Client code.
So no Linux you install or Tweak will help that, until the Staff further optimizes the Client Code.

Gray Handcock
Gray Handcock
Joined: 11 Mar 05
Posts: 211
Credit: 135,567
RAC: 0

Hello Just a short

Hello

Just a short comparison:

On my dual boot system, I have recently done WU in 4 hours 40 minutes in Gentoo Linux. In Windows XP this same size WU is done in 5 hours 18 minutes. The difference may well lie in my use of antivirus + firewall in Windows XP, while Gentoo Linux has neither.

Gray

Michael Karlinsky
Michael Karlinsky
Joined: 22 Jan 05
Posts: 888
Credit: 22,245,334
RAC: 0

RE: Hello Just a short

Message 24880 in response to message 24879

Quote:

Hello

Just a short comparison:

On my dual boot system, I have recently done WU in 4 hours 40 minutes in Gentoo Linux. In Windows XP this same size WU is done in 5 hours 18 minutes. The difference may well lie in my use of antivirus + firewall in Windows XP, while Gentoo Linux has neither.

Gray

Still looking for an AMD system for comparison. And with the upcoming
new application based on akosf optimization we might have to restart
this thread...

Michael

user32094
user32094
Joined: 22 Feb 05
Posts: 22
Credit: 74,648
RAC: 0

It seems that these kernel

Message 24881 in response to message 24880

It seems that these kernel configuration options should benefit scientific computation system at least a few percents. That's what I use anyway and the optimised S39L albert runs at expected speed under Wine 0.9.9 and BOINC 4.19 on AMD XP 2100 when comparing to Win98SE, AMD XP 2800. The expected speed factor is 1733 MHz/2083 MHz, which I got when running Cinebench and POV-Ray benchmark on both systems.

Processor type and features

  • * Processor family (Athlon/Duron/K7, ...)
    * Preemption Model (No Forced Preemption)
    * Timer frequency (100 Hz)

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.