Safe to switch to Linux now?

th3
th3
Joined: 24 Aug 06
Posts: 208
Credit: 2,208,434
RAC: 0

RE: But are you saying here

Message 70911 in response to message 70910

Quote:
But are you saying here that you just don't trust the cpu seconds reported for a work unit by WIN vs LINUX?


Yes, seconds of CPU time used is not reliable for comparision,
very big difference on the same PC:

WinXP 49,483 sec, 389.75 credit WU's
Linux ~49,500 sec, 433.63 credit WU's

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5,227
Credit: 44,564,853,502
RAC: 39,563,270

RE: Yes, seconds of CPU

Message 70912 in response to message 70911

Quote:

Yes, seconds of CPU time used is not reliable for comparision,
very big difference on the same PC:

WinXP 49,483 sec, 389.75 credit WU's
Linux ~49,500 sec, 433.63 credit WU's

My observation (based on quite a number of different machines) is that Linux currently is between about 1% to 10% faster than Windows XP on the same hardware. It does vary from architecture to architecture so you can't put down a single definitive % value. However, in your case, it's quite possible that the observed 11% difference is due mainly to Linux and has nothing to do with any supposed inaccuracy of the CPU seconds reported. Also it's a bit risky to base your figure for WinXP on a single currently observable result unless you are sure that several other (now deleted) results were reporting the same time. There can be some variation in the time reported by results having exactly the same credit.

Cheers,
Gary.

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3,516
Credit: 457,880,261
RAC: 89,388

RE: RE: But are you

Message 70913 in response to message 70911

Quote:
Quote:
But are you saying here that you just don't trust the cpu seconds reported for a work unit by WIN vs LINUX?

Yes, seconds of CPU time used is not reliable for comparision,
very big difference on the same PC:

WinXP 49,483 sec, 389.75 credit WU's
Linux ~49,500 sec, 433.63 credit WU's

But this doesn't necessarily show that Linux is reporting CPU seconds differently than Win. It could also mean that the version of the Linux app this figure was measured on runs significantly faster than the Windows app.

From what I see in your result pages, the Windows app and the Linux app theses figures were taken with are from different "generations" of the sooftware (4.21 and 4.30) so you can't directly compare the results.

CU

BRM

th3
th3
Joined: 24 Aug 06
Posts: 208
Credit: 2,208,434
RAC: 0

That XP result was very

That XP result was very typical, yes.

Maybe Linux is a bit faster, i dont know, yet. It is faster than windows in most other things so it would be natural that it is crunching a bit faster.

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3,516
Credit: 457,880,261
RAC: 89,388

RE: That XP result was very

Message 70915 in response to message 70914

Quote:

That XP result was very typical, yes.

Maybe Linux is a bit faster, i dont know, yet. It is faster than windows in most other things so it would be natural that it is crunching a bit faster.

It doesn't have to do anything with the operating system itself, it's more related to the compilers used. So the generated code is different on both platforms, and some machines will like the output from one compiler more than from the other compiler. Usually the differences are smaller now than in your case, tho.

CU

BRM

th3
th3
Joined: 24 Aug 06
Posts: 208
Credit: 2,208,434
RAC: 0

I would think compilers is

I would think compilers is one of the reasons why Linux is faster in general, arent the compilers also better because Linux is a better OS? =)

Im planning to order a Pentium E2140 + a Gigabyte G31 mATX board (very cheap, Intel chipset at Via/SiS price) and use some old parts here to set up a mATX computer, that could be a good opportunity to test both OS on a computer not being used for anything else.

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3,516
Credit: 457,880,261
RAC: 89,388

RE: I would think compilers

Message 70917 in response to message 70916

Quote:
I would think compilers is one of the reasons why Linux is faster in general, arent the compilers also better because Linux is a better OS? =)

:-) Weeeell, the Intel compiler is good on any platform, even Windows ... unless you have an AMD CPU :-), but that's another story.

Quote:

Im planning to order a Pentium E2140 + a Gigabyte G31 mATX board (very cheap, Intel chipset at Via/SiS price) and use some old parts here to set up a mATX computer, that could be a good opportunity to test both OS on a computer not being used for anything else.

That should be a nice cruncher with reasonable power consumption.

Please let us know the results of the comparison when the box is ready.

CU
BRM

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1,314
Credit: 1,750,580,441
RAC: 924,174

In s5r1 when the hot loops of

In s5r1 when the hot loops of all the x86 platforms were written in identical assembly and thus freed from issues with the difference in compiler behavior there weren't any significant cross OS performance differences. The code that actually took 99% of hte CPU time was indentical and none of the OSes used a large enough fraction of the CPU time for overhead to be a non trivial factor. The old s4 applications were significantly faster in windows than *nix even before akos started tweaking them.

Dave Burbank
Dave Burbank
Joined: 30 Jan 06
Posts: 275
Credit: 1,548,376
RAC: 0

I find the Linux app to be a

I find the Linux app to be a bit faster. My dual-booted Q6600 runs a 449.41 credit WU in (average) 53,600s under Linux, while the Windows app runs a 407.18 credit WU in 50,000s.

Hypothetically the Linux app would run 23.5 minutes faster than the Windows app for that same 407 credit WU.

Not much I admit, but for a system running 24/7 those saved minutes turn into extra WU's crunched.

There are 10^11 stars in the galaxy. That used to be a huge number. But it's only a hundred billion. It's less than the national deficit! We used to call them astronomical numbers. Now we should call them economical numbers. - Richard Feynman

th3
th3
Joined: 24 Aug 06
Posts: 208
Credit: 2,208,434
RAC: 0

Just finished the first WU

Just finished the first WU with the Linux 4.35 app, considerably slower than 4.21.

Computers results
The one reported on "23 Aug 2007 5:16:07 UTC" is the first 4.35 result, looks like the ones in progress will be very close to that CPU time.

Comment viewing options

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