Linux slower then Windows ( both 64 bit)

[Wacky]

Originally Posted by pepi37

Since I have quad, it shows 4 same named processes

It would, perhaps be better if you would actually provide the details of your observations and not just your conclusions. For example, what CPU are you using? Are the hard drives equal in speed and internal caching, how much RAM does the system have, etc.

The base execution times (Windows vs Linux) are similar. I find it suspect that on Windows you find little variance, but significantly more variance on Linux.

This points to some additional process running on Linux. (Hence the previous request for "top" information.)

[Dubslow]

What he's saying is that even though you're using the same hardware, it could be that for whatever reason Linux doesn't like a certain kind of processor. Just copy and paste what top gives you, and you can use the "[ code ]Code goes here[/ code]" (without the spaces) to make it look better.

Without the spaces you'll see

Code goes here

[axn]

In some linux variants, the default setting is to throttle the CPU down when _only_ low priority processes are running. Could be that.

@pepi37: the core part of the computation of PFGW (which is what PRPNet uses, I think), uses the same assembly routines in both Windows and Linux versions. So you're quite wrong in concluding that the problem is with "non-optimized" linux version.

[retina]

Originally Posted by axn

@pepi37: the core part of the computation of PFGW (which is what PRPNet uses, I think), uses the same assembly routines in both Windows and Linux versions. So you're quite wrong in concluding that the problem is with "non-optimized" linux version.

Erm, core computations would not be concerned with the OS so I don't think your example says anything. To optimise for an OS would involve setting things like priority and affinity (and lots of other things) to give the core code the best chance to get lots of CPU cycles.

[axn]

Originally Posted by retina

Erm, core computations would not be concerned with the OS so I don't think your example says anything.

Ermmm... What example?

[retina]

Originally Posted by axn

Ermmm... What example?

Originally Posted by axn

the core part of the computation of PFGW (which is what PRPNet uses, I think), uses the same assembly routines in both Windows and Linux versions.

.

[axn]

Originally Posted by retina

.

Ah. Not an example -- I was replying to the topic of discussion. To wit:

Originally Posted by pepi37

I am on PRPNet now.
<snip>
windows
ppsLow server WU is finished from 79-80 second less then 1 sec
linux - same WU 79-90 second.

I was also trying to disabuse the OP of the notion that there is some kind of anti-linux bias involved.

Originally Posted by pepi37

Problem is always not well optimized application for Linux, and very well optimized app for Windows.

[pepi37]

Originally Posted by Wacky

It would, perhaps be better if you would actually provide the details of your observations and not just your conclusions. For example, what CPU are you using? Are the hard drives equal in speed and internal caching, how much RAM does the system have, etc.

The base execution times (Windows vs Linux) are similar. I find it suspect that on Windows you find little variance, but significantly more variance on Linux.

This points to some additional process running on Linux. (Hence the previous request for "top" information.)

Since there is no GUI on Linux ( I think that will be fastest version) I cannot ( or dont know) how to copy and upload top information on this forum.
Since my kernel is so small, I can assure you that there is no undesirable process. As I say before top shows 99-100 % ( in most cases 100%) using of core per process, so CPU utilization is on maximum.
Last thing: I always used one memory module , and last night I added second one. Computation times are dropped in bot cases ( Windows and Linux) and now is nearly same: difference is less then seconds in Windows or Linux. So it looks like that activating dual channel mode on motherboard solved all problems. And also looks like windows are not so sensitive like like to that option.

[Xyzzy]

Since there is no GUI on Linux ( I think that will be fastest version) I cannot ( or dont know) how to copy and upload top information on this forum.

$ top -n 1 > file.txt

[chris2be8]

Originally Posted by pepi37

Since there is no GUI on Linux ( I think that will be fastest version) I cannot ( or dont know) how to copy and upload top information on this forum.
Since my kernel is so small, I can assure you that there is no undesirable process. As I say before top shows 99-100 % ( in most cases 100%) using of core per process, so CPU utilization is on maximum.
Last thing: I always used one memory module , and last night I added second one. Computation times are dropped in bot cases ( Windows and Linux) and now is nearly same: difference is less then seconds in Windows or Linux. So it looks like that activating dual channel mode on motherboard solved all problems. And also looks like windows are not so sensitive like like to that option.

We still don't have enough information to tell what's going on.
What CPU do you have?
What sort of motherboard is it on?
How big are the memory modules and how fast are they?
What disk(s) do you have?
How much swap space (on LInux type "free -t")?
What version of Windows are you running?
What Linux distribution is it? Post output from "uname -a" and "ls -l /etc/*release".
What program versions are you running? Where did they come from? How do you switch from Windows to Linux (dual boot)?

If you are using a Linux distro that uses more memory than the Windows version it might have been paging before you added memory. That could explain the different speed.

Chris K