Is it just my imagination .....

taso
taso
Joined: 21 Mar 05
Posts: 3
Credit: 77,757
RAC: 0
Topic 190843

.... or do Windows clients claim more credit than Linux/Unix clients for processing the same work unit?

Michael Roycraft
Michael Roycraft
Joined: 10 Mar 05
Posts: 846
Credit: 157,718
RAC: 0

Is it just my imagination .....

Not your imagination at all - the benchmark code in BOINC doesn't fare as well under Linux/Unix, as compared to Windows. Since claimed credit is a function of the sum of the two benchmarks multiplied by processing time, that suffers as well. The difference used to be much worse with the one of the previous versions of the BOINC code, and efforts have been underway to sort the benchmarking business once and for all. A matter of time...

microcraft
"The arc of history is long, but it bends toward justice" - MLK

taso
taso
Joined: 21 Mar 05
Posts: 3
Credit: 77,757
RAC: 0

RE: Not your imagination at

Message 25506 in response to message 25505

Quote:
Not your imagination at all - the benchmark code in BOINC doesn't fare as well under Linux/Unix, as compared to Windows. Since claimed credit is a function of the sum of the two benchmarks multiplied by processing time, that suffers as well. The difference used to be much worse with the one of the previous versions of the BOINC code, and efforts have been underway to sort the benchmarking business once and for all. A matter of time...


Seems to me there are two distinct issues:

1. Given the same hardware and work unit, I would expect a result in
approximately the same amount of time regardless of OS. This was
nowhere near happening when I first looked at e@home but it now seems
to be more or less Ok.

2. Given a work unit, I would expect any computer to claim approximately
the same credit regardless of hardware or OS. This doesn't seem to be the
case. The bizarreness of it all is that you can have 6 identical hardware
configs - 3 Windows and 3 Linux. If 3 Windows hosts process the WU, they
will receive more credit than 3 Linux host processing the same WU.

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

RE: 1. Given the same

Message 25507 in response to message 25506

Quote:


1. Given the same hardware and work unit, I would expect a result in
approximately the same amount of time regardless of OS. This was
nowhere near happening when I first looked at e@home but it now seems
to be more or less Ok.

2. Given a work unit, I would expect any computer to claim approximately
the same credit regardless of hardware or OS. This doesn't seem to be the
case. The bizarreness of it all is that you can have 6 identical hardware
configs - 3 Windows and 3 Linux. If 3 Windows hosts process the WU, they
will receive more credit than 3 Linux host processing the same WU.

1. How do you know your computers have the "exact" same work unit? In E@H there are 3 types of workunits right now, and each vary wildly in the amount of relative time it takes to complete each one.

2. As said before, this is because you are probably not using "optimized" boinc client. When you first run boinc it tests your computer's speed(this independent of E@H). If you are using the ordinary i386 binaries for Linux, then it isn't showing the real benchmark of the system.

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

arcturus
arcturus
Joined: 11 Feb 05
Posts: 44
Credit: 1,008,160
RAC: 0

Here's a specific, actionable

Here's a specific, actionable solution !

Download & run optimized linux client: http://www.guntec.de/Crunch3r/boincx86.html

See claimed credit rise. How much? Check out my WU's reported March 2 and note how it triples.

http://einsteinathome.org/host/555885/tasks

Now everyone get back to work and stop overanalyzing.

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

RE: Now everyone get back

Message 25509 in response to message 25508

Quote:

Now everyone get back to work and stop overanalyzing.

Calm down.

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

arcturus
arcturus
Joined: 11 Feb 05
Posts: 44
Credit: 1,008,160
RAC: 0

RE: RE: Now everyone get

Message 25510 in response to message 25509

Quote:
Quote:

Now everyone get back to work and stop overanalyzing.

Calm down.

No happy pill today huh?

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

RE: RE: RE: Now

Message 25511 in response to message 25510

Quote:
Quote:
Quote:

Now everyone get back to work and stop overanalyzing.

Calm down.

No happy pill today huh?

I am the one having fun :) No pills even!

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

Gerry Rough
Gerry Rough
Joined: 1 Mar 05
Posts: 102
Credit: 1,847,066
RAC: 0

RE: RE: RE: RE: Now

Message 25512 in response to message 25511

Quote:
Quote:
Quote:
Quote:

Now everyone get back to work and stop overanalyzing.

Calm down.

No happy pill today huh?

I am the one having fun :) No pills even!

I don't get too much time on the message boards, but I would be very much a happy camper if someone told me of a windows optimized client for E@H. Does such a thing exist, and where can I find it? How much better are the results of said upgrade?

Gerry Rough


(Click for detailed stats)

taso
taso
Joined: 21 Mar 05
Posts: 3
Credit: 77,757
RAC: 0

RE: 1. How do you know

Message 25513 in response to message 25507

Quote:

1. How do you know your computers have the "exact" same work unit? In E@H there are 3 types of workunits right now, and each vary wildly in the amount of relative time it takes to complete each one.

It's not hard to compare how other different computers handle the same work unit the results are published. Eventually you get a feel for how your computer(s) performs against other similar ones (hardware). With the early e@home clients my Linux system was taking way longer to produce a result than much slower computers running Windows. I lost interest at that point - if I was going to contribute to global (and local) warming I expected the e@home developers to maximize the return per unit of pollution. With the current e@home client Linux
clients are at least in the same ball park as Windows ones.

Quote:

2. As said before, this is because you are probably not using "optimized" boinc client. When you first run boinc it tests your computer's speed(this independent of E@H). If you are using the ordinary i386 binaries for Linux, then it isn't showing the real benchmark of the system.

My Linux computer seems to produce results as efficiently as a Windows one. It is an anomaly that it doesn't claim as much credit as a similar computer running Windows, but it doesn't bother me too much.

KSMarksPsych
KSMarksPsych
Moderator
Joined: 15 Oct 05
Posts: 2,702
Credit: 4,090,227
RAC: 0

RE: RE: RE: RE: Quote

Message 25514 in response to message 25512

Quote:
Quote:
Quote:
Quote:
Quote:

Now everyone get back to work and stop overanalyzing.

Calm down.

No happy pill today huh?

I am the one having fun :) No pills even!

I don't get too much time on the message boards, but I would be very much a happy camper if someone told me of a windows optimized client for E@H. Does such a thing exist, and where can I find it? How much better are the results of said upgrade?

Gerry Rough

ATM, there is no E@H optimized app. THis is b/c the science app isn't open source.

Kathryn

Kathryn :o)

Einstein@Home Moderator

Comment viewing options

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