Windows S5R4 SSE2 power App 6.05 available

Stan Pope
Stan Pope
Joined: 22 Dec 05
Posts: 80
Credit: 426811575
RAC: 0

RE: Could you identify

Message 86620 in response to message 86619

Quote:
Could you identify which system is the system with the thermal problem?


Yes ... 1414352 is running at about 30% cuz if I go much higher, it does a thermal shutdown at odd times. Fan runs high continuously. CPU cooling fins appear to be clear of dust, but I haven't disassembled to be sure ... too much to do at the moment to try that. I assumed that when it started the behavior (a year after startup) that I just needed to blow the dust out, but that did not suffice. I think that most of the last reported WU was run on the SSE2 version. Ahhh... it may have finished processing using the version in the "slots" folder? BOINC has put the correct version in the "slots" folder.

6.05 installation procedure identical on all 9 Einstein boxes. Put new exe, dbg and app_info files on a thumb drive, stop BOINC, copy the three files into the einstein folder, and restart BOINC.

The other is 1107723 housed in a mini-case. It like the remaining 8, run 100% 7X24 and only go down for a power outage or required maintenance reboot.

Stan

Byron S Goodgame
Byron S Goodgame
Joined: 16 Jan 06
Posts: 187
Credit: 56581
RAC: 0

RE: 6.05 installation

Message 86621 in response to message 86620

Quote:
6.05 installation procedure identical on all 9 Einstein boxes. Put new exe, dbg and app_info files on a thumb drive, stop BOINC, copy the three files into the einstein folder, and restart BOINC.

The version of BOINC your using 5.10.13 is pretty dated. I'd try going up to version 6.2.19 or 5.10.45 at the very least and see if you don't get better results with that.

Stan Pope
Stan Pope
Joined: 22 Dec 05
Posts: 80
Credit: 426811575
RAC: 0

RE: RE: 6.05 installation

Message 86622 in response to message 86621

Quote:
Quote:
6.05 installation procedure identical on all 9 Einstein boxes. Put new exe, dbg and app_info files on a thumb drive, stop BOINC, copy the three files into the einstein folder, and restart BOINC.

The version of BOINC your using 5.10.13 is pretty dated. I'd try going up to version 6.2.19 or 5.10.45 at the very least and see if you don't get better results with that.

Been up to later 5.10.xx release, but backed off because I didn't like how it ran. I'd be really surprised if version-up there helped performance in the app ... but stranger things have happened.

Stan

Byron S Goodgame
Byron S Goodgame
Joined: 16 Jan 06
Posts: 187
Credit: 56581
RAC: 0

RE: Been up to later

Message 86623 in response to message 86622

Quote:
Been up to later 5.10.xx release, but backed off because I didn't like how it ran. I'd be really surprised if version-up there helped performance in the app ... but stranger things have happened.

To be honest I'm not very confident that it will work either. Truth is, I've seen the 2nd pc results and they're as unchanged as the 1st one to the new app. I've looked at what info is available in the WU's as well as on your pc's info, and see nothing that's a clear sign to me why you aren't getting better results. I'm scraping the bottom of the barrel here, and pretty much am at the state where you are. If it turns out this happens to be the cause, stranger things have happened.
[Edit] I guess we could always go into if both pc's have the same type of memory and doing test on those as well as other test, but your stock times don't look abnormal to me, so I'm doubtful there too.[/Edit]

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6588
Credit: 316668077
RAC: 353103

RE: 2. Went looking early

Message 86624 in response to message 86615

Quote:
2. Went looking early in the thread for a large chunk of text I recalled seeing earlier. It was missing ... Was it moved? deleted?


Longer threads will have sited above the most distant one displayed on your screen ( at the bottom for me, as I have 'Newest post first' selected ) a message like:

Only the first post and the last Y posts (of the X posts in this thread) are displayed.

and a hotlink like:

Click here to also display the remaining posts.

Cheers, Mike.

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

Byron S Goodgame
Byron S Goodgame
Joined: 16 Jan 06
Posts: 187
Credit: 56581
RAC: 0

RE: but stranger things

Message 86625 in response to message 86622

Quote:
but stranger things have happened.

Looks like you did get a little improvement on this one, running full 6.0.5, (though I was expectng it to be a bit more dramitic), and runnng the BOINC 5.10.13. I do see more memory in this one so could be that has something to do with the difference as well as this being the one I think you said you throtle.

Byron S Goodgame
Byron S Goodgame
Joined: 16 Jan 06
Posts: 187
Credit: 56581
RAC: 0

RE: but stranger things

Message 86626 in response to message 86625

Quote:
but stranger things have happened.

Looks like you did get a little improvement on this one, running full 6.0.5, (though I was expectng it to be a bit more dramitic, maybe because it's throtled it's not up where it should be), and runnng the BOINC 5.10.13. I do see more memory ( and maybe different? ) in this one, so could be that has something to do with the difference from the other AMD.

Brian Silvers
Brian Silvers
Joined: 26 Aug 05
Posts: 772
Credit: 282700
RAC: 0

RE: RE: but stranger

Message 86627 in response to message 86626

Quote:
Quote:
but stranger things have happened.

Looks like you did get a little improvement on this one, running full 6.0.5, (though I was expectng it to be a bit more dramitic, maybe because it's throtled it's not up where it should be), and runnng the BOINC 5.10.13. I do see more memory ( and maybe different? ) in this one, so could be that has something to do with the difference from the other AMD.

Bear in mind the cyclic nature of the runtimes of these tasks. There is a way to calculate cycle period, but I don't know if the people that work on that (archae86, Richard Haselgrove, Mike Hewson, Gary Roberts, et al) have fully perfected it for this science run (S5R4), but the basic idea is that you're going to have a peak and a trough in runtimes. Initially I thought there was not much improvement with my AMD, as times were hovering pretty consistently around 39000-40000 seconds. Now, however, it is only taking 32000 seconds.

Byron S Goodgame
Byron S Goodgame
Joined: 16 Jan 06
Posts: 187
Credit: 56581
RAC: 0

RE: Bear in mind the cyclic

Message 86628 in response to message 86627

Quote:
Bear in mind the cyclic nature of the runtimes of these tasks. There is a way to calculate cycle period, but I don't know if the people that work on that (archae86, Richard Haselgrove, Mike Hewson, Gary Roberts, et al) have fully perfected it for this science run (S5R4), but the basic idea is that you're going to have a peak and a trough in runtimes. Initially I thought there was not much improvement with my AMD, as times were hovering pretty consistently around 39000-40000 seconds. Now, however, it is only taking 32000 seconds.

Well I hope I can look forward to that additional gain as well, since mine have been and are hovering around where yours were.. I was doing more of a comparison of his and my pc's from the standpoint that his was a 2.5 AMD and mine is a 2.4 (though his is a dual core), along with my Intel 2.0 dual (which completes them in about the same time as he's getting now), to try and gauge what a pc like his should be able to achieve. I figured his cpu time should be closer to my AMD than my Intel, though I know we're crunching different WU's, over all I thought there should be a similar comparison. I know it wouldn't be precise and his numbers are dropping more into the ball park from the last look I had.

archae86
archae86
Joined: 6 Dec 05
Posts: 3157
Credit: 7221924931
RAC: 956394

RE: Bear in mind the cyclic

Message 86629 in response to message 86627

Quote:
Bear in mind the cyclic nature of the runtimes of these tasks. There is a way to calculate cycle period, but I don't know if the people that work on that (archae86, Richard Haselgrove, Mike Hewson, Gary Roberts, et al) have fully perfected it for this science run (S5R4), but the basic idea is that you're going to have a peak and a trough in runtimes.


I don't like the functional form which seems more complicated than ought to be necessary, but I still think the estimator I posted some weeks ago gives usefully accurate answers for the cycle length for S5R4

cycle length(freq)=0.42+.00029*(ceiling(freq,10)+1)^2

Making conclusions about performance comparison without taking into account the cycle can certainly be very misleading.

Comment viewing options

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