Lunatics Windows Installer v0.40 release notes

Message boards : Number crunching : Lunatics Windows Installer v0.40 release notes
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 4 · 5 · 6 · 7 · 8 · 9 · 10 . . . 14 · Next

AuthorMessage
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1211290 - Posted: 28 Mar 2012, 22:41:49 UTC - in response to Message 1211285.  
Last modified: 28 Mar 2012, 22:45:00 UTC

Will those work to fix this on the ATI client as well? I don't understand why half of my WUs run normally and half of them crap out.



Worth a try. Looks like Something OS scheduling related has changed, so we may well be dealing with prioity issues all around.

Fred's efMer Priority tool (v1.2) here

Process lasso (Free & commercial versions I beleive
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1211290 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1211291 - Posted: 28 Mar 2012, 22:47:46 UTC - in response to Message 1211285.  

Will those work to fix this on the ATI client as well? I don't understand why half of my WUs run normally and half of them crap out.

What do you mean crap out? Error out? or just low GPU load?

Claggy
ID: 1211291 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1211292 - Posted: 28 Mar 2012, 22:48:44 UTC - in response to Message 1211286.  

Again, these were all done with 1 cpu core free.


Thanks,
nothing obvious except the one running longer for no apparent reason.

Wonder what the difference is between my machine, alpha testers & that machine. Well worth finding out.

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1211292 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1211295 - Posted: 28 Mar 2012, 22:52:01 UTC

Images all OK here too. John's most recent was http://62.131.164.180/seti/afterburner4.jpg - I don't know whether the IP address, rather than the more typical url-style, could have caused your problems?
ID: 1211295 · Report as offensive
JLConawayII

Send message
Joined: 2 Apr 02
Posts: 188
Credit: 2,840,460
RAC: 0
United States
Message 1211296 - Posted: 28 Mar 2012, 22:52:56 UTC - in response to Message 1211291.  
Last modified: 28 Mar 2012, 23:01:07 UTC

Will those work to fix this on the ATI client as well? I don't understand why half of my WUs run normally and half of them crap out.

What do you mean crap out? Error out? or just low GPU load?

Claggy



It's doing the same thing as the CUDA clients, low utilization and dropouts. I linked some of the WU's above. I've since run some more, with the same results. A few go splat due to noise, about half run normally at 86-90% utilization and take a little over 3 minutes, and half show the same splotchy, terrible utilization as shown for those CUDA clients, and run several times longer than normal. At first I thought it was a VLAR issue, since I noticed it doing this on VLARS sent to my card, but now I see it's just doing it randomly on normal WUs too.

these ran normally
http://setiathome.berkeley.edu/result.php?resultid=2373271874
http://setiathome.berkeley.edu/result.php?resultid=2373271870

and these didn't
http://setiathome.berkeley.edu/result.php?resultid=2373271868
http://setiathome.berkeley.edu/result.php?resultid=2373271862
ID: 1211296 · Report as offensive
S@NL - John van Gorsel
Volunteer tester
Avatar

Send message
Joined: 5 Jul 99
Posts: 193
Credit: 139,673,078
RAC: 0
Netherlands
Message 1211299 - Posted: 28 Mar 2012, 22:57:15 UTC - in response to Message 1211279.  

Thanks,
Can you jack up the cuda process priorities (with one of the tools mentioned) & compare ?

BTW: for some reason I'm having trouble getting to your images.

Jason


My bad. My firewall was blocking everything from Turkey on eastward. Made some changes and you should be able to see the images now.

I will try to boost cuda priorities tomorrow. If I remember correctly, Khangollo posted some info in a Linux related discussion about raising thread priorities in Linux so I will try the same with x41g on the Linux pc's
ID: 1211299 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1211300 - Posted: 28 Mar 2012, 22:58:23 UTC - in response to Message 1211295.  

Images all OK here too. John's most recent was http://62.131.164.180/seti/afterburner4.jpg - I don't know whether the IP address, rather than the more typical url-style, could have caused your problems?


I think so. Will have to investigate the cable later today.

On the priority issue, playing with Fred's priority tool between regular below normal & normal does make a noticeable impact on GPU usage. Really wish I hadn't been pressured into dropping those priorities now, oh well, better put some command line options in the next build.
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1211300 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1211302 - Posted: 28 Mar 2012, 23:01:33 UTC - in response to Message 1211299.  
Last modified: 28 Mar 2012, 23:02:24 UTC

Thanks,
Can you jack up the cuda process priorities (with one of the tools mentioned) & compare ?

BTW: for some reason I'm having trouble getting to your images.

Jason


My bad. My firewall was blocking everything from Turkey on eastward. Made some changes and you should be able to see the images now.

I will try to boost cuda priorities tomorrow. If I remember correctly, Khangollo posted some info in a Linux related discussion about raising thread priorities in Linux so I will try the same with x41g on the Linux pc's



That's better!

OK, yup, looks like priority to me. Fred's tool seems to be working for me, raising the x41 processes to 'normal'. I'll jimmy some options into the next build somehow

Jason
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1211302 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1211303 - Posted: 28 Mar 2012, 23:02:39 UTC - in response to Message 1211296.  
Last modified: 28 Mar 2012, 23:04:23 UTC

Will those work to fix this on the ATI client as well? I don't understand why half of my WUs run normally and half of them crap out.

What do you mean crap out? Error out? or just low GPU load?

Claggy



It's doing the same thing as the CUDA clients, low utilization and dropouts. I linked some of the WU's above. I've since run some more, with the same results. A few go splat due to noise, about half run normally at 86-90% utilization and take a little over 3 minutes, and half show the same splotchy, terrible utilization as shown for those CUDA clients, and run several times longer than normal.

There is a low ATI GPU usage Bug about, i have had it since Cat 10.5, ATI_HD5_ r374 and also on ATI AP apps since r521 all on my i7, But my E8500 when it had the HD5770 fitted didn't suffer from it,
Don't know if it's a driver Bug, an app Bug, a Windows scheduler Bug or what, I suspect it's an app Bug since PrimeGrid's OpenCl app hasn't suffered this on my HD5770,

If you suspend the task with low GPU usage, wait for it to exit, then resume you should get the GPU usage up again,

Claggy
ID: 1211303 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1211307 - Posted: 28 Mar 2012, 23:05:50 UTC - in response to Message 1211300.  

Images all OK here too. John's most recent was http://62.131.164.180/seti/afterburner4.jpg - I don't know whether the IP address, rather than the more typical url-style, could have caused your problems?

I think so. Will have to investigate the cable later today.

On the priority issue, playing with Fred's priority tool between regular below normal & normal does make a noticeable impact on GPU usage. Really wish I hadn't been pressured into dropping those priorities now, oh well, better put some command line options in the next build.

Don't forget that the priorities were dropped because some people want to use their computers for purposes other than BOINC, and have less than the latest and greatest hardware - then, screen lag and usability also matters. IIRC, you only raised the priority in the first place to ensure a level playing field in Beta testing with other players who were also gaming the priority arms race: x38g was also bounced from Beta into Release status at rather short notice.
ID: 1211307 · Report as offensive
JLConawayII

Send message
Joined: 2 Apr 02
Posts: 188
Credit: 2,840,460
RAC: 0
United States
Message 1211308 - Posted: 28 Mar 2012, 23:05:50 UTC

Meh, if that's the case I'm not going to babysit the thing to fix a random bug that may or may not show up every 3 minutes. I guess I will stick with the CPU clients for now, and by the time I build a new Nvidia machine hopefully the issue there will no longer exist.
ID: 1211308 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1211316 - Posted: 28 Mar 2012, 23:16:13 UTC - in response to Message 1211307.  
Last modified: 28 Mar 2012, 23:21:18 UTC

Images all OK here too. John's most recent was http://62.131.164.180/seti/afterburner4.jpg - I don't know whether the IP address, rather than the more typical url-style, could have caused your problems?

I think so. Will have to investigate the cable later today.

On the priority issue, playing with Fred's priority tool between regular below normal & normal does make a noticeable impact on GPU usage. Really wish I hadn't been pressured into dropping those priorities now, oh well, better put some command line options in the next build.

Don't forget that the priorities were dropped because some people want to use their computers for purposes other than BOINC, and have less than the latest and greatest hardware - then, screen lag and usability also matters. IIRC, you only raised the priority in the first place to ensure a level playing field in Beta testing with other players who were also gaming the priority arms race: x38g was also bounced from Beta into Release status at rather short notice.


All true. I have to stop short of asking the 'victims' here for lists of all the programs & services they run. If they want regular full GPU usage, then for now they'll have to either:
- locate what's shoving the Cuda threads aside & stop the culprits, (e.g. BlackViper their systems)
- use priority tools such as those posted, or
- put up with erratic GPU usage

I'm still reluctant to add complexity to the X series with command line options, considering that its main successes are attributable to its simplicity. I feel left with no choice at this point to offer the option to these guys... For example Victor was borderline abusive over the issue,.

Well, we'll see how the best course reveals itself over the next week, as some GTX 680 related cosmetic fixes are incorporated also.

Jason
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1211316 · Report as offensive
S@NL - John van Gorsel
Volunteer tester
Avatar

Send message
Joined: 5 Jul 99
Posts: 193
Credit: 139,673,078
RAC: 0
Netherlands
Message 1211319 - Posted: 28 Mar 2012, 23:22:26 UTC - in response to Message 1211307.  

Don't forget that the priorities were dropped because some people want to use their computers for purposes other than BOINC, and have less than the latest and greatest hardware - then, screen lag and usability also matters.


A minor addition to the v0.40 readme file can save a lot of dissapointment:

"For the best performance without compromising usability it is advisable to leave one cpu thread free by setting the option On multiprocessor systems use at most 75% of the processors. (for quadcores that is..)
Those who aim for the highest production should install a priority raising programme such as..."
ID: 1211319 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1211320 - Posted: 28 Mar 2012, 23:22:43 UTC - in response to Message 1211316.  

Images all OK here too. John's most recent was http://62.131.164.180/seti/afterburner4.jpg - I don't know whether the IP address, rather than the more typical url-style, could have caused your problems?

I think so. Will have to investigate the cable later today.

On the priority issue, playing with Fred's priority tool between regular below normal & normal does make a noticeable impact on GPU usage. Really wish I hadn't been pressured into dropping those priorities now, oh well, better put some command line options in the next build.

Don't forget that the priorities were dropped because some people want to use their computers for purposes other than BOINC, and have less than the latest and greatest hardware - then, screen lag and usability also matters. IIRC, you only raised the priority in the first place to ensure a level playing field in Beta testing with other players who were also gaming the priority arms race: x38g was also bounced from Beta into Release status at rather short notice.

All true. I have to stop short of asking the 'victims' here for lists of all the programs & services they run. If they want regular full GPU usage, then for now they'll have to either:
- locate what's shoving the Cuda threads aside & stop the culprits, (e.g. BlackViper their systems)
- use priority tools such as those posted, or
- put up with erratic GPU usage

I'm still reluctant to add complexity to the X series with command line options, considering that it's main successes are attributable to its simplicity. I feel left with no choice at this point to offer the option to these guys... For example Victor was borderline abusive over the issue,.

Well, we'll see how the best course reveals itself over the next week, as some GTX 680 related cosmetic fixes are incorporated also.

Jason

Maybe - just brainstorming here - we ought to consider building something into the installer. We could have an option switch, or two versions, or ... whatever, let me sleep on it.

But if there was a choice of "built for comfort" versus "built for speed" - would that help people? I haven't yet seen Victor's comments (although I've asked him for feedback), so I don't know exactly what the problem is.
ID: 1211320 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1211323 - Posted: 28 Mar 2012, 23:29:09 UTC - in response to Message 1211321.  
Last modified: 28 Mar 2012, 23:29:39 UTC

/HUG Jason


LoL, thanks, helps a lot :D

...Now everyone hold hands, lol.


Well I think indeed sleeping on the problem is going to be the best idea. There are the mentioned workarounds of freeing cores & raising priorties, so we're not stuck yet :)

Thanks all
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1211323 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1211326 - Posted: 28 Mar 2012, 23:42:26 UTC - in response to Message 1211319.  

Don't forget that the priorities were dropped because some people want to use their computers for purposes other than BOINC, and have less than the latest and greatest hardware - then, screen lag and usability also matters.

A minor addition to the v0.40 readme file can save a lot of dissapointment:

"For the best performance without compromising usability it is advisable to leave one cpu thread free by setting the option On multiprocessor systems use at most 75% of the processors. (for quadcores that is..)
Those who aim for the highest production should install a priority raising programme such as..."

At this point, you have to start thinking about who the installer (as opposed to the applications it contains) was designed for - and what sort of people run BOINC in the first place.

Some figures. The new 0.40 installer has been downloaded 680 times since we launched it, less than 60 hours ago. The previous version 0.39, released at the end of November 2011, had been downloaded 5,300 times when we started on this one. Older versions, which were available for longer, were downloaded over 10,000 times.

How many people have posted in this thread? Without counting, I'd guess at fewer than 20. What do the other 660 think?

John, your thoughts are great, and we'll certainly consider them. But they are all geared towards increasing the GPU loading. What about people who want, or need, to reduce it for whatever reason? I, for example, choose to use only GPU apps at SETI: my CPUs are all busy on other BOINC projects. Fewer than a third of my total BOINC credits are shown alongside my name here. So, I would resist advice to reduce CPU usage to favour SETI. (Though I have no objection to information of that nature: IF you do this, the CONSEQUENCE might be... - or IF you want to achieve that, the techniques you could use INCLUDE the other.)

We're going to need a lot of group hugs, and a lot of sleep, before we work out the answer to this one ;P
ID: 1211326 · Report as offensive
Profile zoom3+1=4
Volunteer tester
Avatar

Send message
Joined: 30 Nov 03
Posts: 65709
Credit: 55,293,173
RAC: 49
United States
Message 1211329 - Posted: 28 Mar 2012, 23:48:31 UTC

That is not true Jason, I sent the image file to You, If I was angry the text would have been in all caps, it was not, I did not mean or try to be anything I said, considering I am now running x38g, sure I'm a bit frustrated, but not abusive, nor angry. So quit this. And Yes I know who the guy who is making the Lunatics apps(x38g, x41g, x41u) is, It's You Jason a little birdie told Me...

I can always post My PM here if You'd like, right out in the open.
The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's
ID: 1211329 · Report as offensive
Profile arkayn
Volunteer tester
Avatar

Send message
Joined: 14 May 99
Posts: 4438
Credit: 55,006,323
RAC: 0
United States
Message 1211333 - Posted: 28 Mar 2012, 23:54:12 UTC - in response to Message 1211326.  



Some figures. The new 0.40 installer has been downloaded 680 times since we launched it, less than 60 hours ago. The previous version 0.39, released at the end of November 2011, had been downloaded 5,300 times when we started on this one. Older versions, which were available for longer, were downloaded over 10,000 times.


Add another 12 from my site as well, I think I was over 800 there on the previous version.


ID: 1211333 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1211335 - Posted: 28 Mar 2012, 23:58:24 UTC - in response to Message 1211329.  

That is not true Jason, I sent the image file to You, If I was angry the text would have been in all caps, it was not, I did not mean or try to be anything I said, considering I am now running x38g, sure I'm a bit frustrated, but not abusive, nor angry. So quit this. And Yes I know who the guy who is making the Lunatics apps(x38g, x41g, x41u) is, It's You Jason a little birdie told Me...

I can always post My PM here if You'd like, right out in the open.

A little yellow birdie, was it? Yes, I don't think it's ever been an actual secret - just seemed easier to say so explicitly in the message I sent you with the versions to test. No need to institute a leak enquiry over that one.
ID: 1211335 · Report as offensive
Profile zoom3+1=4
Volunteer tester
Avatar

Send message
Joined: 30 Nov 03
Posts: 65709
Credit: 55,293,173
RAC: 49
United States
Message 1211336 - Posted: 29 Mar 2012, 0:02:42 UTC - in response to Message 1211335.  

That is not true Jason, I sent the image file to You, If I was angry the text would have been in all caps, it was not, I did not mean or try to be anything I said, considering I am now running x38g, sure I'm a bit frustrated, but not abusive, nor angry. So quit this. And Yes I know who the guy who is making the Lunatics apps(x38g, x41g, x41u) is, It's You Jason a little birdie told Me...

I can always post My PM here if You'd like, right out in the open.

A little yellow birdie, was it? Yes, I don't think it's ever been an actual secret - just seemed easier to say so explicitly in the message I sent you with the versions to test. No need to institute a leak enquiry over that one.

Sorry I was just trying to keep You out of trouble, as I'd assume that a good number have x38g somewhere, even if their not running it.
The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's
ID: 1211336 · Report as offensive
Previous · 1 . . . 4 · 5 · 6 · 7 · 8 · 9 · 10 . . . 14 · Next

Message boards : Number crunching : Lunatics Windows Installer v0.40 release notes


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.