Advanced search

Message boards : Graphics cards (GPUs) : BOINC runs 2.2% of time

Author Message
Alain Maes
Send message
Joined: 8 Sep 08
Posts: 63
Credit: 1,437,484,959
RAC: 69,868
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 4804 - Posted: 23 Dec 2008 | 16:40:23 UTC
Last modified: 23 Dec 2008 | 16:41:15 UTC

Looks like a new problem

Linux 64, 6.5.0, 177 driver, hostID16551 (Q6600, 4 GB, GTX260)

Was OK for a while with up to 4 WUs, but now I cannot get any new ones.

Server believes I wil not finish in time: BOINC runs 2.2% of time, computation enabled 100% of that.

On my 24/7 dedicated cruncher BOINC actually runs all the time. So looks to me as if the 2.2% refers to the time used for GPUGRID only (4 other WU - ABC, PG and Cosmology - use the rest).

Will try some more manual updates before my last WU finishes.

kind regards

Alain

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 4805 - Posted: 23 Dec 2008 | 16:47:24 UTC - in response to Message 4804.

Have you tried setting the BOINC resource share for GPUGRID higher?

Alain Maes
Send message
Joined: 8 Sep 08
Posts: 63
Credit: 1,437,484,959
RAC: 69,868
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 4806 - Posted: 23 Dec 2008 | 17:10:25 UTC - in response to Message 4805.

Many thanks, had not thought of that.

Raised from 100 to 500% (too bad for the already out of line long term debt) and got 2 WU with manual update (x2).

Thanks again

Alain

Alain Maes
Send message
Joined: 8 Sep 08
Posts: 63
Credit: 1,437,484,959
RAC: 69,868
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 4822 - Posted: 24 Dec 2008 | 9:49:30 UTC - in response to Message 4804.

Think I figured out what is happening.

This morning my on fraction had grown to 8.8%, so I still had to do a manual update to report 2 WU finished and to get a new one. This succeeded.

A bit later I tried to get another new WU to notice that on fraction was now less than before, even close to zero.

What happened? Being on leave I am playing around with my machines, including the Linux one (hostID 16551). So somewhat unusual I restarted this one a couple of times. And checking on the client_state.xml I noticed that with every restart of the BOINC client the ON_frac value is reset to zero!

I believe this is an unwanted feature of 6.5.0. On previous versions it was an average over time, which makes sense.

Can anyone please confirm this?

Many thanks.

Alain

Profile Lazarus-uk
Send message
Joined: 16 Nov 08
Posts: 29
Credit: 122,821,515
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 4830 - Posted: 24 Dec 2008 | 18:48:09 UTC - in response to Message 4822.
Last modified: 24 Dec 2008 | 18:49:51 UTC

Not sure if this related to the aforementioned problem or not. But, it seems to be similar, so here goes... Since installing Boinc 6.5.0 I noticed that I am having trouble getting work from any project, I have the work cache set to 0.6 days. Currently running 1 'USPME type' GPU and 4 prime grid which take about 12 mins each. BOINC manager is only allowing me 1 or 2 extra PG tasks as cache when there should be about 100 or more.

Also noticed these messages when starting Boinc manager

Wed 24 Dec 2008 18:15:00 GMT||Starting BOINC client version 6.5.0 for x86_64-pc-linux-gnu
Wed 24 Dec 2008 18:15:00 GMT||[error] bad value -1.000000 of time stats connected_frac; ignoring
Wed 24 Dec 2008 18:15:00 GMT||[error] bad value -1.000000 of time stats active_frac; ignoring

Wed 24 Dec 2008 18:15:00 GMT||Processor: 4 GenuineIntel Intel(R) Core(TM)2 Quad CPU Q9450 @ 2.66GHz [Family 6 Model 23 Stepping 7]
Wed 24 Dec 2008 18:15:00 GMT||Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall lm constant_tsc arch_perfmon pebs bts rep_good nopl pni monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr sse4
Wed 24 Dec 2008 18:15:00 GMT||OS: Linux: 2.6.27-9-generic
Wed 24 Dec 2008 18:15:00 GMT||Memory: 3.87 GB physical, 953.66 MB virtual
Wed 24 Dec 2008 18:15:00 GMT||Disk: 26.58 GB total, 21.83 GB free


Also noted that my 'on' time was reading 0.0145....... after it had been on for 5 hours solid.

On another note... I noticed that the GPU task was waiting for memory, so I increased its allowance to 90% and memory usage climbed to 2.3GB. I have since rebooted and it is currently @ 380MB and climbing. Apps are left suspended in memory.


Hope some of this can help.


Mark

Profile mike047
Send message
Joined: 21 Dec 08
Posts: 47
Credit: 7,330,049
RAC: 0
Level
Ser
Scientific publications
watwatwatwatwatwatwat
Message 4832 - Posted: 24 Dec 2008 | 20:29:56 UTC - in response to Message 4830.

If you will run the boinc benchmark, it will give you the memory back..you won't have to reboot. This is on 6.4.2.

mike

Profile Lazarus-uk
Send message
Joined: 16 Nov 08
Posts: 29
Credit: 122,821,515
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 4833 - Posted: 24 Dec 2008 | 21:06:33 UTC - in response to Message 4832.


If you will run the boinc benchmark, it will give you the memory back..you won't have to reboot. This is on 6.4.2.

mike



Thanks for that Mike. I'll remember that next time I use Linux. I'm currently back running GPU on Windoze with the Boinc 6.5.0 release and notice that it does not suffer the same problems as Linux does. Seems much more stable on Windoze and appears to use much less CPU.

Mark



Alain Maes
Send message
Joined: 8 Sep 08
Posts: 63
Credit: 1,437,484,959
RAC: 69,868
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 4842 - Posted: 25 Dec 2008 | 10:29:40 UTC - in response to Message 4822.

Checked again this morning. on_frac reset only happens after restart on 6.5.0 Linux 64 bit.

After downgrade to 6.4.5 on_frac stays as it was and continues to increase.

My VISTA 32, unfortunately without a suitable GPU, is also OK even on 6.5.0.

So staying with 6.4.5 for now since my machines will have to be shutdown for the night till my visitors are gone (they sleep in my PC room).

Merry Christmas to all of you.

Alain

Alain Maes
Send message
Joined: 8 Sep 08
Posts: 63
Credit: 1,437,484,959
RAC: 69,868
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 4852 - Posted: 25 Dec 2008 | 15:25:56 UTC - in response to Message 4842.

Checked again this morning. on_frac reset only happens after restart on 6.5.0 Linux 64 bit.

After downgrade to 6.4.5 on_frac stays as it was and continues to increase.

My VISTA 32, unfortunately without a suitable GPU, is also OK even on 6.5.0.

So staying with 6.4.5 for now since my machines will have to be shutdown for the night till my visitors are gone (they sleep in my PC room).

Merry Christmas to all of you.

Alain


Confirmed by Dagorath (thank you) who put it also on the boinc-alpha list.
See BOINC forum

Kind regards

Alain

Post to thread

Message boards : Graphics cards (GPUs) : BOINC runs 2.2% of time

//