Fixed: BOINC connection mode: When tasks get suspended and the thread count goes to 0, detect Suspended and don't add the tasks to the throttle list.
Fixed: As of V 3.2, One tasks isn't assigned to any throttling regulator at all and is running at 100%.
When you are running a Beta version: I recommend you update to this version A.S.A.P., because there is a serious bug in the older Beta versions.
1 CPU + 1 GPU tasks are running. After snoozing the whole computation, the machine goes correctly silent. Just the logging:
----------------------------------------------------------------------------------------------------------------------------------------------
30 March 2011 - 00:43:56 Počet zhodných programov (procesov): 2
CPU:0, GPU:1, PID:8408 (3) Slot:4 http://boinc.thesonntags.com/collatz/ collatz_2368747883356106762600_824633720832
CPU:1, GPU:0, PID:12956 (3) Slot:1 http://boinc.thesonntags.com/collatz/ collatz_2368747395894498535784_103079215104
----------------------------------------------------------------------------------------------------------------------------------------------
30 March 2011 - 01:43:21 Počet zhodných programov (procesov): 1
Suspended: CPU:0, GPU:1, PID:8408 (0) Slot:4 http://boinc.thesonntags.com/collatz/ collatz_2368747883356106762600_824633720832
CPU:1, GPU:0, PID:12956 (3) Slot:1 http://boinc.thesonntags.com/collatz/ collatz_2368747395894498535784_103079215104
----------------------------------------------------------------------------------------------------------------------------------------------
If the GPU task is marked as Suspended, I'd expect to see the same on CPU task(s)...
Otherwise - everything seems just fine.
Quote from: Pepo on March 29, 2011, 11:52:35 PM
1 CPU + 1 GPU tasks are running. After snoozing the whole computation, the machine goes correctly silent. Just the logging:
----------------------------------------------------------------------------------------------------------------------------------------------
30 March 2011 - 00:43:56 Počet zhodných programov (procesov): 2
CPU:0, GPU:1, PID:8408 (3) Slot:4 http://boinc.thesonntags.com/collatz/ collatz_2368747883356106762600_824633720832
CPU:1, GPU:0, PID:12956 (3) Slot:1 http://boinc.thesonntags.com/collatz/ collatz_2368747395894498535784_103079215104
----------------------------------------------------------------------------------------------------------------------------------------------
30 March 2011 - 01:43:21 Počet zhodných programov (procesov): 1
Suspended: CPU:0, GPU:1, PID:8408 (0) Slot:4 http://boinc.thesonntags.com/collatz/ collatz_2368747883356106762600_824633720832
CPU:1, GPU:0, PID:12956 (3) Slot:1 http://boinc.thesonntags.com/collatz/ collatz_2368747395894498535784_103079215104
----------------------------------------------------------------------------------------------------------------------------------------------
If the GPU task is marked as Suspended, I'd expect to see the same on CPU task(s)...
Otherwise - everything seems just fine.
For some reasons, the GPU tasks are still handled differently.
The GPU task is gone, as there are 0 threads running, but the CPU task is still there, 3 threads.
Quote from: fred on March 30, 2011, 12:35:09 AM
Quote from: Pepo on March 29, 2011, 11:52:35 PM
If the GPU task is marked as Suspended, I'd expect to see the same on CPU task(s)...
For some reasons, the GPU tasks are still handled differently.
The GPU task is gone, as there are 0 threads running, but the CPU task is still there, 3 threads.
Indeed.
When I snooze just GPU computing, then TTh will completely discard GPU tasks from the list, CPU tasks are not differentiated from a running state.
When I snooze all computing, then a the-same-way not-anymore-existing GPU task (0 processes 0 threads) is just marked as Suspended, whereas suspended CPU tasks still look the same. Why are the GPU tasks still on-list?
Maybe if TTh is aware of which existing processes belong to suspended tasks (while being connected to BOINC client), then it could mark them as such?
If TTh is connected to the client, would it notice any runaway tasks' processes, which is the client not (anymore) aware of? I believe that in the standalone mode the tasks are being noticed. (Maybe this could be noticed e.g. by killing the client, if TTh would react sooner than after the tasks' 30 second timeout? Or running some task in stand-alone mode?)
Quote from: Pepo on March 30, 2011, 03:17:26 PM
Indeed.
When I snooze just GPU computing, then TTh will completely discard GPU tasks from the list, CPU tasks are not differentiated from a running state.
When I snooze all computing, then a the-same-way not-anymore-existing GPU task (0 processes 0 threads) is just marked as Suspended, whereas suspended CPU tasks still look the same. Why are the GPU tasks still on-list?
Maybe if TTh is aware of which existing processes belong to suspended tasks (while being connected to BOINC client), then it could mark them as such?
If TTh is connected to the client, would it notice any runaway tasks' processes, which is the client not (anymore) aware of? I believe that in the standalone mode the tasks are being noticed. (Maybe this could be noticed e.g. by killing the client, if TTh would react sooner than after the tasks' 30 second timeout? Or running some task in stand-alone mode?)
Processes are marked suspended when there are 0 threads. When there are thread, something is still going on. But as BOINC has several snooze states it's sometimes hard to predict what will happen. But it works and that's what counts.
Only tasks that the clients sees, that are running in BoincTasks are throttled.
So runaway tasks are not throttled.
Disabling the BOINC client mode is the only option in this case. Or setting up a rule at a higher threshold. Or setting an alarm.
Have v3.41 on 11 machines and it's running very well. Seems to use quite a bit lees CPU than previous versions. Thanks!
I'd love to try out v3.41 except I can't seem to find it on the site? All I get is the message "At this moment there is no beta version. " when I click on the "Download beta version" link on the Download page... ???
Quote from: jave200372 on April 01, 2011, 10:33:55 AM
I'd love to try out v3.41 except I can't seem to find it on the site? All I get is the message "At this moment there is no beta version. " when I click on the "Download beta version" link on the Download page... ???
No idea ??? Maybe I'm allowed to access the page and you are not? ;) :P
Or you see some other cached version of the Beta download page?
Anyway,Quote from: http://www.efmer.eu/boinc/boinc_tasks/download.htmlBeta test version
For testing only. Download BoincTasks beta 32/64 Bit (http://www.efmer.eu/boinc/boinc_tasks/download_beta.html)
points toQuote from: http://www.efmer.eu/boinc/boinc_tasks/download_beta.htmleFMer BoincTasks Beta Version
[.....]
BoincTasks 1.00 Beta:
Download site 1: Europe (//http://)
Download site 2: North America (//http://)
Go try again... :-X
Peter, he's talking about TThrottle, not BoincTasks. It's early :D
There doesn't seem to be a link to 3.41 anymore. Don't know why?
Edit: jave200372, as Peter says below, it's still on the server. Use this link:
http://www.efmer.eu//download/boinc/tthrottle/unified/setup_32_64_tthrottle_3_4_1.exe
Quote from: Beyond on April 01, 2011, 12:31:15 PM
Peter, he's talking about TThrottle, not BoincTasks. It's early :D
:-[ I can not blame even the daytime (it was already pretty afternoon) - but I'm somehow tired and inattentive today... True, a very different application...
QuoteThere doesn't seem to be a link to 3.41 anymore. Don't know why?
jave200372, send me a PM with your e-mail address & I'll send the 3.41 file to you.
I haven't noticed it for some time already - as there are nowhere links to download areas, I'm always taking the last download link (e.g. look for "TThrottle" on my chrome://downloads (http://chrome://downloads) page (
??? does not work correctly as a direct link?), take the most recent TTh's link) and just overwrite the digits, like "
.../tthrottle_3_4_1.exe". Fortunately Fred does keep the whole history.
:-[ Forgot to update the download page. Fixed.
The update function works.
You may need to refresh the download page, normally with F5.
Quote from: fred on April 01, 2011, 11:06:20 PM
:-[ Forgot to update the download page. Fixed.
The update function works.
You may need to refresh the download page, normally with F5.
I have downloaded v3.41 from the main download link. Thanks! :)
Quote from: fred on March 30, 2011, 12:35:09 AM
Quote from: Pepo on March 29, 2011, 11:52:35 PM
1 CPU + 1 GPU tasks are running. After snoozing the whole computation, the machine goes correctly silent. Just the logging:
----------------------------------------------------------------------------------------------------------------------------------------------
30 March 2011 - 01:43:21 Počet zhodných programov (procesov): 1
Suspended: CPU:0, GPU:1, PID:8408 (0) Slot:4 http://boinc.thesonntags.com/collatz/ collatz_2368747883356106762600_824633720832
CPU:1, GPU:0, PID:12956 (3) Slot:1 http://boinc.thesonntags.com/collatz/ collatz_2368747395894498535784_103079215104
----------------------------------------------------------------------------------------------------------------------------------------------
For some reasons, the GPU tasks are still handled differently.
The GPU task is gone, as there are 0 threads running, but the CPU task is still there, 3 threads.
I believe what you are seeing is the difference between GPU and CPU tasks suspending when it comes to the
"leave applications in memory when suspended" parameter. GPU tasks are never left in memory but are always fully shutdown. (To free up any GPU memory and other resources to, for example, a resource hungry game.) CPU tasks can be left in memory (if enabled by the parameter), just waiting to be re-enabled to get going again. This happens no matter what the suspend reason of a task is.
I just wanted to update one machine's 3.41 -> 3.51 (is running since 28.3., main dialog window mostly unattended but closed, the temperatures graph window constantly open) and noticed that the log window has lost its scrollbars - never seen before. The machine is not "overwhelmed with tasks", just a couple of them are usually around:
Quote11 April 2011 - 14:41:59 Počet zhodných programov (procesov): 4
CPU:1, GPU:0, PID:18096 (3) Slot:16 http://climateprediction.net/ famous_xtzq_1199_200_007100701
CPU:1, GPU:0, PID:15860 (4) Child: famous_um_6.11_windows_intelx86.exe
CPU:1, GPU:0, PID:15000 (3) Slot:13 http://cpdnbeta.oerc.ox.ac.uk/ famous_v0ia_1799_200_000223519
CPU:1, GPU:0, PID:17524 (4) Child: famous_um_6.11_windows_intelx86.exe
CPU:1, GPU:0, PID:11992 (3) Slot:2 http://www.primegrid.com/ llrCUL_83826417
CPU:1, GPU:0, PID:7452 (1) Child: primegrid_cllr.exe
CPU:0, GPU:1, PID:2736 (3) Slot:6 http://einstein.phys.uwm.edu/ PM0070_01511.dm_348
----------------------------------------------------------------------------------------------------------------------------------------------
11 April 2011 - 15:45:29 Počet zhodných programov (procesov): 4
CPU:1, GPU:0, PID:18096 (3) Slot:16 http://climateprediction.net/ famous_xtzq_1199_200_007100701
CPU:1, GPU:0, PID:15860 (4) Child: famous_um_6.11_windows_intelx86.exe
CPU:1, GPU:0, PID:11992 (3) Slot:2 http://www.primegrid.com/ llrCUL_83826417
CPU:1, GPU:0, PID:7452 (1) Child: primegrid_cllr.exe
CPU:0, GPU:1, PID:2736 (3) Slot:6 http://einstein.phys.uwm.edu/ PM0070_01511.dm_348
CPU:1, GPU:0, PID:21568 (3) Slot:9 http://setiweb.ssl.berkeley.edu/beta/ 07mr11ah.23742.16018.206158430211.14.147
----------------------------------------------------------------------------------------------------------------------------------------------
11 April 2011 - 16:13:06 Počet zhodných programov (procesov): 4
CPU:1, GPU:0, PID:15000 (3) Slot:13 http://cpdnbeta.oerc.ox.ac.uk/ famous_v0ia_1799_200_000223519
CPU:1, GPU:0, PID:17524 (4) Child: famous_um_6.11_windows_intelx86.exe
CPU:1, GPU:0, PID:11992 (3) Slot:2 http://www.primegrid.com/ llrCUL_83826417
CPU:1, GPU:0, PID:7452 (1) Child: primegrid_cllr.exe
CPU:0, GPU:1, PID:2736 (3) Slot:6 http://einstein.phys.uwm.edu/ PM0070_01511.dm_348
CPU:1, GPU:0, PID:21568 (3) Slot:9 http://setiweb.ssl.berkeley.edu/beta/ 07mr11ah.23742.16018.206158430211.14.147
----------------------------------------------------------------------------------------------------------------------------------------------
11 April 2011 - 16:16:09 Počet zhodných programov (procesov): 2
CPU:0, GPU:1, PID:2736 (3) Slot:6 http://einstein.phys.uwm.edu/ PM0070_01511.dm_348
CPU:3, GPU:0, PID:1404 (5) Slot:11 http://aqua.dwavesys.com/ fp_06apr11_bm_16_003_50_000-1_358
----------------------------------------------------------------------------------------------------------------------------------------------
11 April 2011 - 16:34:34 Počet zhodných programov (procesov): 2
CPU:0, GPU:1, PID:22456 (3) Slot:10 http://www.primegrid.com/ pps_sr2sieve_15793585
CPU:3, GPU:0, PID:1404 (5) Slot:11 http://aqua.dwavesys.com/ fp_06apr11_bm_16_003_50_000-1_358
----------------------------------------------------------------------------------------------------------------------------------------------
11 April 2011 - 17:03:25 Počet zhodných programov (procesov): 4
CPU:1, GPU:0, PID:15000 (3) Slot:13 http://cpdnbeta.oerc.ox.ac.uk/ famous_v0ia_1799_200_000223519
CPU:1, GPU:0, PID:17524 (4) Child: famous_um_6.11_windows_intelx86.exe
CPU:1, GPU:0, PID:20104 (3) Slot:5 http://einstein.phys.uwm.edu/ PM0079_01751.dm_360
CPU:0, GPU:1, PID:22456 (3) Slot:10 http://www.primegrid.com/ pps_sr2sieve_15793585
CPU:1, GPU:0, PID:15732 (3) Slot:4 http://www.worldcommunitygrid.org/ X0000074150700200609201430
----------------------------------------------------------------------------------------------------------------------------------------------
It is also not possible to jump up/down with PageUp/PageDown keys - there is simply nothing more to scroll through, just (14:41:59 - 16:34:34).
A 20 minutes later (17:03:25) a new entry appeared and the whole displayed contents became scrollable. But nothing visible from the history. I remember having seem much longer histories in TTh.
Hey Fred,
Well I am simply amazed! And that is rarely said from my lips!
Out of the box (once I guessed at the settings) it worked fantastically! Ha ha!
Maybe an inline usage help - or a link to a website would help newbies like me!
Needs a display throttle? I don't need real-time updates. A 5 second refresh is good enough for me.
Now I have to figure out how to make a PayPal account to send you a few bucks!
Thanks for an excellent tool...and my hot baby thanks you as well...!!!
Smilin' KraftR
Here is my log data if it helps... Thanks again!
13 June 2011 - 10:45:32 Driver installed properly. Driver Version: 2.0
Program version: 3.41 64Bit
Microsoft Service Pack 1 (build 7601)
Language: User: 1033 ENU ,System: 1033 ENU
Amd/Ati: found 1 devices
Amd/Ati: ATI Mobility Radeon HD 5870 P (0-0)
13 June 2011 - 10:45:32 AdapterInfo: 0-0, Index: 0, ATI Mobility Radeon HD 5870
Pos: 0, Temperature: 70
Vendor ID: GenuineIntel
Vendor: INTEL
HighestIntegerValue: 0000000B - Processor Signature: 000106E5
Misc. info: 02100800
Feature Flags1 0098E3FD
Feature Flags2 BFEBFBFF
Processor: Intel(R) Core(TM) i7 CPU Q 740 @ 1.73GHz
Processor: Family: 6h, Model: 1E, Stepping: 5
Core Temperature: 61 °C, Raw Data: 88270000
88280000,88280000,882a0000,882a0000,88290000,88290000,88270000,88270000,
88270000,88270000,88270000,88270000,88270000,88270000,88270000,88270000,
0123456777777777
Quote from: KraftR on June 13, 2011, 04:06:17 PM
Hey Fred,
Well I am simply amazed! And that is rarely said from my lips!
Out of the box (once I guessed at the settings) it worked fantastically! Ha ha!
Maybe an inline usage help - or a link to a website would help newbies like me!
Needs a display throttle? I don't need real-time updates. A 5 second refresh is good enough for me.
Now I have to figure out how to make a PayPal account to send you a few bucks!
Thanks for an excellent tool...and my hot baby thanks you as well...!!!
Smilin' KraftR
The manual with all the links is installed and give access to this forum.
And for a donation you don't need any PayPal account, unless you don't have a Credit Card. You can pay directly by Credit Card.
Hey Fred,
The help info must be hidden somewhere else?
Or I missed it when it launched? Nothing inside the tool.
And (BTW) if you hit F1 you get a "no help here" message.
I am happy with how it is running - just wanna know what I am looking at?
Smilin'
KraftR
Quote from: KraftR on June 13, 2011, 04:25:40 PM
Hey Fred,
The help info must be hidden somewhere else?
Or I missed it when it launched? Nothing inside the tool.
And (BTW) if you hit F1 you get a "no help here" message.
I am happy with how it is running - just wanna know what I am looking at?
Smilin'
KraftR
All programs -> eFMer TThrottle -> Manual
About -> Internet address at the bottom.
But I will catch the F1 button in the next 4.3 version, to redirect it to the manual.
Bingo! I didn't even bother to look there for help... Ha!
BTW, the graph display interval is not preserved after a reboot.
It goes back to the default setting - just an oversight...
Think about a polling interval (in milliseconds) for reading the temperature sensors.
Not sure what your polling interval is set to now - but it seems almost real-time...?
Sure it makes for nice smooth operations and more even graphs...
...but doesn't a short polling interval itself add some overhead?
I use another tool to monitor the temperature sensors.
And it seems to work just fine at 3-5K milliseconds.
Thanks, KraftR
Quote from: KraftR on June 13, 2011, 05:42:53 PM
Think about a polling interval (in milliseconds) for reading the temperature sensors.
Indeed they are, but the overhead is very low. I hate wasted CPU time.
But TThrottle is designed as a temperature regulator, so it wants up to date info.
Quote from: fred on June 14, 2011, 05:09:41 AM
Quote from: KraftR on June 13, 2011, 05:42:53 PM
Think about a polling interval (in milliseconds) for reading the temperature sensors.
Indeed they are, but the overhead is very low. I hate wasted CPU time.
But TThrottle is designed as a temperature regulator, so it wants up to date info.
TThrottle CPU usage is very low on all my machines. I agree that real time or very short interval polling is important for this app.
Ok Folks...been running very clean since I started running a few days ago!
Here are a few of my observations...
I don't mind the real-time polling of the temperature sensor.
Just don't let me see it displaying that fast in the utility... Ha Ha!
Fake it out for the user so they won't think the same thing...
At least bring the display updates down to once a second?
Look at that Temperature window jumpin' all around! Smilin'
I have had this line in the first slot of each of my logging blocks since I last rebooted.
What is "slot 18" all about? I only have four cores (8 CPUs).
CPU:1, GPU:0, PID:7044 (3) Slot:18 http://setiathome.berkeley.edu/ ap_06ap11af_B1_P1_00612_20110604_11517.wu
And another observation...choppy CPU utilization working with Boinc Manager.
I set the Boinc Manager to a MAX of 90% CPU usage before I went to bed last night.
I noticed a periodic oscillation in the (8) CPU graphic monitor I have running in a Gadget...
The overall CPU oscillation seemed to be on a 60 second interval as per Boinc's checkpoint value?
If I set Boinc back to 100% CPU utilization then this oscillation smooths out to nothing...
And another observation...it seems CPU (1 & 2) are being picked on by somebody?
The heat levels are consistent across all (4) Cores. But CPU (1 & 2) are being throttled hard? maybe?
There is a SETI job assigned to each of the (8) CPUs - so there is always work running.
But CPU (1 & 2) are only 25% utilized - even though the others are running at 75% or more.
Tthrotle has done its job well...the maximum Core temp has stayed at my 88C setting...
Now...if I suspend Boinc...watching all the CPUs fall back to zero...and go have a cig now...
All of the CPU temps drop down to 65C -> which is my normal base temp for this ASUS laptop.
(yes I know -> don't debate about the heat sinks again...this baby likes to run hot. Ha ha!)
I then un-suspend Boinc -> bang all CPUs scream back to 100% - reaching 88C in 20 seconds.
Then Tthrolle kicks in doing its job - but CPU (1 & 2) gradually drops to 25% over a few minutes?
The other CPUs get throttled back to 75-80% utilization and all temps smooth out nicely...
I am going to reboot to see if this abnormal behavior for CPU (1 & 2) goes away or not...???
Note also I have Boinc set to suspend if I use more then 15% of the CPU for other tasks.
I do have about 50 websites open in seven other browser sessions -> idling away at 5% usage.
Boinc kicks in and suspends everything nicely when I start doing work in the various windows...
I am not ready to reboot quite yet - I still have a lot of work pending in those 50 windows.
So I will shutdown/restart Boinc and those 8 SETI tasks -> to reset their CPU affinity now.
No change after restarting the SETI tasks... I am shutting down the 50 websites now...
That seems to have fixed it! Strange? maybe two of the browsers were idling too fast?
Which brings me to another question?
When (or if) does Tthrottle - throttle NON-Boinc processes?
Thanks KraftR
Quote from: KraftR on June 15, 2011, 01:18:23 PM
Ok Folks...been running very clean since I started running a few days ago!
Here are a few of my observations...
1) The slots are folders that hold the WU and the programs that belong to it.
2) BOINC itself uses a very crude on off regulator that switches everything off for a couple of seconds and lets everything running again, this can cause some large spikes, certainly on a I7 notebook.
3) Work is assigned to a certain core by Windows, not much you can do about. Some programs, like the browser can take up a certain core.
You may want to try out V 4.2, it runs a lot better on my own I7 notebook. It's still beta, but I think there are no bugs with 4.20.
What you can try is another browser, some browsers use less CPU time than others. I used Firefox, but Chrome uses a bit less, not a problem on a regular PC, but notable on a laptop.
And there are a lot of other programs running, more than you like.