Progress Nonlinearity

Message boards : Number crunching : Progress Nonlinearity
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Rich Burgan, WC8J

Send message
Joined: 19 May 99
Posts: 5
Credit: 109,718
RAC: 0
United States
Message 14180 - Posted: 6 Aug 2004, 1:05:42 UTC
Last modified: 6 Aug 2004, 1:17:35 UTC

The 'progress' field under the 'work' tab on the BOINC GUI expresses a % complete value. This is very nonlinear. Essentially it goes up to 42% in the first 5 minutes of computation and then rises to 100% at completion.

100...................................+
.80.......................+
.60...........+
.40.+
.20+
..0+
..0 5 -elapsed time minutes- 145
ID: 14180 · Report as offensive
Profile Bruno Moretti IK2WQA
Avatar

Send message
Joined: 15 May 99
Posts: 284
Credit: 49,167
RAC: 0
Italy
Message 14185 - Posted: 6 Aug 2004, 1:24:15 UTC
Last modified: 8 Aug 2004, 2:02:15 UTC

Yes, Rich WC8J.

It's a know secondary bug and not important.

--... ...--


Clear skies from SETI ITALIA [/url]
Cicognola Astronomical Observatory & IK2WQA Radio Station
SETI League IARA GAT IDA CieloBuio IARU ARI
ID: 14185 · Report as offensive
Petit Soleil
Avatar

Send message
Joined: 17 Feb 03
Posts: 1497
Credit: 70,934
RAC: 0
Canada
Message 14189 - Posted: 6 Aug 2004, 1:34:08 UTC

CPU temp shows that nonlinear processing. Temp is slowly
decreasing as the process goes. I have noticed 68 at start
and 62 Deg C at the end. So the CPU "works harder" at let
say 10% then 90% progress.

Dah dah di di dit
di di di da dah dah

Marc
ID: 14189 · Report as offensive
Rich Burgan, WC8J

Send message
Joined: 19 May 99
Posts: 5
Credit: 109,718
RAC: 0
United States
Message 14192 - Posted: 6 Aug 2004, 1:56:06 UTC

For the first 43% the fan on my laptop runs slow then speeds up (max?) for the rest of the time. Is the temperature going down because the fan is running faster?

--... ...--
ID: 14192 · Report as offensive
Petit Soleil
Avatar

Send message
Joined: 17 Feb 03
Posts: 1497
Credit: 70,934
RAC: 0
Canada
Message 14196 - Posted: 6 Aug 2004, 2:40:35 UTC

Actualy no. The fan on my laptop is running at same speed
between 60 - 70 Deg. I have lift up the laptop using home
made stand with cooling fan underneth providing cool air
to the bottom of the laptop. Before that it was reaching
72 and I could ear the fan speeding up at 70.

I can clearly see that the CPU works harder when starting
a new WU. I have watched closely MBM during WU transition.
At 99% cmpletion the temp is 62 and it rise at 68 when it
start crunching the next WU.

As I am writing this a WU is at 99.6% completion and temp
is at 61. It will then jump to 62, 63, 64 up to 68 and then
it will slowly decrease back to 61,62 with the fan running
at same speed. I monitor the fan with MBM and by the sound.

So to me the WU are more power demanding at the beginning.

ID: 14196 · Report as offensive
Rich Burgan, WC8J

Send message
Joined: 19 May 99
Posts: 5
Credit: 109,718
RAC: 0
United States
Message 14200 - Posted: 6 Aug 2004, 3:12:47 UTC

Your experience with CPU temperature is interesting. I will have to watch mine more closely. Maybe I have mis-interpreted what I'm seeing.

Cheers!
ID: 14200 · Report as offensive
Petit Soleil
Avatar

Send message
Joined: 17 Feb 03
Posts: 1497
Credit: 70,934
RAC: 0
Canada
Message 14298 - Posted: 6 Aug 2004, 12:25:51 UTC
Last modified: 6 Aug 2004, 13:46:13 UTC

@Rich

Keep me informed of your dicoveries.

I have made a close watch of a complete WU process and
took temp snapshots. Completion time is 5 hours

0 -10% 62-63-64-65-66-67 deg.C
10-40% 68
40-50% 67-66
50-70% 66-65
70-90% 64-63
90-100% 62

END OF WU STARTING NEW

0 -10% 62-63-64-65-66-67 deg.C etc.

If the CPU work load was the same thruout the process the
temp would not change at the beginning of a new WU and the
temp would stay at 62.

Marc



ID: 14298 · Report as offensive
Profile Bruno Moretti IK2WQA
Avatar

Send message
Joined: 15 May 99
Posts: 284
Credit: 49,167
RAC: 0
Italy
Message 14315 - Posted: 6 Aug 2004, 14:23:50 UTC
Last modified: 8 Aug 2004, 2:03:06 UTC

With BOINC running 24/24 & 7/7
my temp is always 52-54 deg Celsius
on all my machines,
since Dec 2002.

In about 25 deg Celsius air conditioning site. ;-)

--... ...--


Clear skies from SETI ITALIA [/url]
Cicognola Astronomical Observatory & IK2WQA Radio Station
SETI League IARA GAT IDA CieloBuio IARU ARI
ID: 14315 · Report as offensive
Petit Soleil
Avatar

Send message
Joined: 17 Feb 03
Posts: 1497
Credit: 70,934
RAC: 0
Canada
Message 14320 - Posted: 6 Aug 2004, 14:46:25 UTC - in response to Message 14315.  
Last modified: 6 Aug 2004, 14:50:10 UTC

> With BOINC running 24/24 & 7/7
> my temp is always 52-54 deg Celsius
> on all my machines,
> since Dec 2002.
>
> In about 25 deg Celsius air conditionig site. ;-)
>

Thanks Bruno for the info.

It's weird. Maybe specific to my machine ?
some kind of power management ?
Or it could be that my fan decrease speed during the
small time lap of WU transition witch makes the
temp uncreasing at around 68 deg Celsius untill fan
speed increase. I have to check again on that maybe
MBM is not correctly setup and fan reading is wrong
and / or I have earing problems...
I'll have a deeper look.

Thanks
Marc
ID: 14320 · Report as offensive
Chelski
Avatar

Send message
Joined: 3 Jan 00
Posts: 121
Credit: 8,979,050
RAC: 0
Malaysia
Message 14346 - Posted: 6 Aug 2004, 17:43:25 UTC

I see the same thing on both my machines. Don't think it is temp though.

Maybe because the way the progress is calculated i.e. % steps completed are linear but the computation to process the steps may varies from short to long as we crunch?
ID: 14346 · Report as offensive
Profile Christopher Hauber
Avatar

Send message
Joined: 10 Feb 01
Posts: 196
Credit: 71,611
RAC: 0
United States
Message 14366 - Posted: 6 Aug 2004, 19:27:58 UTC - in response to Message 14320.  

Until you said this was on a laptop this didn't make much sense to me. Regardless of where you are in processing the workunit, it processes at 100% usage and unless there is more to it than I realize, that *should* indicates that the processor is using just as much power at as at any other stage which would make produce the same amount of heat. I can't see it making a significant difference in temperature at all. Except that most if not all laptops have some sort of variable speed fan to save power. So when it gets hot, it will turn the fan on as high as it needs to to cool the processor. When it cools it to a certain point, it lowers the fan speed basically until it maintains the temperature within a certain range. That is really the only explanation that I can see explaining it based on my experience with laptops.

Chris


>
> Thanks Bruno for the info.
>
> It's weird. Maybe specific to my machine ?
> some kind of power management ?
> Or it could be that my fan decrease speed during the
> small time lap of WU transition witch makes the
> temp uncreasing at around 68 deg Celsius untill fan
> speed increase. I have to check again on that maybe
> MBM is not correctly setup and fan reading is wrong
> and / or I have earing problems...
> I'll have a deeper look.
>
> Thanks
> Marc
>
>
ID: 14366 · Report as offensive
Petit Soleil
Avatar

Send message
Joined: 17 Feb 03
Posts: 1497
Credit: 70,934
RAC: 0
Canada
Message 14395 - Posted: 6 Aug 2004, 22:57:40 UTC - in response to Message 14366.  

That is really the only
> explanation that I can see explaining it based on my experience with laptops.
>
> Chris
>
>
Thanks Chris

I agree with you. I have tested the fan and it shows same speed
in the 60-68 deg celsius range. I first thaugh it was 60-70 Deg.
Please note that my laptop is always pluged in and it's not affected
by any power saver unless I run on the batery.

I still don't understand why it goes from 62 to 68 when it start
processing a new WU. I have watched the temp in the system tray
during the last WU transition. Actually I can tell when it starts
a new one even when boinc is minimized. Progress was at 99% and temp
was 62 when I minimized boinc and kept an eye on sys tray temp reading.
It started to rise slowly from 62 to 68 and I then knew it was starting
to crunch a new WU.

Grrrrrrr!!! I should have think of that before !
There could be one good explanation. The interval between MBM
reading is about 5 seconds. If the transition between WU keeps
the processor idle long enough the temp might go under 60 but I
don't see it. the fan speed would decrease and for some reason
the sensor system don't kicks in at 60 but at 68. This could be
a good explanation only if there is a processing "pause" between
WU transition. I have to check that. The only way to be 100% sure
would be to set the fan at a fixed RPM and check the temp during
transition. I'm not sure if I can do it with my machine.
I will check the bios.

Marc




ID: 14395 · Report as offensive
Pete49

Send message
Joined: 28 Jul 04
Posts: 64
Credit: 250,376
RAC: 0
United States
Message 14416 - Posted: 6 Aug 2004, 23:37:48 UTC

Another possible explanation could be the algorithm the SETI program is using.
Even though a processor is being externally utilized 100%, not all of the internal registers, caches, etc are being used. As floating point calculations increase in complexity, the internal usage of the processor increases which would result in an increase in temp.


ID: 14416 · Report as offensive
ric
Volunteer tester
Avatar

Send message
Joined: 16 Jun 03
Posts: 482
Credit: 666,047
RAC: 0
Switzerland
Message 14419 - Posted: 6 Aug 2004, 23:46:32 UTC - in response to Message 14395.  

When a cpu is getting hot while processing (a wu), after stopping the high usage, it will take much more time as 5 seconds to go realy down the temp.

please also include this: many motherboard offers different location, where the temp sensor is placed, some are on the MB, just near the cpu, some more away.

Your monitoring software, does it allow to select what sensor is used for the mesurement? Don't ask me in details, but only newer CPUs (amd and intel)
have a sensor build in AND can deliver the temp status. This can produce a other value of temp than effectively is.

and.. some time ago, an "important" pc-magazine made some laboratory test concerning temp mesurement, they mesured differences till 8% from the effective values of temp. Same in the u min of the air cooling units

wishing cool cpus

ric
ID: 14419 · Report as offensive
Petit Soleil
Avatar

Send message
Joined: 17 Feb 03
Posts: 1497
Credit: 70,934
RAC: 0
Canada
Message 14438 - Posted: 7 Aug 2004, 2:31:25 UTC - in response to Message 14416.  
Last modified: 7 Aug 2004, 2:34:07 UTC

@Pete49
That was my feeling too when I wrote "more power demanding" at the begining
of a WU. I have the feeling that the calculations are more intensive in the first
40% of the WU.

@RichiMeister
I know my CPU would not cool completely in 5 seconds but it might be enough
to cool from 62 to 60. the fan speed is decreasing at 60 so if there is a long enough
time lap between two WU that could have been an expanation. My machine is a 2 years
old Compaq Presario 2800 P4M 1.4 GHz Its got one sensor on the cooling block and
another on the CPU. I monitor both with MBM.

----

It could well be both reasons. Algorith and fan speed. One thing for shure the temp goes
from 62 to 68 in about 4 minutes when starting a new WU.

Thanks to all. My quest for the reasons continue

Best regards
Marc
ID: 14438 · Report as offensive
Petit Soleil
Avatar

Send message
Joined: 17 Feb 03
Posts: 1497
Credit: 70,934
RAC: 0
Canada
Message 14574 - Posted: 7 Aug 2004, 16:11:24 UTC
Last modified: 7 Aug 2004, 16:33:57 UTC

OK I am now 100% sure that those non linear WU has a higher
usage of the processor ressources at the begining of the process.

The first pic is a screen capture showing a boinc SETI WU at
99% progress. Note the MBM temperature in the system tray.
The temperature reading from left to right are BLOCK and CPU.
47 and 62 deg celsius. The second pic shows a second screen
capture taken 9 minutes later. The new WU progress is 40% and
the temperatures are 51 and 69

I have just finished crunching a normal (linear progress %) work
unit and the temperature stayed at 65 deg. Celsius during complete
process. There is no doubt in my mind that the temperatures reading
reflects that those nonlinear progress WU are also non constant as
for Processor usage.


ID: 14574 · Report as offensive
ric
Volunteer tester
Avatar

Send message
Joined: 16 Jun 03
Posts: 482
Credit: 666,047
RAC: 0
Switzerland
Message 15374 - Posted: 11 Aug 2004, 8:22:39 UTC

ping "threat still alive?"
ID: 15374 · Report as offensive
Petit Soleil
Avatar

Send message
Joined: 17 Feb 03
Posts: 1497
Credit: 70,934
RAC: 0
Canada
Message 15383 - Posted: 11 Aug 2004, 8:39:22 UTC - in response to Message 15374.  

> ping "threat still alive?"
>
Which threat ?



-.-. --.- -.. -..- . - .-.-. -.- --... ...--
ID: 15383 · Report as offensive
ric
Volunteer tester
Avatar

Send message
Joined: 16 Jun 03
Posts: 482
Credit: 666,047
RAC: 0
Switzerland
Message 15387 - Posted: 11 Aug 2004, 8:47:18 UTC - in response to Message 15383.  

> > ping "threat still alive?"
> >
> Which threat ?
>
>
>
> -.-. --.- -.. -..- . - .-.-. -.- --... ...--

sorry, not threat, 2nd chance, thread (still learning english)..
but it's to admit, I was able to copy 5 of 6 chars from the adress line,
it's not to bad fo me.
ID: 15387 · Report as offensive
Profile Masher1

Send message
Joined: 24 Jun 01
Posts: 1
Credit: 454
RAC: 0
Canada
Message 15388 - Posted: 11 Aug 2004, 8:47:51 UTC

the W/U feed is not allways a stable linear thing sometimes the tele-scope is moving excessively {hence the 2 sec wu) or speeding up or slowing down {the non-linear wu} the data somtimes is less than optmial for the search these thing can all have the effect you see
ID: 15388 · Report as offensive
1 · 2 · Next

Message boards : Number crunching : Progress Nonlinearity


 
©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.