The above is taking 49-55% CPU - is this correct?

The above is taking 49-55% CPU - is this correct?

I have seen the same thing for both of my models too. Currently windows taskmanager has four CPDN processes listed:

hadsm3_um_5.06_windows_intelx86.exe (20-25%)
hadsm3_5.06_windows_intelx86.exe (20-25%)
hadcm3transum_5.44_windows_intelx86.exe (20-25%)
hadcm3_5.44_windows_intelx86.exe (20-25%)

My system is a Dual Core Intel @2.16Ghz with two CPDN models running.

Anyone understand what\'s going on???

The above is taking 49-55% CPU - is this correct?

Peter,

Welcome to the message boards.

Your machine looks to be the same as my home machine, which has a 3 GHz Pentium 4. These machines have a \'hyperthreading\' capability, which improves throughput by up to 20%, but which appears to Windows as if there are two processors. So a process running at full speed will be reported by Windows as consuming only 50% of the CPU.

Iain

I have seen the same thing for both of my models too. Currently windows taskmanager has four CPDN processes listed:

hadsm3_um_5.06_windows_intelx86.exe (20-25%)
hadsm3_5.06_windows_intelx86.exe (20-25%)
hadcm3transum_5.44_windows_intelx86.exe (20-25%)
hadcm3_5.44_windows_intelx86.exe (20-25%)

My system is a Dual Core Intel @2.16Ghz with two CPDN models running.

Anyone understand what\'s going on???

Tim,

Normally, only the \'hadcm3transum\' process is consuming CPU. The other process consumes CPU if the graphics are running (e.g. rotating view). If it\'s consuming CPU without the graphics being displayed then that\'s probably because the graphics are being displayed in another session to the logged on session (this happens to me on a server I manage remotely). Try switching users to look at any other live sessions, or it may be related to the service install.

Iain

Tim wrote:

hadsm3_um_5.06_windows_intelx86.exe (20-25%)
hadsm3_5.06_windows_intelx86.exe (20-25%)
hadcm3transum_5.44_windows_intelx86.exe (20-25%)
hadcm3_5.44_windows_intelx86.exe (20-25%)

I sometimes get the same thing after the graphics have been displayed. I think what happens is that the thread which displays the graphics is not sent the message to stop showing the graphics, and keeps generating screenshots (which are unused). I simply display the graphics very infrequently, and never use the screensaver.

Tim, which firewall do you use?


Peter, could I suggest you upgrade to Boinc 5.10.30? It contains a fix for the 0xC00000142 error which sometimes happens as the PC is shutting down, or just about to crash for some reason. (This is unrelated to the 50% issue, see Iain\'s post for more details).

---

I'm a volunteer and my views are my own.
News and Announcements and FAQ

Thank you both Ian and Mike.

I ended up reverting the service settings back to the original way it was setup without graphics. I presume until I can find another method of finding the timestep without graphics, I\'ll just have to wait for the trickles.

Overall, it seemed to be exactly as you guys had suggested that the command to shutdown the graphics was not reaching the program because even an attempt to shutdown the process via taskmanager resulted in an immediate restart of it without me telling it to pointing to the idea that I wasn\'t in control of it (great, my computer found a way to outsmart me).

In response to your final question Mike, I use avg free antivirus and everyone\'s favorite windows firewall on each windows machine but also have a Netgear hardware SPI firewall setup between the modem and my switch.

OK, thanks for that. I was wondering if ZoneAlarm could have been implicated somehow (it was running on my PCs when I experienced the problem), but it sounds like its something else.

---

I'm a volunteer and my views are my own.
News and Announcements and FAQ

Tim,

If your reason for wanting to know the timestep is to schedule closedowns, reboots or backups, then there is another way. There should be a file named cc_config.xml in your BOINC folder. Mine looks like this:

<cc_config>
   <log_flags>
      <task>1</task>
      <file_xfer>1</file_xfer>
      <sched_ops>1</sched_ops>
      [color=blue]<checkpoint_debug>1</checkpoint_debug>[/color]
   </log_flags>
   <options>
      <save_stats_days>90</save_stats_days>
   </options>
</cc_config>

The highlighted line causes a message to be added in BOINC Manager\'s messages tab each time the model hits a checkpoint (every six days in HADCM3). That\'ll give you the timestep in a round-about sort of way.

Only change the file when BOINC is not running, then restart BOINC Manager and select \'Advanced | Read config file\' to load it.

You\'ll then get stuff like this:

15/02/2008 00:48:43|climateprediction.net|[checkpoint_debug] result hadcm3iozn_cpgy_2000_80_125898779_2 checkpointed
15/02/2008 01:17:24|climateprediction.net|[checkpoint_debug] result hadcm3iozn_cpgy_2000_80_125898779_2 checkpointed
15/02/2008 01:41:44|climateprediction.net|[checkpoint_debug] result hadcm3iozn_cpgy_2000_80_125898779_2 checkpointed
15/02/2008 02:08:35|climateprediction.net|[checkpoint_debug] result hadcm3iozn_cpgy_2000_80_125898779_2 checkpointed

Iain

I\'d like to make a slight correction to this:
\"... should be a file ...\" > \"... needs to be a file ...\"

This is a debuging file, and won\'t exist unless you create it.

And just to be quite clear:

(every six days in HADCM3)

refers to MODEL days.

Thank you Ian and Les.

Both your posts are very helpful. Backups are exactly what I\'m worried about and I will try this the next time I have a free moment and let you know how it goes!

Ian,

I setup the debug file and the flags are streaming in. This system gives me a very good approximation of the timestep for the HADCM3 model and I appreciate it!!

Curious though, I have a slab model running to that it also flags... any idea the number of model days for each of those?

Thanks again!

...Curious though, I have a slab model running to that it also flags... any idea the number of model days for each of those? ...

The slab is an awkward customer: it checkpoints every three model days and trickles every 10,802 steps. Each day is 48 atmosphere update steps, so 75 checkpoints is 75 x 48 x 3 steps = 10,800, which is two steps short of a trickle!

There are 24 trickles per phase, so 24 x 10,802 = 259,248 steps = 5,401 days = 15 years + 1 day (where a year is 360 days).

I\'ve no idea why! I\'m sure MikeMarsUK has been through this somewhere before.

Each day is 48 atmosphere update steps, so 75 checkpoints is 75 x 48 x 3 steps = 10,800, which is two steps short of a trickle!

There are 24 trickles per phase, so 24 x 10,802 = 259,248 steps = 5,401 days = 15 years + 1 day (where a year is 360 days).

I\'ve no idea why!

LOL. You\'re absolutely right... that is bizarre. But hey, can\'t beat the math I guess. Thanks again!!

-Tim

We may find wider use for the phrase \'being two timesteps short of a trickle\'......

---

Cpdn news