Indeed. It appears as if the estimated GFLOPS setting is substantially too low.

Also, although this isn't a problem per se, these tasks seem to have a linear and well behaved progress bar. It would be beneficial for them to utilize BOINC's <fraction_done_exact/> mechanism, which would override BOINC's default "smart" estimation. This would compensate for the bad rsc_fpops_est setting.

Fortunately, YOU can set <fraction_done_exact/> yourself by using app_config.xml. I haven't tried it myself for this beta, but this should work:

<app_config>
  <app>
    <name>beta27</name>
    <fraction_done_exact>1</fraction_done_exact>
  </app>
</app_config>

That file goes in the projects/www.worldcommunitygrid.org directory.

Once the task starts running, BOINC will compute the estimated total time as exactly <elapsed time> / <progress>. On these beta tasks, that should give a fairly accurate remaining time estimate.

It won't work until the task has started running, and you may need to restart BOINC for it to take effect.

I think beta27 is the correct app name. Can anyone confirm that?

This assumes you're using the standard BOINC client. All bets are off if you're using a customized version of BOINC, which might or might not support this feature.

Correct.


I changed it myself in client_state, because I already lost 2 tasks with exceeded time limit as I mentioned in an earlier post and don't want to loose tasks for that reason. Hopefully new beta-tasks will have a more realistic rsc_fpops_est.
Beneficial side effect: Testing the restart of tasks.

Before I could change it on this machine another task with the exceeded time limit error:

https://www.worldcommunitygrid.org/ms/device/...og.do?resultId=1000181925

World Community Grid 31 May 16:03:33 Aborting task BETA_ARP1_0000370_000_1: exceeded elapsed time limit 153481.54 (547904.24G/3.57G)

The current error and invalid rate for this beta is low and looking good. We will not be adding any new work units until probably early next week. We will add the automation pieces to the mix that takes the validated results and builds the next 48 hour simulation period. Once this is enabled all of the validated results from the current 2000 simulations will have work available for the continuation. Then as those finish and validate the next 48 hours will be built and loaded. For beta we will probably go 3 - 5 levels deep or in terms of simulation time 6 to 10 days of simulation. The current plan for production is to simulate an entire calendar year.

Thanks,
armstrdj

Hi devs, WU 1142300482
(https://www.worldcommunitygrid.org/ms/device/....do?workunitId=1142300482)

Host: WCG deviceId 5429224

- dual E5-2660v2 w/HT on dedicated to only WCG
- all 40 threads run WCG random units
- CPU threads are capped at 95% each using systemd
- Debian 9 basic install, which places all threads into +20 nice
- CPU temps remain about 75C on this chassis but is +20C between CPUs:
- BOINC is configured in "run always" mode and prefs override sets all values to 100% (as systemd will take care of limiting them now)

$ sensors | grep Physical
Physical id 1: +75.0°C (high = +76.0°C, crit = +86.0°C)
Physical id 0: +57.0°C (high = +76.0°C, crit = +86.0°C)

Chassis: Supermicro X9DRi-LN4+/X9DR3-LN4+

Please let me know if I can provide any more useful information about the host which ran the beta WU. It appears I've only been farmed out this one.

Edit: added note about configuration of BOINC client

I got rid of my old machine that had dual Pentium-III orocessors some years ago. Likewise, I had a machine with two 32-bit 3.06 GHz Xeon processors that died after 10 year or so when tropical storm Sand took it out. I now have a slower (1.8 GHz) four-core 64-bit Xeon processor and 16 GBytes of RAM. -- 8 modules of 2 GByte RAM, the kind that must be installed in pairs. This kind:

Capacity: 2GB Module - ECC Reg - DDR3-10600 (PC3-1333)
https://www.amazon.com/gp/product/B00HH8FOEK/...o06_s00?ie=UTF8&psc=1

If I remember correctly, I paid about 10x as much for the first 8GBytes of that stuff when I got that machine.

Detected 1795.700 MHz processor.
Calibrating delay loop (skipped), value calculated using timer frequency.. 3591.40 BogoMIPS (lpj=1795700)
pid_max: default: 32768 minimum: 301
Security Framework initialized

CPU0: Intel(R) Xeon(R) CPU E5-2603 0 @ 1.80GHz stepping 07
Performance Events: PEBS fmt1+, 16-deep LBR, SandyBridge events, full-width counters, Intel PMU driver.
... version: 3
... bit width: 48
... generic registers: 8
... value mask: 0000ffffffffffff
... max period: 0000ffffffffffff
... fixed-purpose events: 3
... event mask: 00000007000000ff
NMI watchdog enabled, takes one hw-pmu counter.
Booting Node 0, Processors #1 #2 #3
Brought up 4 CPUs
Total of 4 processors activated (14365.60 BogoMIPS).