not getting cpu tasks

merle van osdol
merle van osdol
Joined: 1 Mar 05
Posts: 513
Credit: 60724446
RAC: 0
Topic 197790

I can't get any cpu tasks and I don't understand the message that is referenced:
11/10/2014 5:52:37 AM | Einstein@Home | see scheduler log messages on http://einstein5.aei.uni-hannover.de/EinsteinAtHome/host_sched_logs/11681/11681266

merle

What is freedom of expression? Without the freedom to offend, it ceases to exist.

— Salman Rushdie

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5874
Credit: 118286929430
RAC: 25014689

not getting cpu tasks

The scheduler log contains the line

2014-11-10 12:50:19.1115 [PID=2015 ] [send] CPU: req 0.00 sec, 0.00 instances; est delay 0.00 which shows that your BOINC client was requesting 0.00 secs of CPU work.

You have two AMD GPUs in that host, of which at least one is listed as "Tahiti" class with 3GB RAM. I presume you are running multiple GPU tasks concurrently. If you happen to be running 8 concurrent GPU tasks in total, that would explain why BOINC is not requesting CPU work.

The default is for each GPU task to need the 'assistance' of 0.5 CPU cores - ie 4 cores for 8 GPU tasks. You could change (lower) that default and that would allow some CPU tasks to run. If you did, GPU performance would likely suffer quite severely. Another alternative would be to run 7 GPU tasks instead of 8. That would allow 1 CPU task to start up, but at the expense of lower GPU output. A third alternative would be to upgrade to a CPU with more threads - ie 4 core 8 threads would give 4 CPU tasks with 8 GPU tasks but once again, GPU performance would probably suffer to some extent.

If you want to have 2 GPUs in that machine, then running 8 concurrent and zero CPU tasks may well turn out to be best from a crunching perspective, particularly if they are both Tahiti class GPUs. Is your machine a 'work' machine or is it dedicated to crunching?

Cheers,
Gary.

merle van osdol
merle van osdol
Joined: 1 Mar 05
Posts: 513
Credit: 60724446
RAC: 0

I just made some adjustments.

I just made some adjustments. I made some stupid mistakes. (Had cpu usage at 50% instead of 75%) I now have one cpu task running while not downgrading my gpu performance. I use it as primarily a cruncher especially in the cooler weather but I also like to use it for ordinary purposes. Running the way I have it now with some free additional capacity for general tasks seems about perfect for what I was looking for.
I have an i7 4790K with HT turned off to help push thru gpu tasks. Thanks very much for your input and concern. See ya around.

--edit
I have 2 Tahiti's a 270x and a 280x. 850 watt gold PSU.

merle

What is freedom of expression? Without the freedom to offend, it ceases to exist.

— Salman Rushdie

Claggy
Claggy
Joined: 29 Dec 06
Posts: 560
Credit: 2723818
RAC: 1334

RE: I can't get any cpu

Quote:
I can't get any cpu tasks and I don't understand the message that is referenced:
11/10/2014 5:52:37 AM | Einstein@Home | see scheduler log messages on http://einstein5.aei.uni-hannover.de/EinsteinAtHome/host_sched_logs/11681/11681266


You're got plenty of CPU tasks on that host (now), i make it you're got 125 Gamma-ray pulsar search #4 v1.04 (FGRP4-SSE2) tasks on that host:

In progress Gamma-ray pulsar search #4 tasks for computer 11681266

But you also keep aborting them:

Error Gamma-ray pulsar search #4 tasks for computer 11681266

You're got a quad core processor, But you're limited to only using 2 cores, Boinc isn't asking for CPU work, only GPU work, You're got the min cache set for 10 days work,
You're asking for 10.5 days of GPU work for each GPU, this project has maximum deadlines of 14 days,
The other use of the min cache settings is how many days Boinc is going to have unavailability of internet access,
So Boinc will need to do that work 10 days early, meaning Boinc has 4 days to get that 10 days of work done,
The scheduler should refuse to send you more work if you can't possibly get it done in time,
Set a more reasonable cache size, a couple of days is all you need at this project:

Quote:
2014-11-10 12:50:19.1108 [PID=2015] Request: [USER#xxxxx] [HOST#11681266] [IP xxx.xxx.xxx.93] client 7.2.42
2014-11-10 12:50:19.1115 [PID=2015 ] [send] effective_ncpus 2 max_jobs_on_host_cpu 999999 max_jobs_on_host 999999
2014-11-10 12:50:19.1115 [PID=2015 ] [send] effective_ngpus 2 max_jobs_on_host_gpu 999999
2014-11-10 12:50:19.1115 [PID=2015 ] [send] Not using matchmaker scheduling; Not using EDF sim
2014-11-10 12:50:19.1115 [PID=2015 ] [send] CPU: req 0.00 sec, 0.00 instances; est delay 0.00
2014-11-10 12:50:19.1115 [PID=2015 ] [send] ATI: req 1822018.04 sec, 0.00 instances; est delay 544613.28
2014-11-10 12:50:19.1115 [PID=2015 ] [send] work_req_seconds: 0.00 secs
2014-11-10 12:50:19.1115 [PID=2015 ] [send] available disk 3.34 GB, work_buf_min 864000
2014-11-10 12:50:19.1115 [PID=2015 ] [send] active_frac 0.999936 on_frac 0.990347 DCF 0.733791
2014-11-10 12:50:19.1223 [PID=2015 ] [send] [HOST#11681266] is reliable
2014-11-10 12:50:19.1225 [PID=2015 ] [send] set_trust: error rate 0.310944 > 0.050000, don't trust
2014-11-10 12:50:19.1225 [PID=2015 ] [mixed] sending non-locality work first (0.4721)
2014-11-10 12:50:19.1513 [PID=2015 ] [version] Checking plan class 'FGRP4-SSE2'
2014-11-10 12:50:19.1533 [PID=2015 ] [version] reading plan classes from file '/BOINC/projects/EinsteinAtHome/plan_class_spec.xml'
2014-11-10 12:50:19.1533 [PID=2015 ] [version] numerical Windows version: 601760100 (Microsoft Windows 7 Ultimate x64 Edition, Service Pack 1, (06.01.7601.00))
2014-11-10 12:50:19.1533 [PID=2015 ] [version] plan class ok
2014-11-10 12:50:19.1533 [PID=2015 ] [version] Don't need CPU jobs, skipping version 104 for hsgamma_FGRP4 (FGRP4-SSE2)
2014-11-10 12:50:19.1533 [PID=2015 ] [version] no app version available: APP#27 (hsgamma_FGRP4) PLATFORM#9 (windows_x86_64) min_version 0
2014-11-10 12:50:19.1533 [PID=2015 ] [version] no app version available: APP#27 (hsgamma_FGRP4) PLATFORM#2 (windows_intelx86) min_version 0
2014-11-10 12:50:19.1636 [PID=2015 ] [mixed] sending locality work second
2014-11-10 12:50:19.1666 [PID=2015 ] [debug] [HOST#11681266] MSG(high) No work sent
2014-11-10 12:50:19.1666 [PID=2015 ] [debug] [HOST#11681266] MSG(high) see scheduler log messages on http://einstein5.aei.uni-hannover.de/EinsteinAtHome/host_sched_logs/11681/11681266
2014-11-10 12:50:19.1666 [PID=2015 ] Sending reply to [HOST#11681266]: 0 results, delay req 60.00
2014-11-10 12:50:19.1676 [PID=2015 ] Scheduler ran 0.060 seconds

Claggy

merle van osdol
merle van osdol
Joined: 1 Mar 05
Posts: 513
Credit: 60724446
RAC: 0

RE: RE: I can't get any

Quote:
Quote:
I can't get any cpu tasks and I don't understand the message that is referenced:
11/10/2014 5:52:37 AM | Einstein@Home | see scheduler log messages on http://einstein5.aei.uni-hannover.de/EinsteinAtHome/host_sched_logs/11681/11681266

You're got plenty of CPU tasks on that host (now), i make it you're got 125 Gamma-ray pulsar search #4 v1.04 (FGRP4-SSE2) tasks on that host:

In progress Gamma-ray pulsar search #4 tasks for computer 11681266

But you also keep aborting them:

Error Gamma-ray pulsar search #4 tasks for computer 11681266

You're got a quad core processor, But you're limited to only using 2 cores, Boinc isn't asking for CPU work, only GPU work, You're got the min cache set for 10 days work,
You're asking for 10.5 days of GPU work for each GPU, this project has maximum deadlines of 14 days,
The other use of the min cache settings is how many days Boinc is going to have unavailability of internet access,
So Boinc will need to do that work 10 days early, meaning Boinc has 4 days to get that 10 days of work done,
The scheduler should refuse to send you more work if you can't possibly get it done in time,
Set a more reasonable cache size, a couple of days is all you need at this project:

Quote:
2014-11-10 12:50:19.1108 [PID=2015] Request: [USER#xxxxx] [HOST#11681266] [IP xxx.xxx.xxx.93] client 7.2.42
2014-11-10 12:50:19.1115 [PID=2015 ] [send] effective_ncpus 2 max_jobs_on_host_cpu 999999 max_jobs_on_host 999999
2014-11-10 12:50:19.1115 [PID=2015 ] [send] effective_ngpus 2 max_jobs_on_host_gpu 999999
2014-11-10 12:50:19.1115 [PID=2015 ] [send] Not using matchmaker scheduling; Not using EDF sim
2014-11-10 12:50:19.1115 [PID=2015 ] [send] CPU: req 0.00 sec, 0.00 instances; est delay 0.00
2014-11-10 12:50:19.1115 [PID=2015 ] [send] ATI: req 1822018.04 sec, 0.00 instances; est delay 544613.28
2014-11-10 12:50:19.1115 [PID=2015 ] [send] work_req_seconds: 0.00 secs
2014-11-10 12:50:19.1115 [PID=2015 ] [send] available disk 3.34 GB, work_buf_min 864000
2014-11-10 12:50:19.1115 [PID=2015 ] [send] active_frac 0.999936 on_frac 0.990347 DCF 0.733791
2014-11-10 12:50:19.1223 [PID=2015 ] [send] [HOST#11681266] is reliable
2014-11-10 12:50:19.1225 [PID=2015 ] [send] set_trust: error rate 0.310944 > 0.050000, don't trust
2014-11-10 12:50:19.1225 [PID=2015 ] [mixed] sending non-locality work first (0.4721)
2014-11-10 12:50:19.1513 [PID=2015 ] [version] Checking plan class 'FGRP4-SSE2'
2014-11-10 12:50:19.1533 [PID=2015 ] [version] reading plan classes from file '/BOINC/projects/EinsteinAtHome/plan_class_spec.xml'
2014-11-10 12:50:19.1533 [PID=2015 ] [version] numerical Windows version: 601760100 (Microsoft Windows 7 Ultimate x64 Edition, Service Pack 1, (06.01.7601.00))
2014-11-10 12:50:19.1533 [PID=2015 ] [version] plan class ok
2014-11-10 12:50:19.1533 [PID=2015 ] [version] Don't need CPU jobs, skipping version 104 for hsgamma_FGRP4 (FGRP4-SSE2)
2014-11-10 12:50:19.1533 [PID=2015 ] [version] no app version available: APP#27 (hsgamma_FGRP4) PLATFORM#9 (windows_x86_64) min_version 0
2014-11-10 12:50:19.1533 [PID=2015 ] [version] no app version available: APP#27 (hsgamma_FGRP4) PLATFORM#2 (windows_intelx86) min_version 0
2014-11-10 12:50:19.1636 [PID=2015 ] [mixed] sending locality work second
2014-11-10 12:50:19.1666 [PID=2015 ] [debug] [HOST#11681266] MSG(high) No work sent
2014-11-10 12:50:19.1666 [PID=2015 ] [debug] [HOST#11681266] MSG(high) see scheduler log messages on http://einstein5.aei.uni-hannover.de/EinsteinAtHome/host_sched_logs/11681/11681266
2014-11-10 12:50:19.1666 [PID=2015 ] Sending reply to [HOST#11681266]: 0 results, delay req 60.00
2014-11-10 12:50:19.1676 [PID=2015 ] Scheduler ran 0.060 seconds

Claggy

Will do. And thanks Claggy.

merle

What is freedom of expression? Without the freedom to offend, it ceases to exist.

— Salman Rushdie

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.