O3 almost not using any GPU power on a Mac

[AF>Le_Pommier] Jerome_C2005
[AF>Le_Pommier]...
Joined: 1 May 10
Posts: 38
Credit: 111591916
RAC: 0

Hello I'm trying to get *any*

Hello I'm trying to get *any* kind of GPU work (no CPU) for my iMac now, and I can't.

2024-04-07 15:49:42.9374 [PID=1197045]   Request: [USER#xxxxx] [HOST#12895206] [IP xxx.xxx.xxx.200] client 7.24.2
2024-04-07 15:49:42.9374 [PID=1197045] [debug]  [HOST#0] rpc_time:0 timezone:7200 d_total:379999997952.000000 d_free:131385344000.000000 d_boinc_used_total:39952949248.000000 d_boinc_used_project:976257024.000000 d_boinc_max:0.000000
2024-04-07 15:49:42.9465 [PID=1197045] [debug]   have_master:0 have_working: 1 have_db: 1
2024-04-07 15:49:42.9465 [PID=1197045] [debug]   using working prefs
2024-04-07 15:49:42.9465 [PID=1197045] [debug]   have db 1; dbmod 1670364416.000000; global mod 1710067846.000000
2024-04-07 15:49:42.9465 [PID=1197045] [debug]   sending db prefs in reply
2024-04-07 15:49:42.9466 [PID=1197045]    [send] DCF=11.511993; setting to 10
2024-04-07 15:49:42.9466 [PID=1197045]    [send] effective_ncpus 9 max_jobs_on_host_cpu 999999 max_jobs_on_host 999999
2024-04-07 15:49:42.9466 [PID=1197045]    [send] effective_ngpus 1 max_jobs_on_host_gpu 999999
2024-04-07 15:49:42.9466 [PID=1197045]    [send] Not using matchmaker scheduling; Not using EDF sim
2024-04-07 15:49:42.9466 [PID=1197045]    [send] CPU: req 0.00 sec, 0.00 instances; est delay 0.00
2024-04-07 15:49:42.9467 [PID=1197045]    [send] ATI: req 25920.00 sec, 1.00 instances; est delay 0.00
2024-04-07 15:49:42.9467 [PID=1197045]    [send] work_req_seconds: 0.00 secs
2024-04-07 15:49:42.9467 [PID=1197045]    [send] available disk 22.79 GB, work_buf_min 8640
2024-04-07 15:49:42.9467 [PID=1197045]    [send] active_frac 0.998515 on_frac 0.975540 DCF 11.511993
2024-04-07 15:49:42.9474 [PID=1197045]    [mixed] sending locality work first (0.4854)
2024-04-07 15:49:42.9508 [PID=1197045]    [send] send_old_work() no feasible result older than 336.0 hours
2024-04-07 15:49:42.9609 [PID=1197045]    [send] send_old_work() no feasible result younger than 188.4 hours and older than 168.0 hours
2024-04-07 15:49:43.2105 [PID=1197045] [CRITICAL]   get_working_set_filename(fasthost): pattern not found (file list empty)
2024-04-07 15:49:43.2105 [PID=1197045] [CRITICAL]   get_working_set_filename(fasthost): pattern not found (file list empty)
2024-04-07 15:49:43.2105 [PID=1197045] [CRITICAL]   get_working_set_filename(fasthost): pattern not found (file list empty)
2024-04-07 15:49:43.2105 [PID=1197045] [CRITICAL]   get_working_set_filename(fasthost): pattern not found (file list empty)
2024-04-07 15:49:43.2106 [PID=1197045]    [mixed] sending non-locality work second
2024-04-07 15:49:43.2548 [PID=1197045]    [version] Don't need CPU jobs, skipping version 160 for einsteinbinary_BRP4G ()
2024-04-07 15:49:43.2549 [PID=1197045]    [version] no app version available: APP#25 (einsteinbinary_BRP4G) PLATFORM#10 (x86_64-apple-darwin) min_version 0
2024-04-07 15:49:43.2549 [PID=1197045]    [version] no app version available: APP#25 (einsteinbinary_BRP4G) PLATFORM#6 (i686-apple-darwin) min_version 0
2024-04-07 15:49:43.2549 [PID=1197045]    [version] no app version available: APP#57 (einsteinbinary_BRP7) PLATFORM#10 (x86_64-apple-darwin) min_version 0
2024-04-07 15:49:43.2549 [PID=1197045]    [version] no app version available: APP#57 (einsteinbinary_BRP7) PLATFORM#6 (i686-apple-darwin) min_version 0
2024-04-07 15:49:43.2550 [PID=1197045]    [version] Checking plan class 'opencl-intel_gpu-lion-Beta'
2024-04-07 15:49:43.2563 [PID=1197045]    [version] reading plan classes from file '/BOINC/projects/EinsteinAtHome/plan_class_spec.xml'
2024-04-07 15:49:43.2563 [PID=1197045]    [version] beta test app versions not allowed in project prefs.
2024-04-07 15:49:43.2563 [PID=1197045]    [version] no app version available: APP#19 (einsteinbinary_BRP4) PLATFORM#10 (x86_64-apple-darwin) min_version 0
2024-04-07 15:49:43.2563 [PID=1197045]    [version] no app version available: APP#19 (einsteinbinary_BRP4) PLATFORM#6 (i686-apple-darwin) min_version 0
2024-04-07 15:49:43.2564 [PID=1197045]    [version] no app version available: APP#60 (einsteinbinary_BRP4A) PLATFORM#10 (x86_64-apple-darwin) min_version 0
2024-04-07 15:49:43.2564 [PID=1197045]    [version] no app version available: APP#60 (einsteinbinary_BRP4A) PLATFORM#6 (i686-apple-darwin) min_version 0
2024-04-07 15:49:43.2565 [PID=1197045]    [version] Checking plan class 'FGRPSSE'
2024-04-07 15:49:43.2565 [PID=1197045]    [version] plan class ok
2024-04-07 15:49:43.2565 [PID=1197045]    [version] Don't need CPU jobs, skipping version 111 for hsgamma_FGRP5 (FGRPSSE)
2024-04-07 15:49:43.2565 [PID=1197045]    [version] no app version available: APP#46 (hsgamma_FGRP5) PLATFORM#10 (x86_64-apple-darwin) min_version 0
2024-04-07 15:49:43.2565 [PID=1197045]    [version] no app version available: APP#46 (hsgamma_FGRP5) PLATFORM#6 (i686-apple-darwin) min_version 0
2024-04-07 15:49:43.2609 [PID=1197045] [debug]   [HOST#12895206] MSG(high) No work sent
2024-04-07 15:49:43.2609 [PID=1197045] [debug]   [HOST#12895206] MSG(high) Binary Radio Pulsar Search (MeerKAT) is not available for your type of computer.
2024-04-07 15:49:43.2609 [PID=1197045] [debug]   [HOST#12895206] MSG(high) see scheduler log messages on https://einsteinathome.org/host/12895206/log
2024-04-07 15:49:43.2609 [PID=1197045] [debug]   [HOST#12895206] MSG(high) Binary Radio Pulsar Search (Arecibo,GBT,A) is not available for your type of computer.
2024-04-07 15:49:43.2609 [PID=1197045]    Sending reply to [HOST#12895206]: 0 results, delay req 60.00
2024-04-07 15:49:43.2610 [PID=1197045]    Scheduler ran 0.328 seconds

I enabled everything in the preferences (for GPU).

Any idea why ?

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3937
Credit: 46461792642
RAC: 64116211

[AF>Le_Pommier wrote:

[AF>Le_Pommier wrote:

Jerome_C2005]

Hello I'm trying to get *any* kind of GPU work (no CPU) for my iMac now, and I can't.


 

I enabled everything in the preferences (for GPU).

Any idea why ?



there are only two projects supporting AMD GPUs, BRP7 and O3AS.

BRP7 does not have any applications for Mac.

O3AS does have an app for Mac (intel) with AMD, but the O3AS project seems to have run out of work last night or there's something wrong with the work generator. you'll have to wait for them to get more work in the pipeline, probablly not until tomorrow morning at the earliest

_________________________________________________________________________

[AF>Le_Pommier] Jerome_C2005
[AF>Le_Pommier]...
Joined: 1 May 10
Posts: 38
Credit: 111591916
RAC: 0

Thanks for your quick answer,

Thanks for your quick answer, I'll wait :)

 

It's strange because I thought O3 was not the only option for Mac, in the past (I think O3 didn't always exist, right ?) I had some work for the mac, the 100M+ credits I have were mainly cruncher on my mac(s) !

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3937
Credit: 46461792642
RAC: 64116211

Previously there was the

Previously there was the gamma ray pulsar search #1 project that had been running for many years. But that project is now ended and has not had work for many months. 

_________________________________________________________________________

[AF>Le_Pommier] Jerome_C2005
[AF>Le_Pommier]...
Joined: 1 May 10
Posts: 38
Credit: 111591916
RAC: 0

OK very clear thanks, so O3

OK very clear thanks, so O3 is the last of the mohicans for me...

And with CPU I see those 3 apps seem to have mac app

- Gamma-ray pulsar search #5 (FGRP5)

- Binary Radio Pulsar Search (Arecibo,GBT,long) (BRP4G)

- Binary Radio Pulsar Search (Arecibo,GBT) (BRP4)

do you know if they are still active / with work ? 

(if I read the server status correctly it seems most of the apps have some work) 

[AF>Le_Pommier] Jerome_C2005
[AF>Le_Pommier]...
Joined: 1 May 10
Posts: 38
Credit: 111591916
RAC: 0

FYI I started to receive

FYI I started to receive tasks 3 days ago, they are very long to run with increasing run time (it started at 11 hours and is now around 19 hours). Each one of them is processed so much quickly (30mn or less) by the wingman (always a much bigger GPU : NVIDIA TITAN V, GeForce RTX 3090).

I run 3 in parallel because this is the setup I had done (you can read previous page of this topic), with same apparent symptom as before (hardly no GPU used), which might be "not real" (see previous page), but still it's the only project I run on GPU where this happens. 

I'll probably won't try too long, it's too depressing.

mikey
mikey
Joined: 22 Jan 05
Posts: 12670
Credit: 1839069036
RAC: 4127

[AF>Le_Pommier wrote:

[AF>Le_Pommier wrote:

Jerome_C2005]

FYI I started to receive tasks 3 days ago, they are very long to run with increasing run time (it started at 11 hours and is now around 19 hours). Each one of them is processed so much quickly (30mn or less) by the wingman (always a much bigger GPU : NVIDIA TITAN V, GeForce RTX 3090).

I run 3 in parallel because this is the setup I had done (you can read previous page of this topic), with same apparent symptom as before (hardly no GPU used), which might be "not real" (see previous page), but still it's the only project I run on GPU where this happens. 

I'll probably won't try too long, it's too depressing.

Try running the next batch, just do a few dozen, at 1 at a time with the default settings and see if it uses the gpu more?

GWGeorge007
GWGeorge007
Joined: 8 Jan 18
Posts: 3060
Credit: 4958144353
RAC: 1340351

Hi Jerome!  I'll chime in

Hi Jerome!  I'll chime in again for a bit.

[AF>Le_Pommier wrote:

Jerome_C2005]

FYI I started to receive tasks 3 days ago, they are very long to run with increasing run time (it started at 11 hours and is now around 19 hours). Each one of them is processed so much quickly (30mn or less) by the wingman (always a much bigger GPU : NVIDIA TITAN V, GeForce RTX 3090).

I run 3 in parallel because this is the setup I had done (you can read previous page of this topic), with same apparent symptom as before (hardly no GPU used), which might be "not real" (see previous page), but still it's the only project I run on GPU where this happens. 

I'll probably won't try too long, it's too depressing.

First of all, yes, you do have 40 GB of system memory.  But that does not mean that you can run 5 tasks simultaneously, 3 as of yesterday, without careful planning.  Also, listed in your details page for your computer, you have "0" (zero) cache!  That is surprising.  Is that normal for Apple/Mac computers?  And what does your project preferences show for each project you're trying to use?  If you are using this computer as a daily driver also, do you have the statement in your Computing Preferences as:


Suspend when non-BOINC CPU usage is above: ??? %  ( the ??? is a figure that you can set, default is 25% )

( This will )  Suspend computing when your computer is busy running other programs.


Remember, if you are using the website to set your 'preferences', the LAST preference set will be the preferences for ALL projects.  Also, make sure that your venues match in both Computing and Project for each project you are running.

My recommendations:

If you ARE using this computer as your daily driver, set the "Suspend when non-BOINC CPU usage is above..." to something higher than the default.  I have mine set to 80%, but I am running multiple GPUs with a 32 core AMD CPU.

Try running just a single GPU task at just 1X and see what it gives in times.

Limit your CPU use to something less than your 20 cores if (because of the zero cache) you can't run multi-threads - or 40 threads if you can run multi-threads.  I would limit your CPU to 30 threads, or 15 cores, max.

Try these and then get back to us with the results, please.

IHTH!

George

Proud member of the Old Farts Association

[AF>Le_Pommier] Jerome_C2005
[AF>Le_Pommier]...
Joined: 1 May 10
Posts: 38
Credit: 111591916
RAC: 0

I don't have "0 cache" (you

I don't have "0 cache" (you refer to the parameter bottom left of my screenshot, right ?), I have 0.1 day, because I never have internet issue so I don't see the point in accumulating a big number of tasks in advance. I use local preferences and I have no limitation in boinc compute power, all is set to 100% (none of the "suspend when" options are selected).

Global CPU usage is set to 45% because I have a linux VM running (for boinc) at the same time (projects without a macOS client) = 10 threads and I prefer to let 5% not assigned to boinc so this is 9 CPU threads left for boinc.

Below I show the last 7 days stats of my mac, you can see when I started to use O3 instead of PrimeGrid early last week.

GPU memory (out of 16 GB)

GPU usage (you clearly see when I stopped PG and started O3)

Global memory usage (out of 40 GB)

So nothing seems "saturated" at all.

But task duration keeps increasing, now the last 3 took 1 day 7 hours to complete !! this is crazy.

For the 3 I have running currently, one has completed 53% after 17 hours and the other 2 have between 2 and 5% after 1 hour to 1.5 hour.

So I can try to let only 1 running at a time and see if it changes anything...

[AF>Le_Pommier] Jerome_C2005
[AF>Le_Pommier]...
Joined: 1 May 10
Posts: 38
Credit: 111591916
RAC: 0

(I have to kill those

(I have to kill those on-going tasks because even after removing my app_config and restarting boinc it was still running 3 at a time, so I cancelled them and now it's OK I have 1 task only running)

Comment viewing options

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