climateprediction.net home page
New work Discussion

New work Discussion

Message boards : Number crunching : New work Discussion
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 30 · 31 · 32 · 33 · 34 · 35 · 36 . . . 80 · Next

AuthorMessage
Profile Bonsai911

Send message
Joined: 9 Sep 04
Posts: 226
Credit: 29,990,607
RAC: 0
Message 60987 - Posted: 25 Sep 2019, 12:23:23 UTC
Last modified: 25 Sep 2019, 12:49:31 UTC

No wu transfer?

Boinc Manager shows:

25.09.2019 14:18:47 | climateprediction.net | No tasks sent
25.09.2019 14:18:47 | climateprediction.net | No tasks are available for UK Met Office Coupled Model Full Resolution Ocean
25.09.2019 14:18:47 | climateprediction.net | No tasks are available for UK Met Office HadCM3 short

But on the server status are 1607 UK Met Office HadCM3 short are shown.

Is this a server failure?

Ahhhhh, only for Linux/x86[/color] [color=indigo]!!!!
(Layout tags shown?!)
Greetings

bonsai911
ID: 60987 · Report as offensive     Reply Quote
Les Bayliss
Volunteer moderator

Send message
Joined: 5 Sep 04
Posts: 7578
Credit: 24,034,754
RAC: 70
Message 60988 - Posted: 25 Sep 2019, 12:48:42 UTC - in response to Message 60987.  

"short" are now Linux only.
See this page.
ID: 60988 · Report as offensive     Reply Quote
Profile Bonsai911

Send message
Joined: 9 Sep 04
Posts: 226
Credit: 29,990,607
RAC: 0
Message 60989 - Posted: 25 Sep 2019, 12:50:32 UTC - in response to Message 60988.  

Got it. Thanks Les
ID: 60989 · Report as offensive     Reply Quote
Thomas Wiegand

Send message
Joined: 4 Jul 19
Posts: 31
Credit: 252,192
RAC: 0
Message 60990 - Posted: 25 Sep 2019, 13:02:37 UTC

yeah, some work
23 new tasks here - let it calculate
ID: 60990 · Report as offensive     Reply Quote
Profile Dave Jackson
Volunteer moderator

Send message
Joined: 15 May 09
Posts: 3340
Credit: 9,948,339
RAC: 791
Message 60991 - Posted: 25 Sep 2019, 14:03:31 UTC - in response to Message 60990.  
Last modified: 25 Sep 2019, 14:12:16 UTC

yeah, some work
23 new tasks here - let it calculate


Batch 835 hadcm3s for Linux 7470tasks.

Grab them while you can because they won't last long. Fast machines should finish them in about three or four days.

Edit: Server status showing 1427 unsent at 13:16UTC so about an hour ago so not all uploaded to server at that point. Because of some issues I won't go into, the last of them may not make the server for an hour or so yet.
ID: 60991 · Report as offensive     Reply Quote
Jim1348

Send message
Joined: 15 Jan 06
Posts: 606
Credit: 26,339,909
RAC: 853
Message 60999 - Posted: 26 Sep 2019, 2:27:28 UTC - in response to Message 60991.  
Last modified: 26 Sep 2019, 2:37:48 UTC

Grab them while you can because they won't last long. Fast machines should finish them in about three or four days..

My fastest machine, an i7-9700 estimates almost 5 days total after 12 hours of running time.
My Ryzen 2600 estimates a little over 5 days.
My Ryzen 1700 estimates almost 6 days.

Edit: And my i7-8700 almost 9 days.

If these estimates hold up (based on % completed, all cores fully loaded), it appears to me that these work units do well on the Ryzens, relatively speaking, considering that the i7-9700 is using full cores whereas all the others are on virtual cores. Maybe the final results will be a bit different though.
ID: 60999 · Report as offensive     Reply Quote
Thomas Wiegand

Send message
Joined: 4 Jul 19
Posts: 31
Credit: 252,192
RAC: 0
Message 61000 - Posted: 26 Sep 2019, 3:14:18 UTC - in response to Message 60991.  
Last modified: 26 Sep 2019, 3:15:40 UTC

Batch 835 hadcm3s for Linux 7470tasks.

404 left to take, 3320 in work (? 7470 - 404 - 3320 = 3.746 area stored at computers for later start, like at me 18 I guess)

here are
* i7-4770 : 8 on work (7d10h est), 4 more ready to start
* i7-4770 : 7 on work (7d11h est), 14 more waiting to start (mean ready in another more 16 days)
* i5-3470 : 3 still Weather At Home 2 (wah2) v8.24 windows_intelx86 (one failed earlier)
* i3-2100 : 4 on work (11d14h)
I do my best and let them run also night, not only on sunpower
ID: 61000 · Report as offensive     Reply Quote
Jim1348

Send message
Joined: 15 Jan 06
Posts: 606
Credit: 26,339,909
RAC: 853
Message 61013 - Posted: 26 Sep 2019, 12:23:21 UTC - in response to Message 60999.  
Last modified: 26 Sep 2019, 12:24:43 UTC

If these estimates hold up (based on % completed, all cores fully loaded), it appears to me that these work units do well on the Ryzens, relatively speaking,

While we are on the subject, I tried my new Ryzen 3700x. I could get only a single hadcm3s, so loaded up the other cores with WCG/MCM. But the estimated time is an even four days; very nice. So that is the one I will use, hoping that the CPDN server can figure out how to send the right work units. I have only 16 GB on that machine, though could increase it to 32 GB.

The Ryzens do not normally have such a big lead on the Intels, and the Intels do better in some projects. I think it has to do with the cache and memory managers, which seems to favor the Ryzens here.
ID: 61013 · Report as offensive     Reply Quote
Profile Dave Jackson
Volunteer moderator

Send message
Joined: 15 May 09
Posts: 3340
Credit: 9,948,339
RAC: 791
Message 61014 - Posted: 26 Sep 2019, 13:23:08 UTC - in response to Message 61013.  
Last modified: 26 Sep 2019, 14:36:21 UTC

I could get only a single hadcm3s,


That is because of problems uploading batch 835 to the servers. There are about another 5,000 still to be uploaded. Sarah is working on that at the moment.

Edit: The rest will be 837 when they go out. (If nothing changes)
ID: 61014 · Report as offensive     Reply Quote
Jim1348

Send message
Joined: 15 Jan 06
Posts: 606
Credit: 26,339,909
RAC: 853
Message 61017 - Posted: 26 Sep 2019, 15:41:58 UTC - in response to Message 61014.  
Last modified: 26 Sep 2019, 15:47:42 UTC

Edit: The rest will be 837 when they go out. (If nothing changes)

Yes, I just picked up 15 of them, which gives me a full load on my Ryzen 3700x even though I have nine hours of WCG work left.
And I don't use a large buffer either, just 0.1 + 0.5 days. It will keep me busy for a while, which is very nice.
ID: 61017 · Report as offensive     Reply Quote
Profile geophi
Volunteer moderator

Send message
Joined: 7 Aug 04
Posts: 2096
Credit: 57,971,859
RAC: 5,479
Message 61021 - Posted: 26 Sep 2019, 16:39:33 UTC - in response to Message 61013.  

While we are on the subject, I tried my new Ryzen 3700x. I could get only a single hadcm3s, so loaded up the other cores with WCG/MCM.

When I first attach a PC, it will often only download one task. Then, if the task is still running an hour later, it will request enough to load up the rest of the cores. Maybe a boinc server config thing?
ID: 61021 · Report as offensive     Reply Quote
Jim1348

Send message
Joined: 15 Jan 06
Posts: 606
Credit: 26,339,909
RAC: 853
Message 61023 - Posted: 26 Sep 2019, 17:58:28 UTC - in response to Message 61021.  

When I first attach a PC, it will often only download one task. Then, if the task is still running an hour later, it will request enough to load up the rest of the cores.

Yes, but I just caught it at the wrong time. They had run out before I could get more.
However, I am now loaded up again with 837, so should be able to do an accurate test.
ID: 61023 · Report as offensive     Reply Quote
Profile Dave Jackson
Volunteer moderator

Send message
Joined: 15 May 09
Posts: 3340
Credit: 9,948,339
RAC: 791
Message 61028 - Posted: 27 Sep 2019, 6:09:29 UTC

Edit: The rest will be 837 when they go out. (If nothing changes)


Well about 75% of the rest went out. The others got caught up in the file naming problem that was responsible for them not all going out initially and not until this week for any of them. So expect a new batch at some point. (The main cause of the problem was having tasks for several different experiments in one batch. Next time the person responsible will do separate smaller batches to start with.)
ID: 61028 · Report as offensive     Reply Quote
Thomas Wiegand

Send message
Joined: 4 Jul 19
Posts: 31
Credit: 252,192
RAC: 0
Message 61033 - Posted: 27 Sep 2019, 11:04:53 UTC

suddenly 6 tasks crashed with a computer freeze, reason not known, as not any other program run as normal.
I guess this 6 task goes back, and will be run somewhere ale again. Good luck
ID: 61033 · Report as offensive     Reply Quote
Jean-David Beyer

Send message
Joined: 5 Aug 04
Posts: 601
Credit: 9,422,989
RAC: 376
Message 61035 - Posted: 27 Sep 2019, 12:30:27 UTC

I got these recently:

Wed 25 Sep 2019 09:25:53 AM EDT Starting task hadcm3s_hd57_190012_240_835_011892279_0 using hadcm3s version 836 in slot 4
Wed 25 Sep 2019 09:25:53 AM EDT Starting task hadcm3s_hd55_190012_240_835_011892277_0 using hadcm3s version 836 in slot 5

They seen to be running just fine and both have done two trickles and seem to be about 11.2% done.
ID: 61035 · Report as offensive     Reply Quote
Jim1348

Send message
Joined: 15 Jan 06
Posts: 606
Credit: 26,339,909
RAC: 853
Message 61063 - Posted: 28 Sep 2019, 23:44:10 UTC

I have rebooted my i7-8700 (Ubuntu 18.04) a couple of times to do updates, and each time picked up a few errors on 835 and 837.
I think the errors are unrelated to the updates themselves, which appear to have nothing to do with the work units, but only to the reboots.
https://www.cpdn.org/cpdnboinc/results.php?hostid=1492331

So it seems that they are not very robust, but I usually don't have to reboot, so maybe it will not happen again.
ID: 61063 · Report as offensive     Reply Quote
Profile Dave Jackson
Volunteer moderator

Send message
Joined: 15 May 09
Posts: 3340
Credit: 9,948,339
RAC: 791
Message 61065 - Posted: 29 Sep 2019, 6:35:03 UTC - in response to Message 61063.  

I have rebooted my i7-8700 (Ubuntu 18.04) a couple of times to do updates, and each time picked up a few errors on 835 and 837.


In my experience, reboots after a kernel update are much more likely to produce errors than other reboots. As most updates that require a reboot I notice on my own system are ones that involve a kernel update, wondering if this is the case here?
ID: 61065 · Report as offensive     Reply Quote
Profile Dave Jackson
Volunteer moderator

Send message
Joined: 15 May 09
Posts: 3340
Credit: 9,948,339
RAC: 791
Message 61066 - Posted: 29 Sep 2019, 6:38:43 UTC
Last modified: 29 Sep 2019, 6:40:24 UTC

Batch 838 &50 hadam4 tasks at N144 resolution. (I should really look up exactly what N144 and N216 mean.)
ID: 61066 · Report as offensive     Reply Quote
Jim1348

Send message
Joined: 15 Jan 06
Posts: 606
Credit: 26,339,909
RAC: 853
Message 61068 - Posted: 29 Sep 2019, 7:43:06 UTC - in response to Message 61065.  
Last modified: 29 Sep 2019, 7:45:30 UTC

In my experience, reboots after a kernel update are much more likely to produce errors than other reboots. As most updates that require a reboot I notice on my own system are ones that involve a kernel update, wondering if this is the case here?

Sort of. The kernel stayed the same "Ubuntu 18.04.3 LTS [4.15.0-64-generic|libc 2.27 (Ubuntu GLIBC 2.27-3ubuntu1)]", but there are always a lot of security updates.
That probably explains it. I just won't reboot. It was not really necessary anyway.

EDIT: But I am glad I did. I picked up four "8.09 UK Met Office HadAM4 at N144 resolution" to replace them. They are beginning to use some memory.
ID: 61068 · Report as offensive     Reply Quote
Thomas Wiegand

Send message
Joined: 4 Jul 19
Posts: 31
Credit: 252,192
RAC: 0
Message 61075 - Posted: 29 Sep 2019, 14:41:39 UTC

another freeze in the evening - , same computer- and afternoon 4 tasks with failure computing
(it is a bit like nonsense energy using) hmm - but if continue like this, they are all gone fast, without a result
ID: 61075 · Report as offensive     Reply Quote
Previous · 1 . . . 30 · 31 · 32 · 33 · 34 · 35 · 36 . . . 80 · Next

Message boards : Number crunching : New work Discussion

©2022 climateprediction.net