Welcome to MilkyWay@home

Validation inconclusive

Message boards : Number crunching : Validation inconclusive
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 14 · 15 · 16 · 17 · 18 · Next

AuthorMessage
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 990
Credit: 376,143,149
RAC: 0
Message 74688 - Posted: 20 Nov 2022, 19:49:24 UTC - in response to Message 74686.  

still unclear how they determine inconclusives if it's not being compared to anything lol
I assume that's the small number of yours which are still being randomly tested? And that number drops as you get more trustworthy.
ID: 74688 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Ian&Steve C.
Avatar

Send message
Joined: 18 Nov 22
Posts: 84
Credit: 640,530,847
RAC: 0
Message 74689 - Posted: 20 Nov 2022, 19:56:46 UTC - in response to Message 74687.  

Oh, I thought you had a few of those. What are those ones you have watercooled about 4 in one PC you showed a picture of in Einstein?

i don't run that system anymore, but it was 7x RTX 2080. that system has since been dismantled, parted out, and converted to something else. I don't have any multi-GPU setups that are water cooled like that anymore. my other multi-GPU systems are air cooled at the moment.

Odd, I see it as 7.5Tflops DP, pretty much the same as all my old Tahitis together, and I get 1 maybe 1.5 million credits a day.

theoretical specs vs real world outcome. just look at the leaderboard here and you'll see MANY TitanVs in the 2-2.5M range per GPU. and with tuning you can push it more. mine should come close to 3M if i run it long enough.

ID: 74689 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Keith Myers
Avatar

Send message
Joined: 24 Jan 11
Posts: 715
Credit: 556,164,589
RAC: 54,740
Message 74690 - Posted: 20 Nov 2022, 20:00:32 UTC
Last modified: 20 Nov 2022, 20:01:25 UTC

I believe they fit each result returned into a "not unreasonable result" filter. I believe that is defined by the:
<stream_only_likelihood2>
<search_likelihood2>
values returned for each result in the stderr.txt output. If the values are reasonable, then they don't need an extra validation task, if unreasonable the result goes into inconclusive status and warrants a wingman task.
ID: 74690 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 990
Credit: 376,143,149
RAC: 0
Message 74691 - Posted: 20 Nov 2022, 20:45:55 UTC - in response to Message 74689.  
Last modified: 20 Nov 2022, 20:46:12 UTC

theoretical specs vs real world outcome. just look at the leaderboard here and you'll see MANY TitanVs in the 2-2.5M range per GPU. and with tuning you can push it more. mine should come close to 3M if i run it long enough.
I guess those specs (I get from techpowerup) are raw specs, like looking at GHz of a CPU. Your Titan presumably has a better instruction set MW can take advantage of.

I take it you're considerably richer than me :-/ https://youtu.be/qRKbX_ilsVI?t=4
ID: 74691 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 990
Credit: 376,143,149
RAC: 0
Message 74692 - Posted: 20 Nov 2022, 20:47:28 UTC - in response to Message 74690.  

I believe they fit each result returned into a "not unreasonable result" filter. I believe that is defined by the:
<stream_only_likelihood2>
<search_likelihood2>
values returned for each result in the stderr.txt output. If the values are reasonable, then they don't need an extra validation task, if unreasonable the result goes into inconclusive status and warrants a wingman task.
Sounds good, so when I stick a dodgy GPU into my good host, it'll get spotted for the charlatan it is pretty quickly.
ID: 74692 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 8 May 09
Posts: 3339
Credit: 524,010,781
RAC: 0
Message 74693 - Posted: 20 Nov 2022, 22:16:55 UTC - in response to Message 74692.  

I believe they fit each result returned into a "not unreasonable result" filter. I believe that is defined by the:
<stream_only_likelihood2>
<search_likelihood2>
values returned for each result in the stderr.txt output. If the values are reasonable, then they don't need an extra validation task, if unreasonable the result goes into inconclusive status and warrants a wingman task.


Sounds good, so when I stick a dodgy GPU into my good host, it'll get spotted for the charlatan it is pretty quickly.


Yes it should and the whole host will will then be put on 'watch' again requiring wingmen until you get 10 valid tasks in a row
ID: 74693 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 990
Credit: 376,143,149
RAC: 0
Message 74694 - Posted: 20 Nov 2022, 22:23:26 UTC - in response to Message 74693.  

Yes it should and the whole host will will then be put on 'watch' again requiring wingmen until you get 10 valid tasks in a row
I buy broken cards and resurrect. You can get a busted one for a third of the price. Sometimes just a fan is broken, or the fan controller which you can bypass, or some heatsink gunk and pads (I use the super stuff, carbon nanotubes or something) make it work again.
ID: 74694 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Hurr1cane78

Send message
Joined: 7 May 14
Posts: 57
Credit: 206,540,646
RAC: 2
Message 74812 - Posted: 16 Dec 2022, 21:52:35 UTC

my boinc playlist gpu calculations all videos https://www.youtube.com/watch?v=pYVW9oA9yWk&list=PL_r97NUMjf7x4TgZT3bwQYL8KV_Vbo2J8
ID: 74812 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Septimus

Send message
Joined: 8 Nov 11
Posts: 205
Credit: 2,900,464
RAC: 0
Message 75204 - Posted: 25 Mar 2023, 16:09:50 UTC - in response to Message 72528.  

Is there a bit of a problem waiting for validation is steadily heading north to 700,000 units ?
ID: 75204 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Skillz

Send message
Joined: 28 May 17
Posts: 76
Credit: 4,398,910,125
RAC: 105
Message 75205 - Posted: 25 Mar 2023, 16:41:11 UTC - in response to Message 75204.  

Yes, the pendings started climbing at an alarming rate not too long ago.

https://grafana.kiska.pw/d/boinc/boinc?orgId=1&var-project=milkyway@home&from=now-7d&to=now&viewPanel=3

Even surprising the spike that happened back during the FB Sprint when one team dropped all their bunkers.

I'm currently sitting at nearly 60k pendings myself.
ID: 75205 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 990
Credit: 376,143,149
RAC: 0
Message 75206 - Posted: 25 Mar 2023, 16:44:36 UTC

Does it matter? Just keep crunching and let them pile up, somebody will sort it eventually.

There's always an enormous queue waiting on us, why get upset when there's a tiny queue waiting on them?
ID: 75206 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Septimus

Send message
Joined: 8 Nov 11
Posts: 205
Credit: 2,900,464
RAC: 0
Message 75207 - Posted: 25 Mar 2023, 16:57:49 UTC - in response to Message 75206.  

Does it matter? Just keep crunching and let them pile up, somebody will sort it eventually.

There's always an enormous queue waiting on us, why get upset when there's a tiny queue waiting on them?


No it probably doesn’t, just making an observation..
ID: 75207 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Skillz

Send message
Joined: 28 May 17
Posts: 76
Credit: 4,398,910,125
RAC: 105
Message 75208 - Posted: 25 Mar 2023, 17:07:19 UTC

If the admins aren't aware of the issue then the disk(s) will potentially fill up. When that happens no new tasks will be created and no tasks will be able to be uploaded. Thus, halting the progress of the project until it's sorted.
ID: 75208 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 990
Credit: 376,143,149
RAC: 0
Message 75209 - Posted: 25 Mar 2023, 17:22:00 UTC - in response to Message 75208.  

If the admins aren't aware of the issue then the disk(s) will potentially fill up. When that happens no new tasks will be created and no tasks will be able to be uploaded. Thus, halting the progress of the project until it's sorted.
Nothing we can do about it. If the admins are at work, they'll notice. Unless you know one is on vacation and you can tell him and he can fix it remotely.
ID: 75209 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 8 May 09
Posts: 3339
Credit: 524,010,781
RAC: 0
Message 75212 - Posted: 26 Mar 2023, 12:01:44 UTC - in response to Message 75209.  

If the admins aren't aware of the issue then the disk(s) will potentially fill up. When that happens no new tasks will be created and no tasks will be able to be uploaded. Thus, halting the progress of the project until it's sorted.


Nothing we can do about it. If the admins are at work, they'll notice. Unless you know one is on vacation and you can tell him and he can fix it remotely.


It usually takes a physical on site reboot of the Server to fix things, and they don't stop making new tasks, they just keep on making tasks even though the other Server is no longer validating tasks very fast.
ID: 75212 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 990
Credit: 376,143,149
RAC: 0
Message 75217 - Posted: 26 Mar 2023, 19:02:58 UTC - in response to Message 75212.  
Last modified: 26 Mar 2023, 19:03:51 UTC

If the admins aren't aware of the issue then the disk(s) will potentially fill up. When that happens no new tasks will be created and no tasks will be able to be uploaded. Thus, halting the progress of the project until it's sorted.
Nothing we can do about it. If the admins are at work, they'll notice. Unless you know one is on vacation and you can tell him and he can fix it remotely.
It usually takes a physical on site reboot of the Server to fix things, and they don't stop making new tasks, they just keep on making tasks even though the other Server is no longer validating tasks very fast.
I thought a remote reboot might be possible.

Presumably these tasks don't take up much disk space, if they do, they'd just queue up on our computers to send back.

It's not that many really, a similar number to how many we all have on our computers.
ID: 75217 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 8 May 09
Posts: 3339
Credit: 524,010,781
RAC: 0
Message 75222 - Posted: 27 Mar 2023, 10:38:06 UTC - in response to Message 75217.  

[quote]If the admins aren't aware of the issue then the disk(s) will potentially fill up. When that happens no new tasks will be created and no tasks will be able to be uploaded. Thus, halting the progress of the project until it's sorted.
Nothing we can do about it. If the admins are at work, they'll notice. Unless you know one is on vacation and you can tell him and he can fix it remotely.
It usually takes a physical on site reboot of the Server to fix things, and they don't stop making new tasks, they just keep on making tasks even though the other Server is no longer validating tasks very fast.


I thought a remote reboot might be possible.[/quite]

It should be but Tom the previous Admin said it almost never worked for him

Presumably these tasks don't take up much disk space, if they do, they'd just queue up on our computers to send back.

It's not that many really, a similar number to how many we all have on our computers.


Hopefully, though they did just release new N-Body runs and a couple people says they are taking days to run not hours.
ID: 75222 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 990
Credit: 376,143,149
RAC: 0
Message 75223 - Posted: 27 Mar 2023, 10:55:06 UTC - in response to Message 75222.  
Last modified: 27 Mar 2023, 10:55:41 UTC

I thought a remote reboot might be possible.[/quite]
Quite :-)

It should be but Tom the previous Admin said it almost never worked for him
Probably a lazy server which needed a size 10 boot. I wonder if that's where the term "booting a computer" came from.

Hopefully, though they did just release new N-Body runs and a couple people says they are taking days to run not hours.
14 minutes to 7 hours, last ones I did, February, Ryzen 3900XT 12 cores per task.

I wonder what the long ones mean. New way of calculating? Something interesting found?
ID: 75223 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Speedy51

Send message
Joined: 12 Jun 10
Posts: 57
Credit: 6,221,053
RAC: 3,752
Message 75235 - Posted: 28 Mar 2023, 21:50:46 UTC

I am more than happy to assist with processing any work I get especially tasks with _2 or more. Should I process a more tasks when my current work queue is clear or wait for the validation queue to drop?
ID: 75235 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 8 May 09
Posts: 3339
Credit: 524,010,781
RAC: 0
Message 75237 - Posted: 28 Mar 2023, 21:58:50 UTC - in response to Message 75223.  

I thought a remote reboot might be possible.[/quite]
Quite :-)

It should be but Tom the previous Admin said it almost never worked for him
Probably a lazy server which needed a size 10 boot. I wonder if that's where the term "booting a computer" came from.

Hopefully, though they did just release new N-Body runs and a couple people says they are taking days to run not hours.
14 minutes to 7 hours, last ones I did, February, Ryzen 3900XT 12 cores per task.

I wonder what the long ones mean. New way of calculating? Something interesting found?


Tom always said the beginning of the runs were taking longer calculate than the ends of the runs, don't know if that applies this time though.
ID: 75237 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 . . . 14 · 15 · 16 · 17 · 18 · Next

Message boards : Number crunching : Validation inconclusive

©2024 Astroinformatics Group