The Great Slow Down

Tigers_Dave
Tigers_Dave
Joined: 25 Mar 09
Posts: 231
Credit: 9,248,626,727
RAC: 0

mikey wrote:I'm getting no

mikey wrote:

I'm getting no tasks available for the GRP#1 gpu tasks so I had to go thru and setup a backup project just in case, I have a bunch of tasks, just over 2k in progress over multiple gpu's, but I'm hoping that there are enough tasks to keep them fed even if the cache is smaller than I'd like. My cache is currently set at 1 day plus 1/2 day extra across all my pc's.

mikey

 

Mikey, has the number of GRP#1 tasks you have in progress increased over the last two days?  I think it has increased from 4k to 9k for me, but I can't be certain.  In any event, I am going to try to grab some tasks for the rig that ran out of them earlier this week.  Will post the outcome to this thread.

"I was born in a small town, and I live in a small town." - John Mellencamp

mikey
mikey
Joined: 22 Jan 05
Posts: 12,558
Credit: 1,838,882,787
RAC: 24,500

mikey wrote: I'm getting no

mikey wrote:

I'm getting no tasks available for the GRP#1 gpu tasks so I had to go thru and setup a backup project just in case, I have a bunch of tasks, just over 2k in progress over multiple gpu's, but I'm hoping that there are enough tasks to keep them fed even if the cache is smaller than I'd like. My cache is currently set at 1 day plus 1/2 day extra across all my pc's.

mikey 

No in fact I'm down about 250 to 300 or so but that's only about 10% of my normal number, right now I have 2034 in progress right now

Tigers_Dave
Tigers_Dave
Joined: 25 Mar 09
Posts: 231
Credit: 9,248,626,727
RAC: 0

Mikey, my total pending tasks

Mikey, my total pending tasks went from 9889 to 9987 overnight, but that is less than 2 days of work for my crunchers.  Indeed, I am still unable to keep my fastest cruncher continuously fed.  Once it ran out of work, I suspended E@H and started crunching for AmicableNumbers@H.  I'll check on things next week; hopefully the situation has improved by then.

"I was born in a small town, and I live in a small town." - John Mellencamp

mikey
mikey
Joined: 22 Jan 05
Posts: 12,558
Credit: 1,838,882,787
RAC: 24,500

Tigers_Dave wrote: Mikey, my

Tigers_Dave wrote:

Mikey, my total pending tasks went from 9889 to 9987 overnight, but that is less than 2 days of work for my crunchers.  Indeed, I am still unable to keep my fastest cruncher continuously fed.  Once it ran out of work, I suspended E@H and started crunching for AmicableNumbers@H.  I'll check on things next week; hopefully the situation has improved by then. 

And of course you and other 'heavy hitters' that switch make it more likely the rest of us can soldier on with the GRP#1 tasks as we have been

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4,907
Credit: 18,462,846,412
RAC: 6,037,980

This is true, I am unable to

This is true, I am unable to keep my gpus fed with GR#1 tasks as the limited supply is exhausted too quickly by my hosts.  So I have to run BRP7 now as only they have adequate supply.

So more GR#1 tasks for everyone else with slower hosts.

 

Aurum
Aurum
Joined: 12 Jul 17
Posts: 77
Credit: 3,412,397,040
RAC: 61,617

Since I get hardly any OPNG

Since I get hardly any OPNG WUs I turned on Gamma-ray pulsar binary search #1 (GPU). At first I tried it in RZM but got nothing so I kicked it up to 100% and now I have a couple thousand WUs. I've turned GRP1 off now that I see there's a shortage. Are there classes of GPUs that can run GRP1 but not other other GPU WUs?

I never get any Gravitational Wave search WUs for either CPU or GPU. I recall a post that they'd run them all but a new grad student may decide to look again in a different way. The Applications page doesn't even list it, the Server Status doesn't show any and its work generator is disabled along with a couple of others. I keep them ticked on my Preferences just in case.

Tom M
Tom M
Joined: 2 Feb 06
Posts: 6,272
Credit: 8,961,136,237
RAC: 11,102,632

Brp7 aka MeerKat will run on

Brp7 aka MeerKat will run on windows Nvidia and Linux Nvidia gpus.

I think gpr#5 will run on Intel/and CPUs.

Your computers are hidden so it is difficult to be more specific.

Tom M

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)

Aurum
Aurum
Joined: 12 Jul 17
Posts: 77
Credit: 3,412,397,040
RAC: 61,617

Tom M wrote:Your computers

Tom M wrote:
Your computers are hidden

Not any more. Thanks for looking. I have no problem running Meerkat and there's plenty of them. My question was being curious why some are bemoaning the lack of GRP1s.

Tom M
Tom M
Joined: 2 Feb 06
Posts: 6,272
Credit: 8,961,136,237
RAC: 11,102,632

Aurum wrote: Tom M

Aurum wrote:

Tom M wrote:
Your computers are hidden

Not any more. Thanks for looking.

You have a very nice "farm".

While grp#1 tasks will usually run with 2 per gpu and produce more production I am less sure about "MeerKAT".  I know the custom version I am running shows very little gain at 2x per gpu.

The stock version ran at least 1x fine for my on my NVIDIA gpus. There is a good supply of MeerKAT so you can run as many as you want to download. 

One a single gpu system (rtx 3080 ti or lower) you should be able to run grp#1 tasks steadily.  I am uncertain about a 2 gpu gtx 1080 system however.

My current setup is I am running a pure grp#1 profile on my 3950x/rtx 3080ti.  On my Epyc/ 2 rtx 3080 ti system, I am running a gpr#1 selected with "other projects allowed with nothing else available" which gets me (mostly) a MeerKAT download with occasional gpr#1 tasks. So far when I try a pure gpr#1 profile it runs dry.

I could also enable some "0" resource projects for gpus which would keep them busy when grp#1 ran dry.

Tom M

 

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4,907
Credit: 18,462,846,412
RAC: 6,037,980

The Gravitational Wave

The Gravitational Wave campaign finished long ago.  There will not ever be any more work.

If you look at the server status page, you will see the task ready to sent count of the Gamma Ray tasks is zero constantly.  There is a severe shortage of work generation that has persisted for a month now because some backend servers that do pre-processing of the tasks did not survive the last main power outage the project suffered.

Slow running hosts do not have any issue maintaining a work cache of the GR#1 tasks.  Fast running hosts process work so fast that they exhaust their caches constantly and are never able to rebuild them. They constantly get no work is available because the work buffer is constantly emptied by all the slow running hosts before their hosts can hit the work buffer with work requests.

That is what the volunteers with fast crunchers are complaining about.  They have had to resort to running the BRP7 work since there is plenty of that available from the work buffer.

 

Comment viewing options

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