BOINC Pentathlon 2023

[Mxyzptlk]

3 hours ago, Blqckqut said:

why do you have so many nvidia gpus?

My passion is folding.... Nvidia is where it is at if you want to get the most bang for the buck in folding. 

[gsrcrxsi]

36 minutes ago, Mxyzptlk said:

My passion is folding.... Nvidia is where it is at if you want to get the most bang for the buck in folding. 

and BOINC

[TylerD321]

Is there anything special I need to know about Einstein before I switch to it tonight? 

[leadeater]

2 minutes ago, TylerD321 said:

Is there anything special I need to know about Einstein before I switch to it tonight? 

Only allow 'Gamma-ray pulsar binary search #1' to run

[Mxyzptlk]

1 hour ago, gsrcrxsi said:

and BOINC

F@H is my first love.... But I'll BOINC around when needed

[gsrcrxsi]

22 minutes ago, Mxyzptlk said:

F@H is my first love.... But I'll BOINC around when needed

I meant that Nvidia is where it is at if you want to get the most bang for the buck for BOINC also. most projects favor Nvidia. and some are CUDA only so can't be run on AMD anyway.

[SkillzTA]

I'm all in on Nvidia now also. Used to have a bunch of AMD 280X GPUs specifically for Milkyway@home due to their double-precision compute abilities, but I've since replaced them with much faster and more stable Nvidia GPUs with good DP. 

[Idiot_icLoser55]

Just want to reiterate what was said earlier after figuring some things out.

 

If you are running linux with an Nvidia gpu and set up the Einstein Special App as shown in the link below Gamma Ray #1 can receive a performance increase. For me this dropped around a minute from the runtime.

 

Before doing this your boinc instance must already be running the Gamma Ray #1 tasks. In my case I was having a couple problems getting Einstein to provide me with the correct tasks. It appears that einstein keeps a record of the tasks it assigns you. And changing your preferences doesn't do anything to this record. Until you complete these tasks they will just keep getting sent back to you even if you abort them. The only solution appears to be to run through them all until einstein starts assigning you new tasks.

 

Once it starts assigning you the new tasks you can then move onto the following instructions:

https://einsteinathome.org/content/einstein-fgrpb1g-linuxnvidia-special-app-aio

 

After completing those steps you need to restart the boinc client to allow it to select the new executable. Don't reset the project or else you will have to replace the executable again.

 

And as mentioned in the instructions if you are getting errors try using v0.95

Edited May 16, 2023 by Idiot_icLoser55
Had some details wrong.

[SkillzTA]

8 minutes ago, Idiot_icLoser55 said:

Just want to reiterate what was said earlier after figuring some things out.

 

If you are running linux with an Nvidia gpu and set up the Einstein Special App as shown in the link below Gamma Ray #1 can receive upwards of a 2x performance increase. For me the tasks have gone from completing in 20 minutes to 7 minutes.

 

Before doing this your boinc instance must already be running the Gamma Ray #1 tasks. In my case I was having a couple problems getting Einstein to provide me with the correct tasks. It appears that einstein keeps a record of the tasks it assigns you. And changing your preferences doesn't do anything to this record. Until you complete these tasks they will just keep getting sent back to you even if you abort them. The only solution appears to be to run through them all until einstein starts assigning you new tasks. Once it starts assigning you the new tasks you can then move onto the instructions below.

 

https://einsteinathome.org/content/einstein-fgrpb1g-linuxnvidia-special-app-aio

 

From 20 minutes to 7 minutes? That doesn't sound right.

 

I am assuming you mean this host: https://einsteinathome.org/host/13139955

 

Since it's the only host you have.

 

Looking at your completed tasks none of them have ran for 20 minutes.

 

You got on CPU task that ran for 41,669 tasks. Don't run CPUs, they're not worth it.

Then your highest running task on your GPU was the Multi-Directional Gravitational Wave task that took 732 seconds.

While your fastest tasks running on your GPU has been another Multi-Directional Gravitational Wave task that took 559 seconds.

 

The instructions you posted above doesn't effect any of those tasks.

Those tasks you have 30 of them that timed out - no response, and 31 of them that are "in progress" meaning they haven't been uploaded to the server and reported yet.

 

edit:

The 31 tasks you have in progress are the correct tasks however. So if some of them have completed already you should upload and report them, unless you're being sneaky with a bunker.

[Idiot_icLoser55]

51 minutes ago, SkillzTA said:

 

From 20 minutes to 7 minutes? That doesn't sound right.

 

I am assuming you mean this host: https://einsteinathome.org/host/13139955

 

Since it's the only host you have.

 

Looking at your completed tasks none of them have ran for 20 minutes.

 

You got on CPU task that ran for 41,669 tasks. Don't run CPUs, they're not worth it.

Then your highest running task on your GPU was the Multi-Directional Gravitational Wave task that took 732 seconds.

While your fastest tasks running on your GPU has been another Multi-Directional Gravitational Wave task that took 559 seconds.

 

The instructions you posted above doesn't effect any of those tasks.

Those tasks you have 30 of them that timed out - no response, and 31 of them that are "in progress" meaning they haven't been uploaded to the server and reported yet.

 

edit:

The 31 tasks you have in progress are the correct tasks however. So if some of them have completed already you should upload and report them, unless you're being sneaky with a bunker.

Just uploaded now but they're still pending.

 

Don't know that any of the originals ran to completion though cause I was fiddling with the settings in the middle of it and think they got aborted. 20 mins I think is about what it looked on track for and it's what they were estimating after I changed the settings. Guess it may have come down though if I'd let it run to completion.would have probably made sense to run a couple to completion so I had something concrete to compare. I'll remember that for next time.

 

Also my CPU's are off einstein now. Didn't know which would be best to begin with and had it backwards with GPU running on numbers and CPU on einstein. New to boinc and folding, and only just picking up on the fact that I should probably go and look at the top profiles to see which I should be running.

[SkillzTA]

3 minutes ago, Idiot_icLoser55 said:

Just uploaded now but they're still pending.

 

Don't know that any of the originals ran to completion though cause I was fiddling with the settings in the middle of it and think they got aborted. 20 mins I think is about what it looked on track for and it's what they were estimating after I changed the settings. Guess it may have come down though if I'd let it run to completion.would have probably made sense to run a couple to completion so I had something concrete to compare. I'll remember that for next time.

 

Also my CPU's are off einstein now. Didn't know which would be best to begin with and had it backwards with GPU running on numbers and CPU on einstein. New to boinc and folding, and only just picking up on the fact that I should probably go and look at the top profiles to see which I should be running.

 

The estimated time that the BOINC Manager reports is 99.9% of the time, incorrect. You will need to run 1000s of tasks on that host before the BOINC Manager can accurately calculate the remaining time because in the beginning its just guessing. Never base the time it takes to run a task on the estimated time. Let them finish, get the result file and see how long it actually took.

 

I just looked at the 12 tasks you uploaded that are pending. None of them are using the special app.

 

Here is one of the results that is pending you for.

https://einsteinathome.org/task/1470637670

 

Notice the top section:

<core_client_version>7.20.2</core_client_version><![CDATA[<stderr_txt>22:28:22 (816644): [normal]: This Einstein@home App was built at: Aug 17 2021 16:19:4022:28:22 (816644): [normal]: Start of BOINC application '../../projects/einstein.phys.uwm.edu/hsgamma_FGRPB1G_1.28_x86_64-pc-linux-gnu__FGRPopencl2Pup-nvidia'.

 

This tells you it is using the stock app still.

 

If it was using the special app it would look like this:

<core_client_version>7.16.6</core_client_version><![CDATA[<stderr_txt>02:19:04 (3367): [normal]: This Einstein@home App (v1.0 by petri33) was built at: Apr 28 2022 18:47:1502:19:04 (3367): [normal]: Start of BOINC application '../../projects/einstein.phys.uwm.edu/HSgammaPulsar_x86_64-pc-linux-gnu-opencl_v1.0'.

 

You can run nvidia-smi in the console/terminal/ssh/putty when a task is running and at the bottom it will tell you which app it is using.

 

 

[Chachunka]

A gentle reminder -- we're about to cross the finish line for the Sprint (NumberFields) in approx 20 minutes. Prepare to shift GPUs over to Einstein and CPUs back to yoyo. 

[Ithanul]

Good going team!

 

We snag third in Sprint!

[AamirA]

Should I do yoyo subproject ECM, ECM P2, or both?

[Idiot_icLoser55]

1 hour ago, SkillzTA said:

 

The estimated time that the BOINC Manager reports is 99.9% of the time, incorrect. You will need to run 1000s of tasks on that host before the BOINC Manager can accurately calculate the remaining time because in the beginning its just guessing. Never base the time it takes to run a task on the estimated time. Let them finish, get the result file and see how long it actually took.

 

I just looked at the 12 tasks you uploaded that are pending. None of them are using the special app.

 

Here is one of the results that is pending you for.

https://einsteinathome.org/task/1470637670

 

Notice the top section:

 

<core_client_version>7.20.2</core_client_version><![CDATA[<stderr_txt>22:28:22 (816644): [normal]: This Einstein@home App was built at: Aug 17 2021 16:19:4022:28:22 (816644): [normal]: Start of BOINC application '../../projects/einstein.phys.uwm.edu/hsgamma_FGRPB1G_1.28_x86_64-pc-linux-gnu__FGRPopencl2Pup-nvidia'.

 

 

This tells you it is using the stock app still.

 

If it was using the special app it would look like this:

 

<core_client_version>7.16.6</core_client_version><![CDATA[<stderr_txt>02:19:04 (3367): [normal]: This Einstein@home App (v1.0 by petri33) was built at: Apr 28 2022 18:47:1502:19:04 (3367): [normal]: Start of BOINC application '../../projects/einstein.phys.uwm.edu/HSgammaPulsar_x86_64-pc-linux-gnu-opencl_v1.0'.

 

 

You can run nvidia-smi in the console/terminal/ssh/putty when a task is running and at the bottom it will tell you which app it is using.

 

 

Thanks for this. Needed to restart the client before it would use the new executable. Brought the runtime down from ~450 to ~320. But left it only completing 1 in 6 tasks. Dropped it to v0.95 and now the runtimes are ~380 but there are no errors and it uses less resources.

[HaRdLy007]

I've moved my GPU's to Einstein. Not sure I want to fight with Yoyo tonight

[SkillzTA]

15 minutes ago, AamirA said:

Should I do yoyo subproject ECM, ECM P2, or both?

Probably ECM without knowing much about your host.

 

ECM can use up to 4GB of RAM per task while ECM P2 could go as high as 10GB, but I've only seen it reach 8GB. Still twice as much as non-P2 tasks.

 

Also keep note that the RAM fluctuates while the task is running. So you'll need to monitor the task and RAM usage for probably 30 mins to an hour to see it use the highest point.

[leadeater]

Anyone know the best Einstein task(s) for CPUs?

[SkillzTA]

None of them.

[leadeater]

1 minute ago, SkillzTA said:

None of them.

  Hide contents

 

I know but... since I have to for a little bit just to help pass [H]OCP soon as possible I'm going to do it lol. Not like we can gain on Yoyo anyway.

[Chachunka]

Now that three events are in the books, we have a pretty good idea of how the standings are stacking up. It's clear we are not in contention for overall silver, but need to aggressively defend overall bronze. SETI.Germany did a last minute push in the Sprint and got a good bump in the ranks. We only lead them by 27 points overall right now. 

 

Let's look at some hypothetical scenarios:

SETI.Germany ramps up in Einstein and takes our current 5th place standing. They gain 9 points and we lose 3. Net change: 12 points.

     If they can get 4th place, net change (assuming we get 6th place), 19 points. 

     Third place in Eintsein should be out of reach.

SETI.Germany beats [H] in their close race for 4th place in yoyo. They gain 7 points.

     If they get 3rd place, pushing us to 4th, net change is 27 points. (which would be a tie for 3rd overall only if they remain 8th in Einstein which is unlikely)

 

If they accomplish both 4th in Einstein and 4th in yoyo, we would still be able to hold 3rd overall by a slim 1 point margin. Therefore, we must maintain 3rd in yoyo to guarantee a 3rd overall finish. 

[SkillzTA]

I've actually never ran CPU tasks (intentionally) on E@H before so I honestly do not know.

 

You'd be better off taking the time to dual boot all the hosts you can and running the GPUs on Linux with the Special App. You'll nearly double your output.

[leadeater]

20 minutes ago, SkillzTA said:

You'd be better off taking the time to dual boot all the hosts you can and running the GPUs on Linux with the Special App. You'll nearly double your output.

If any of them actually had GPUs I would 

 

All the Windows servers are only that due to them being VMs on ESXi deployed via zero touch PXE other than the two 6242R servers which had OS's installed but have been sitting doing nothing for 1-2 years. I actually didn't know the Pent was on and joined after the second or third day, spend the next few days racking, adding ram sticks and installing Ubuntu on the actual physical not VM servers.

 

P.S. all my computers are public on every project, you'll see just how little GPUs I have 

[SkillzTA]

Ah, I assumed you had some GPU hosts running also. Well youve got time. Run CPU tasks on them, check how long they take to run and how many credits they are worth. Shouldn't be too hard for you to figure out which ones give the best PPD.

 

Link me to the host(s) after they've reported and validated tasks and I'll do the math for you if you need.

[leadeater]

8 minutes ago, SkillzTA said:

Ah, I assumed you had some GPU hosts running also. Well youve got time. Run CPU tasks on them, check how long they take to run and how many credits they are worth. Shouldn't be too hard for you to figure out which ones give the best PPD.

 

Link me to the host(s) after they've reported and validated tasks and I'll do the math for you if you need.

Sure, happy to be the first idiot to run Einstein on CPU during the Pentathlon. Plus I'm a little curious to see what I have is equivalent to GPU wise.