2018 SETI-Wow! Event

[Gilthanis]

The reason I want to start this thread in here early is to suggest a proposition. A splash of sorts. The event is on August 15th. Typically many people start bunkering 10 days in advanced because the client limits this. However, IIRC SETI work units have a deadline of around 53 days. I was wondering if the team wanted to do an insane major bunkering this year. Obviously, this would mean running multiple clients per host, or having multiple OS installs (swappable hard drives?), VM's for those running on CPU only boxes, etc...

If people started tonight, they would have ~49'ish days of work to process...

Thoughts?

 

[motqalden]

Sounds like fun.

 

[Gilthanis]

I also realize this is during a time when most people are powered down for the summer and it isn't the largest event due to both this and the project choice. I for one typically only run SETI during this event and usually half assed. I am in the process of restructuring my rigs and so may not be able to push it a ton but figured for those ready to go, this would be a great opportunity.

atp1916 , is SETI one of the projects your rigs like? Are you willing to run a single project that long? Would be an incredible showing if the team did. I'm going to try and get focused this weekend and get a box or two done. Then I have to get some DDR4 still. Preferring 64GB worth but that is insanely priced still. Seen a few decent deals on Reg ECC but was really wanting to avoid that so that I could re-purpose in the future if needed...

Also, if you download your work now... the odds of being verified by event time should be pretty good....lol

 

[ChristianVirtual]

But then hiding computers might be a good idea. Else it will be too obvious.

Also what is the risk that “server cancelled” during that time. Never understand when they do it. But yeah, statistically many WU should make 8t and it sure is a great boot camp for bunker practice. I learned my ropes for bunkering last year with wow installing multiple clients at the same time in one sys5em (no VM, because also of GPU).

 

[Gilthanis]

Server cancelled typically means that they didn't need it. Like if you got the work unit due to someone else missing deadline but then completing before you or the admin deletes a batch for some reason. I think the latter is usually labeled server aborted but I don't recall. It is possible that some will but the over all payout of having a crap ton of completed bunker work is the advantage. I know there are times that I have work units waiting to validate for a few months...so... not sure.

 

[motqalden]

So i am going to basically boot up 4 VB per computer and download 10 days worth on each of them. Then i will turn them on one at a time (with hostfile or whatever to block upload) let them crunch untill done all tasks, turn off and then turn on the next one. Should be pretty simple this way and only need 1 VM active per machine at a time so no ram limitations.

 

[atp1916]

Gilthanis Good question about performance on SETI. It was the first project in the FB sprint season, and i know i had gimpy Celerons at the time... So, i'll retest.

However, at this time - i am pretty booked on RL stuff. Can't really handle much more than the FBs.

 

[Gilthanis]

Not a problem. I'm just tossing out ideas.

 

[motqalden]

Testing this out on 1 VB. Seems like it will need babysitting. I downloaded a bunch of work units with Aug 18th deadline, but also some with July 17th. This means i either need to abort / weed those ones out or pause the longer ones and push through / finish and upload the july ones...

 

[pututu]

I learned my ropes for bunkering last year with wow installing multiple clients

I would prefer running optimized GPU apps in linux which I think runs the fastest using multiple clients installation approach. Any quick guide on multiple setup on linux? I only know of how to do this in Windows. Thanks!

In case anyone is looking for the SETI optimized apps in linux, see post here.

 

[pututu]

I just PMed fastgeek to see if he is still in town to help us in the sprint.

 

[motqalden]

Another day or two and i will have ~ 1000 cpu tasks saved up. kinda sucks that i can only get about 100 tasks before it won't give me any more.. (setting config with ncpu=196 didn't make any difference)
. i'm gonna have to buy another HDD to give me more VM space. CPU is not terribly productive for this but its much easier to stockpile as i can't run GPU in VM and need to dual boot or something. I think i will crunch CPU till about 10 days out and then see about bunkering my GPU work at that point. either way i should have a nice stockpile come contest time me thinks.

 

[Gilthanis]

If you aren't able to pull any more, then it may be a project limitation on a per host basis. Hoping that it separates the number between cpu and gpu. I've not even started yet though I was really wanting to. Re-purposing and tearing down of the rigs keeps running into oops scenarios....

 

[motqalden]

If you aren't able to pull any more, then it may be a project limitation on a per host basis. Hoping that it separates the number between cpu and gpu. I've not even started yet though I was really wanting to. Re-purposing and tearing down of the rigs keeps running into oops scenarios....

Yeah the table says task limit of 200 so that must be gpu/cpu spit

 

[phoenicis]

Another day or two and i will have ~ 1000 cpu tasks saved up. kinda sucks that i can only get about 100 tasks before it won't give me any more.. (setting config with ncpu=196 didn't make any difference)
. i'm gonna have to buy another HDD to give me more VM space. CPU is not terribly productive for this but its much easier to stockpile as i can't run GPU in VM and need to dual boot or something. I think i will crunch CPU till about 10 days out and then see about bunkering my GPU work at that point. either way i should have a nice stockpile come contest time me thinks.

Additional boinc client instances may be the way to go on this one. They allow you to crunch both cpu/gpu and are less resource hungry than VMs.

I looked into this following the comment from ChristianVirtual on his success using this approach last year. The guide I used was written by OCN's mmonnin and tictoc here.

According to TAAT's Stefan the only currently active project he's come across that it can't be used for is GPUGrid.

Screenshot is a Threadripper with 3 boinc instances currently brewing an NFS bunker. The only gpu project I've tested so far is Primegrid. It seemed to work fine.

 

[motqalden]

Additional boinc client instances may be the way to go on this one. They allow you to crunch both cpu/gpu and are less resource hungry than VMs.

I looked into this following the comment from ChristianVirtual on his success using this approach last year. The guide I used was written by OCN's mmonnin and tictoc here.

According to TAAT's Stefan the only currently active project he's come across that it can't be used for is GPUGrid.

Screenshot is a Threadripper with 3 boinc instances currently brewing an NFS bunker. The only gpu project I've tested so far is Primegrid. It seemed to work fine.

View attachment 87074

would this be basically the same setup as the multiboinc for goofygrid but with less folders?

 

[phoenicis]

would this be basically the same setup as the multiboinc for goofygrid but with less folders?

I just had a quick look at the goofygrid guide and it would seem so as far as Windows is concerned. I'm probably missing the obvious but I'm not sure how the unique port per instance is established using the guide.

 

[motqalden]

i think its the last part of the start command where it uses a variable?
-gui_rpc_port 50%X

 

[atp1916]

Gentlemen (and ladies )

What is the need for HDD space for this?

I got a bunch of spare laptop HDD spinners (most all >500GB) and smaller SSDs (60-64GB) that i am trying to figure out what to do with....

 

[Gilthanis]

i think its the last part of the start command where it uses a variable?
-gui_rpc_port 50%X

To be fair, my guide (based off of Digi-koz) is the basis mmonnin used for refining the guide they have now. At one time he credited it by mentioning it was based off of a guide over at [H] but in typical mmonnin fashion, he stopped citing his sources. He is the type of guy that likes taking credit for things that he picks up elsewhere. Has to be "the go to person" or gets butt hurt. I believe his last visit here was to call me a prick because I corrected him once about RALPH@home not being Android only. Anyways, it should be pretty similar in conceptual use at other projects. They have their guide a lot more polished and I stopped messing with multi clients outside of goofy.

 

[motqalden]

got my second boinc instance working. this will definitely save me HDD space and time!

 

[Gilthanis]

Can someone test this in a cc_config

<coproc>
<type>some_name</type>
<count>1</count>
<device_nums>0 2</device_nums>
[ <peak_flops>1e10</peak_flops> ]
[ <non_gpu/> ]
</coproc>

Allegedly it lets you edit the number of GPU's that the BOINC client reports similar to the N CPU's option. I've not toyed with it yet and will not be able to for a while. I saw it posted here https://forums.anandtech.com/threads/seti-home-wow-event-2018.2550478/page-2#post-39498532

If it does allow you to lie about the number or GPU's it may be a way to download more work units than the max 100 per card.

 

[pututu]

Interesting. BIONIC wiki has further explanation about the options. Like ncpus, probably once you have enough WUs downloaded, might need to switch the device count to 1, if you have 1 GPU.

<coproc>
specify a coprocessor. Used in combination with the Anonymous platform mechanism. The element has the form
<coproc>
<type>some_name</type>
<count>1</count>
<device_nums>0 2</device_nums>
[ <peak_flops>1e10</peak_flops> ]
[ <non_gpu/> ]
</coproc>

The name given in <type> must match that in the <coproc> element in app_info.xml.
<count> in the number of GPU instances, and <device_nums> is their device numbers. <peak_flops> is the GPU peak FLOPS; it can be omitted if your app_info.xml specifies estimated application FLOPS. If <non_gpu/> is specified, the coprocessor is not treated as a GPU; i.e. "Suspend GPU" does not affect it. This mechanism has two purposes: to provide fine-grained control of the coprocessors recognized by BOINC (NVIDIA, AMD, and Intel GPUs), and to let you use coprocessors not :recognized by BOINC. Requires a client restart.

 

[Gilthanis]

I'm thinking that this years Wow event can be used for testing strategy options for 2019 Pentathlon/WowEvent etc...

 

[motqalden]

I will try this out when i get home from work.

 

[motqalden]

State: All (3291) · In progress (3288) · Validation pending (1) · Validation inconclusive (0) · Valid (0) · Invalid (0) · Error (2)

These are all CPU so far and i am done all but 200 of the "in progress" little less than a month left but i should have a nice bunker.

 

[auntjemima]

Not to look like a complete idiot, but this entire event is only seti@home, correct?

 

[auntjemima]

What is the update interval for this project for a script?

edit: seems to be about 5 minutes. Might as well just let boinc update this one itself.

 

[motqalden]

What is the update interval for this project for a script?

edit: seems to be about 5 minutes. Might as well just let boinc update this one itself.

You do not need a script and there seems to be no reason to trick it into thinking you have "X" CPU's From what i have read and can tell it limits you to 100 tasks total per Boinc instance (CPU tasks) and this is really simple to download in a few minutes. I believe if you enable GPU it will allow an additional 100 GPU tasks per instance, but i have not confirmed this. because 100 is not alot of tasks you can run multiple VM's or more simply multiple instances of BOINC as described here: http://www.overclock.net/forum/1805...uide-setting-up-multiple-boinc-instances.html

I started out doing the VM thing but quickly realized it was gonna take way way to much HDD space up, where the multiboinc thing only needs the space for the core project files and the tasks you download. on my 16 thread machine i find it can crunch approx 200 tasks a day, so to make so that i dont' have to keep setting up more of them throughout the day i run 2 or thee instances at once and limit / spread the available threads between them. IE 2 instances with 7 threads each. This way it will take them twice as long to get through the pile of tasks before i have to make more of them. Even if you don't do the Seti wow thing i highly recommend learning the multiboinc because it can be so useful in a lot of situations.

Yes the entire event is Seti only.

 

[tayunz]

ChristianVirtual If you don't mind sharing, what does your setup for Seti look like? Was just wondering because you were getting pretty good output from what I can tell.

 

[auntjemima]

You do not need a script and there seems to be no reason to trick it into thinking you have "X" CPU's From what i have read and can tell it limits you to 100 tasks total per Boinc instance (CPU tasks) and this is really simple to download in a few minutes. I believe if you enable GPU it will allow an additional 100 GPU tasks per instance, but i have not confirmed this. because 100 is not alot of tasks you can run multiple VM's or more simply multiple instances of BOINC as described here: http://www.overclock.net/forum/1805...uide-setting-up-multiple-boinc-instances.html

I started out doing the VM thing but quickly realized it was gonna take way way to much HDD space up, where the multiboinc thing only needs the space for the core project files and the tasks you download. on my 16 thread machine i find it can crunch approx 200 tasks a day, so to make so that i dont' have to keep setting up more of them throughout the day i run 2 or thee instances at once and limit / spread the available threads between them. IE 2 instances with 7 threads each. This way it will take them twice as long to get through the pile of tasks before i have to make more of them. Even if you don't do the Seti wow thing i highly recommend learning the multiboinc because it can be so useful in a lot of situations.

Yes the entire event is Seti only.

This is exactly what I was looking for, thanks!

 

[ChristianVirtual]

ChristianVirtual If you don't mind sharing, what does your setup for Seti look like? Was just wondering because you were getting pretty good output from what I can tell.

as simple as 1x980Ti and 1x1080Ti on Linux; but the "special source" CUDA 9.0 version of the software. not the stock cuda6 version.

https://arkayn.us/forum/index.php?P...048d7fe046d1d&topic=197.msg4529;topicseen#new

That really makes all the difference; 2 min each WU; bunker really becomes work.

(thats why I have the "lazy-bunker" right now; just keep it running an hope no wingman comes down the line; every wingman is then counted as good for the FB marathon). Just not have time right now to bunker too much.

 

[tayunz]

as simple as 1x980Ti and 1x1080Ti on Linux; but the "special source" CUDA 9.0 version of the software. not the stock cuda6 version.

https://arkayn.us/forum/index.php?P...048d7fe046d1d&topic=197.msg4529;topicseen#new

That really makes all the difference; 2 min each WU; bunker really becomes work.

(thats why I have the "lazy-bunker" right now; just keep it running an hope no wingman comes down the line; every wingman is then counted as good for the FB marathon). Just not have time right now to bunker too much.

Oh okay thanks, I was wondering. Awesome!

 

[Gilthanis]

I won't have a bunker this year. However, if things go as planned I will be a lot more active next year with the sprints. This year has sucked pretty bad in the happenings of real life.

 

[auntjemima]

You do not need a script and there seems to be no reason to trick it into thinking you have "X" CPU's From what i have read and can tell it limits you to 100 tasks total per Boinc instance (CPU tasks) and this is really simple to download in a few minutes. I believe if you enable GPU it will allow an additional 100 GPU tasks per instance, but i have not confirmed this. because 100 is not alot of tasks you can run multiple VM's or more simply multiple instances of BOINC as described here: http://www.overclock.net/forum/1805...uide-setting-up-multiple-boinc-instances.html

I started out doing the VM thing but quickly realized it was gonna take way way to much HDD space up, where the multiboinc thing only needs the space for the core project files and the tasks you download. on my 16 thread machine i find it can crunch approx 200 tasks a day, so to make so that i dont' have to keep setting up more of them throughout the day i run 2 or thee instances at once and limit / spread the available threads between them. IE 2 instances with 7 threads each. This way it will take them twice as long to get through the pile of tasks before i have to make more of them. Even if you don't do the Seti wow thing i highly recommend learning the multiboinc because it can be so useful in a lot of situations.

Yes the entire event is Seti only.

Oh man, this was a bit frustrating, only because I wanted to change the password to what I use for my other BoincTask instances. No worky. Even when changing them in the auth cfg. I removed all the files in boinc2 again and tried connecting with the generic long password and it worked. Now to wait for seti to come back online.

 

[auntjemima]

You do not need a script and there seems to be no reason to trick it into thinking you have "X" CPU's From what i have read and can tell it limits you to 100 tasks total per Boinc instance (CPU tasks) and this is really simple to download in a few minutes. I believe if you enable GPU it will allow an additional 100 GPU tasks per instance, but i have not confirmed this. because 100 is not alot of tasks you can run multiple VM's or more simply multiple instances of BOINC as described here: http://www.overclock.net/forum/1805...uide-setting-up-multiple-boinc-instances.html

I started out doing the VM thing but quickly realized it was gonna take way way to much HDD space up, where the multiboinc thing only needs the space for the core project files and the tasks you download. on my 16 thread machine i find it can crunch approx 200 tasks a day, so to make so that i dont' have to keep setting up more of them throughout the day i run 2 or thee instances at once and limit / spread the available threads between them. IE 2 instances with 7 threads each. This way it will take them twice as long to get through the pile of tasks before i have to make more of them. Even if you don't do the Seti wow thing i highly recommend learning the multiboinc because it can be so useful in a lot of situations.

Yes the entire event is Seti only.

Not to keep quoting this, but each new instance is only using my cpu's, but not my gpu. Is that normal for only one instance to be able to use it? Like, seti@home isnt even trying to grab gpu tasks on my 9 other instances.

 

[motqalden]

Not to keep quoting this, but each new instance is only using my cpu's, but not my gpu. Is that normal for only one instance to be able to use it? Like, seti@home isnt even trying to grab gpu tasks on my 9 other instances.

Are you sure you have GPU enabled in your preferences for the project and also in the boinc manager? If you are using a task manager to connect, maybe try making a shortcut and seeing "whats up"

Edit: Also i have not tried GPU tasks with this so far, but i would assume it works fine

 

[auntjemima]

Are you sure you have GPU enabled in your preferences for the project and also in the boinc manager? If you are using a task manager to connect, maybe try making a shortcut and seeing "whats up"

Edit: Also i have not tried GPU tasks with this so far, but i would assume it works fine

On my initial instance of BOINC I am crunching gpu tasks. No issues there, but I am not on the new instances created today, ie: BOINC2, BOINC3, etc

I am using Boinc Tasks to run the instances. I don't see an option for gpu there, besides "use gpu". Which is selected.

 

[auntjemima]

When running a shortcut for one instances I get "No usable gpu's found" in the eventlog. weird.

 

[motqalden]

yeah i really don't know on this one. Maybe someone else can chime in?