Now that the contests are done...

[Endgame124]

Where is the best place to put our processing power? I’ve got most of my hosts running Rosetta, 2 GPUs folding (though there still doesn’t appear to be enough gpu WUs), and I’ve got 1 old AMD card on Milky Way because it was only pulling like 30k ppd folding.

 

[ao_ika_red]

Wait for FB Sprint announcement, if you're interested in 4 days burst of fun*

Formula Boinc

Formula Boinc Championship results

formula-boinc.org

*it's addictive!

 

[lane42]

I might stay on Ibercivis for awhile for COVID-19 WORK.

 

[Endgame124]

I might stay on Ibercivis for awhile for COVID-19 WORK.

That’s the question I guess - put everything into 1 project? If so, which one? Or just spread it all around?

 

[lane42]

That's really up to you Eg

 

[TennesseeTony]

Yeah just run your personal preferrence of project between contests. We don't usually have a project of the month or anything. I like to look at my stats on Free-DC and see where I could put a bit of effort to get to the next 'milestone'. The contests are always bumping things up, that I was perfectly happy with before, lol.

 

[TennesseeTony]

Man, you are off to a good start!

Distributed Computing Stats from Free-DC

 

[Endgame124]

Man, you are off to a good start!

Distributed Computing Stats from Free-DC

I’m just getting started - most of that was just starting up old gear that was offline or mostly idle. Now that I have a reason to do so, I’ll be picking up much better hardware. I actually considered a threadripper, but 2 Ryzen 4950 (or whatever the model will be called) will be cheaper and will let me reuse the MB and ram from my current 2700X system.

The biggest issue I have at the moment is deciding which projects to support. I’m really not equipped to judge if covid research will be better done through Rosetta, WCG, or F@H.

 

[Assimilator1]

Probably none of us are, and we can't know with any certainty until sometime afterwards when papers begin to emerge.
For now you could just choose based on which have the most reliable supply of WUs. If they all are you could split your power up evenly (roughly).

 

[TennesseeTony]

I’m really not equipped to judge if covid research will be better done through Rosetta, WCG, or F@H.

Well 2 of those can't use GPU, so your GPU project is chosen for you. Now you are down to just choosing a CPU project.

If you run Rosy Red and WCG together, Rosetta has the shorter deadlines and will choke out WCG tasks until they near their deadlines.
An app_config.xml may be in order, using max_concurrent_tasks tag to limit Rosetta. Regarding the creation of config files, I am more of a copy/paste guy.
But if want to go that route, someone will create one for you, just ask.

 

[StefanR5R]

The biggest issue I have at the moment is deciding which projects to support. I’m really not equipped to judge if covid research will be better done through Rosetta, WCG, or F@H.

There are also TN-Grid and Ibercivis. However, TN-Grid has enough contributors relative to its limited pace of work generation, and Ibercivis is currently hampered by serious download failures while no admin is around for the time being.

If you run Rosy Red and WCG together, Rosetta has the shorter deadlines and will choke out WCG tasks until they near their deadlines.
An app_config.xml may be in order, using max_concurrent_tasks tag to limit Rosetta.

Attempts to control this just with project_max_concurrent (or worse, per-application max_concurrent) will eventually leave the computer at low utilization because the client has its own mind.

Perhaps it will work as desired if each application version is set to avg_ncpus=0.01 additionally. But I haven't tested this.

What always works is to use one dedicated boinc client instance for each project that is desired to be run at the same time.

 

[lane42]

As you can see here on ibercivis, over 500 errors,

Error tasks for computer 5114

 

[StefanR5R]

Many of these are not "errors" in the classic sense; they are "Cancelled by server".

This is working as designed: The server sends two or three tasks of a WU initially* (and further replica if tasks fail at the clients or time out at the clients), and as soon as at least one returned result was validated successfully, the server gives notifications to the remaining clients that their tasks are no longer needed for computation. (The server cannot notify the clients by itself, it needs these clients to send scheduler requests. It will hand these server-side cancellations out via its scheduler replies.)

Edit, the clients abort these tasks then if they weren't started yet. If they were started already, boinc is designed to keep them running and complete them, and the Ibercivis server will give credit if the returned results turn out valid too. I have read one or two complaints that server-side cancellations terminated tasks which were already running, but I have not seen actual evidence of this.

*) The admins made this choice because the project has not yet left its testing phase.

However, those tasks in the list you gave, which are "Error while downloading", failed due to a server-side mistake, as per the link in post #11.

Edit, the download errors make it practically impossible to run Ibercivis on many hosts for the time being (until an admin returns and fixes this). Nevertheless there are still several people able to keep the project running: https://www.boincstats.com/stats/187/user/list/12/0/0

 

[Howdy]

@lane42
Is Ibercivis still sending out the long R04 tasks?

 

[lane42]

Hi Howdy, Iam not sure now as I have stopped ibercivis for now
until they get things sorted out. I don't like seeing so many errors,
whatever the reason is. it just seems like wasted cycles.

 

[StefanR5R]

There are still some workunits left from the end-of-April R04 batch. But most WUs are from new batches now. Some people posted that they don't have download problems anymore.

 

[Fardringle]

I'm still getting a few scattered download errors, but most are fine now.