@wolfman1360

I'm not sure what is going on. First things first on another of your computers, your Core 2 8400 is requesting work and all tasks are failing. The stderr on the failed task webpages states a missing library. Try to reinstall the needed compatibility libraries for that version of Ubuntu.

On the i7 4770, the benchmark scores are 1 billion ops/sec for both floating point and integer. While this would not be causing your work fetch problem, it would give ridiculously large estimates for time to task completion in boinc manager. Run a boinc benchmark when you get a chance.

How many other projects are the two work fetch problem computers attached to? And did you attach to the other projects before cpdn?

Perhaps suspend all other projects in boinc manager and then allow boinc to try to request work from cpdn. I'm not sure if that will do anything, but it's worth a try.

One more question - I remember seeing something about figuring out the actual computing speed by something like t/sec, but not sure where that's located or if it's still something to consider. E.g. 20 is faster than 30, I think.
I could be mistaken and I forget the specifics of what the T stands for. Surely not trickles. Maybe the steps per second or something.

The speed can be represented in units of seconds/timestep (sec/TS) and after trickles are uploaded, can be seen on each task's webpage. The lower the average number of sec/TS, the relatively faster the model is running and the less CPU time a completed model will take.

One can also see the sec/TS on running models by going into the .../projects/climateprediction.net/{task name} directory and looking at the file stdout_mon.txt, which is a log of the timesteps throughout the model run. In Linux, one can be in that directory and do a

tail -f stdout_mon.txt

and it will output a display to the terminal window continuously as the model runs. Depending on the Linux distribution and how it handles permissions for that directory, one might need to be a superuser to maneuver to that directory and tail that file.

Edit...For the same PC, the value of the sec/TS for a given model will be dependent on how complex one model type may be relative to another. So for the same PC, the sec/TS for a hadam4 N144 model will be lower than the sec/TS for a hadam4 N216 model which is run at a higher resolution.

Random question because I have forgotten over the years.
Each month is a trickle, so a 12 month task would give 12 total thus 12 bits of credit (when the credit is actually calculated and updated weekly)?
thanks!

Yes though the size of the bits of credit varies depending on the amount of computing that needs to go into it. Credit is currently updated on Thursdays but that has been a moveable feast over the years. At some point Andy plans to introduce a credit script that needs less work by the server running it at which point we should move to daily but there is no news of a date for that to happen and it is something he works on when he has spare work hours rather than being a priority task. I am pretty sure he could have sorted it by now had that been the case.

0% after 4 days means something is seriously wrong. Have you checked how much CPU it is using in Task Manager? I suspect that whatever BOINC is telling you it isn't actually running.

@Wolfman1360

I vaguely remember discussion of Rosetta eating up l3 cache as well, but can't find the discussion anywhere.
Is this still true today and should I be limiting it alongside the n216 and n144?

Jim1348 has referred to local threads where this has come up; if you look in the threads about UK Met Office HadAM4 at N216 resolution and UK Met Office HadAM4 at N144 resolution you'll find several mentions of L3 cache bashing (especially in the N216 thread, but in this message in the N144 thread I actually replied to one of your posts, talking about workload mixes (and again in this message)... Jim1348 (and others) had some good contributions in those threads too. I don't recall many explicit references to Rosetta, but WCG MIP1 (which uses Rosetta) got some dishonourable mentions...

You may also have seen (or even participated in) threads about MIP1 at WCG -- because of the model construction it uses, the rule of thumb is that one MIP1 per 4 or 5 MB of L3 cache! I haven't got time to track those down at the moment - sorry!

For what it's worth, if you run MIP1 alongside N216 you'll see the same sort of hit as if running extra N216 tasks; N144 is nowhere near as bad!

Cheers - Al.

[Edited to fix a broken link, then to fix a typo I'd missed!]

And on my i7-9700 (which has eight full cores), it checkpoints at 23 minutes. But that is again with limiting the N216 to running on only four cores. The other four cores are on TN-Grid, which seems to be an easy project for this purpose.

In general, I find that I need to limit any of my CPUs (Intel coffee lake or Ryzen) to four cores for the N216, but can put just about anything else on the other cores without much ill effect. Beyond four cores, it drops off a cliff.

It's not really about how BOINC sees it. It is complicated to answer because it has to do with how the kernel's process scheduler and memory management more data into and out of RAM to and from the CPUs. The kernel can only get data in/out of RAM at the speed of the RAM and it can only move data between CPU cores at the speed of the CPU die interconnect and it can only move data between CPU sockets at the speed of the motherboard's North Bridge controller.

My advice is to start with less and work up to more until you find a sweet spot.

This definitely isn't going to be something I'll be using full time or even for years to come.

Oops, I made the wrong assumption. If it's not for purpose of crunching 24/7, ignore what I said. :-P EPYC is just an example for power difference. I don't mean to add that to the choices here anyway.

Then latest generation stuff makes no sense in this case. Release of Ryzen 3 dropped Ryzen 2 to a much lower price, like <$150 during holiday season for a new 2700X. You can probably get really good deals on used ones soon. A used Ryzen 2 is likely something you want to consider instead, even if you want something decently recent. Since you have all other parts already, I would go with E5. E5 2670 should be quite better than Opteron 6128.

The office. Now that's a fantastic idea I'll have to consider....maybe this can be written off as a work expense?

LOL. I just realized "the office" is not from your reply...

I would say neither. I grabbed dual E5-2670 when they were dirt cheap three years ago. That 700W system was less productive than my EPYC 7401P system which is like 250W and I've recycled those E5s last year. Ryzen 3 is really awesome in terms of power efficiency. It looks a bit pricer up front, but unless your electricity is free, a 400-500W saving running 24/7 will make up for it in a few month. (Though I guess you are probably not paying power bill for your office, which is nice. :-P) In addition, if you don't have the memory already, eight DDR3-1333 RDIMMs needed to populate all channels for dual E5-2670 is going to cost more than the two DDR4-3600 sticks for Ryzen 3.

While I would really like to test how well my CPUs can cope with multiple N216's

Your Ryzen 3700x will work fine if you limit the N216 to four at a time.

Your main concern (and your children's next seven generations' concern) will be "Performance per Watt"

Though this time of year, I wouldn't nmind a little extra heating in my office!