Hi Gang
I've been BOINC'ing for a few years now, my first was SETI back in 1999. I like feeling that I am contributing in my own small way to the bettement of the race.
I've been Rosetta'ing for some time as well, but I'm afraid I Will be ceasing as of the end of the month. The reason: credits...
I crunch for other projects and get oodles of credit, (am trying to reach 2 mill on each project), but I crunch a block for Rosetta for 6 hours or more and get ...200.
I *know* it's petty, but I feel like I am getting nowhere, so I will go back to World Community, CPDN (Lonnng units!) or perhaps more SETI.
Later Dudes

BobTheBrit

---

Bob, looking at your results, it does seem your machine is not being granted much credit for the time being used. The machine reports in with 12 CPUs and 12GB of memory on an i7. I see some tasks undergoing many restarts, and erratic runtimes between tasks.

Perhaps if you could tells us a bit more about how you use the machine, we could point you to some preferences that better fit your usage. Does the machine run 24hrs a day? Do you run more than one BOINC project? Does the machine have a heavy non-BOINC workload? How are your preferences set for how much memory to allow BOINC to use, and whether to keep tasks in memory?

I'm thinking the machine may be thrashing, where perhaps you are allowing BOINC to try and use so much memory that it has so much work going at the same time that it is conflicting with itself. Are you seeing heavy disk access to the swap file?

Rosetta does use more memory per active task than many other BOINC projects. If you have all 12 cores actively running only Rosetta work units, it might cause memory contention like this. I'd suggest that you add WCG with a 20-30% resource share. Their tasks tend to use much less memory, so having a few of those sprinkled in will help reduce memory contention.

Whatever details you provide and actions you take, the goal would be to see your claimed credit per task and your granted credit to be much closer. It appears currently you are being granted 30-40% of the credit claimed. This implies your machine did not get as much work done in the time period as the benchmarks would have predicted (perhaps because of memory contention).

---

Rosetta Moderator: Mod.Sense

Hi
I have only been running rosetta on this machine for a couple of weeks. It runs 24/7 with very little else going on. Basically it sits and crunches and when I come home from work it may do a little light Steam Gaming. I had inteneded it to be my home NAS box, but got a Synology instead :). Previously I ran on an older I5 box but had the same issues there (ie, low credit return for time spent.

Any suggestions gratefully received. It seems fine with WCG, Seti et al, just rosetta having an issue
I do normally run each Project by itself, (on theory that the more power you give a project the better), but I will now change my prefs to allow WCG to run at the same time.

My memory prefs are currently set to:
in use: 40%
not in use 60%
The leave in memory box is unticked

I did try to cut n paste but for whatever reason CTRL V didn't want to play!

It will be a couple days before Milkyway drains, then I'll start WCG and rosetta.

Talk Soon

Bob The Brit

---

OK, yes, R@h memory requirement is likely why you see low productivity on your machine. Many people comment about how R@h gives less credit per hour than other projects, but in your case, you're getting significantly less than the normal for R@h.

If the machine really is not used for anything when you are not sitting at it, I'd suggest allowing 100% of memory when not in use, and about 70% when in use (reduce if it impacts your gaming reflexes at all). Also, as you sit down and begin to use the machine, BOINC suddenly finds itself using more memory than your preference and so halts work in progress. When this happens, since you've told it not to keep tasks in memory, it will basically terminate tasks until it is living within the memory preference. And these will later have to restart from their last checkpoint. I saw a few of your tasks that were restarted from checkpoints multiple times, and on a machine running 24/7, with one project, that really shouldn't happen much. So, I'd suggest checking the box to leave tasks in memory when suspended. The phrase "in memory" is vague, it really means "virtual memory". The tasks will remain active, but get swapped out, and can then restart without loss of the current work since the last checkpoint.

Doing the above, with a 20-30% resource share to a lower memory project such as WCG, your credit granted per task and per hour of active time should improve nicely. Please post again in a week or so when you have had some time to observe the effects of the changes.

---

Rosetta Moderator: Mod.Sense

I'd reduce the time given to each WU and/or reduce the number of threads running simultaneously (there are only 6 real cores in that CPU after all). Running with all available threads greatly reduces the credits/WU but AFAIK increases overall RAC... at least in dual cores with HT. Not sure if this RAC increase scales to 12 consecutive threads running at the same time (I'd imagine RAM latency and access becomes a bigger bottleneck with 12 threads than with 4 threads). Rosetta is very CPU cache and RAM-heavy compared to say SETI. With 12GB + 12 threads, you're on the "recommended minimum" RAM requirement for Rosetta (1GB per thread). So reducing thread count could alleviate the other RAM-hungry WUs running in parallel.

Maybe reduce to 10 threads instead of 12? (that'd be 82% on BOINC settings).
Leave the "leave in memory" tick checked, so that when you play or suspend BOINC for any other reason, whenever you resume the WU will start right from where they were suspended (if this box is not checked, resuming will start from the last checkpoint... not good... wasted time).

Also, allow as much RAM when not-in-use as possible (90%).
In-use memory could also be increased to 60%. Maybe some WUs are left on "waiting for memory" when you're just browsing the net.
Consider maybe getting more RAM, it's dirt cheap and easy to install. I have 16GB on my 4 thread laptop for instance. 12GB is plenty, but maybe not so for 12 simultaneous threads of Rosetta WUs.

---

Just glancing through your latest completed results, it looks like they are being granted somthing closer to 75% of the claimed credit. So, that is roughly double what you used to be getting, so whatever settings you've chosen to use seem to be making a noticeable improvement.

---

Rosetta Moderator: Mod.Sense