I have a Dell Precision T5500 workstation with Xeon X5690 (6 cores 12 threads). I used the computer with Rosetta for a few months achieve average points >9k. Recently I decided to buy a second CPU and upgrade RAM to 36GB. After installing the second CPU, I started processing 24 jobs at the same time, but my average points stay at nearly the same level. When I analyse points for each tasks they are more less 50% of the pervious achievements. So, it seems that I can process two times more tasks, but I spend 2x less computing power on each task. I upgraded BIOS and I tested two memory access models and it is still the same. Both CPUs work at 100% power, uptime shows >24, so from operating system point of view everything seems to be OK, but it is not. I did not change anything in the boinc/rosetta config files, because I could not find any place which I should change for dual CPI.
Has anybody found similar problem and how it was solved? Any help would be very appreciated.

Has anybody found similar problem and how it was solved? Any help would be very appreciated.

You need to run the BOINC Benchmarks as presently it is using the default values. BOINC Manager (Advanced view), Tools, Run CPU benchmarks.

Your Intel(R) Xeon(R) CPU X5690 @ 3.47GHz

Measured floating point speed 1000 million ops/sec
Measured integer speed        1000 million ops/sec

Your Intel(R) Xeon(R) CPU E5-2680 v2 @ 2.80GHz

Measured floating point speed 4552.11 million ops/sec
Measured integer speed       64769.29 million ops/sec

As they are used for working out Credit, it's resulting very little Credit being granted. Run the benchmarks, and a more appropriate amount of Credit will be awarded.

A quick check shows at least 4 or 5 of your systems are using the default values, so the Credit each of them is getting is much less than it should be.
Re-run the benchmarks on all of your systems to be sure they're all getting as much Credit as they should be.


Given that you are running multiple projects, i would suggest setting your cache to zero

Store at least            0.0 days of work
Store up to an additional 0.01 days of work

and using the default CPU runtime for Rosetta Tasks (which is 8 hours).
When project or system issues occur, or you make other changes to your settings, this will allow your Resource share settings to be honoured in a much shorter time frame (several days to a week) instead of the month(s) it would presently take.

---

Grant
Darwin NT

Thanks for your reply, I performed benchmark tests and I will see some results in two days. Target time in my case is 36h and I don't want to change, because I have to minimise network traffic, so at least now it has to stay as it is.