My P4 2.66 PC has been running for nearly a week after joining and my average credit is now 54. I would like to know what kind of system is producing thousands of average credits? I don't see that this is possible unless you’re reporting a composite from all runtimes of Rosetta on a PC. This assumption is based on reading that every CPU core produces a runtime of Rosetta. If correct all you are reporting is the sum of all of the runtimes that are in one PC which makes it impossible to determine or compare a PC since more processors running more runtimes is the same as having more PCs.

To really see how fast a PC is you would have to see how many average credits it produces per runtime of Rosetta. If you would post the number of runtimes divided in to the PC’s average credit then anyone could see how their PC truly compares to a large multi-processor/multi-core server.

If you crunch 24/7 you get a higher RAC and is a measure of your throughput in time.

Looking at your settings you crunch 24hr runtime and get around 160 credits for that, which is approximatly where your RAC would settle to if you crunched 24/7 over a few weeks.

If you only crunch fro half that time you would get half the RAC.

The very high RAC machines are often the 2,4 or 8 core/cpu configurations.

It is a 'Computer as a whole' score.

The P4 2.6 is very week compared to the newer CPUs.

---

Team mauisun.org

Hi Benito

I think you're assuming that all 'runtimes' - i.e. tasks (as shown under the tasks tab in BOINC Manager) require different amounts of processing. Each task shown consists of a number of 'decoys', and these decoys don't all require the same amount of processing. For example, some of the protein-protein docking jobs have been very computationally intensive, whereas others have been much quicker.

Rosetta will process x number of decoys within a given task, with each starting from a different point (so the same work isn't being repeated), and will continue running decoys until it reaches your 'Target CPU run time' as set in your rosetta preferences. So, if you've set your preferences to 8hrs then your computer might manage 20 decoys, whereas the the same machine will only manage approximately half of these if your preference is half that time (a longer time setting is better for the project as there is less communications and bandwidth to deal with and slightly better for your RAC).

Your computer then submits its benchmark score along with its results, from which your computer determines its claimed credit using: benchmark (i.e. whetstone + dhrystone /2) * time. The granted credit it calculated as the average claimed credit per decoy for that particular Work Unit type from all computers.

More efficient CPUs such as anything based on the Intel Core architecture is more efficient than the benchmarks show and so will usually get more granted credit than it claims, while some others will usually get less.

A computer's RAC will increase almost linearly with the number of cores it has, and a member's RAC is the sum of the RAC's of all the machines under their account.

HTH
Danny

---

ok, if you say it goes linearly as time goes along, then why did I take such a sharp drop and have never recovered since HINGE got started.

I don't have any HINGE work units running at the moment and instead I am running smaller jobs, such as ABRELAX, but even with those jobs running at 8 hrs the results I see in RAC are that I am not really climbing at any significant rate, rather I am gaining about 1 or 2 points over 2-3 days.

Before HINGE I was up at 240-250 range and then took a steep decline of 80-90 points after processing 4 HINGE WU's.

This does not make sense, losing about 100 points over a month, between normal ABRELAX and 4 HINGE work units and then not being able to show any significant gain in 3 days.

Should I expect a gradual increase as each day (i.e. 3 WU in 24hrs) or is it going the same at 1 point average per day or is it going to regain what it lost in one months time or less?

---