I assume that Task Manager shows Rosetta running 49 - 50 % (with hyperthreading) or 99% without hyperthreading. Cpuidle should be 0%. The only 3 possibilities that occur to me offhand do not seem to fit.
1) temperature mandated slowdown.
2) competing process.
3) software enforced idle (such as threadmaster or cpuidle [a misleadingly named utility that ensures cpuidle runs a percentage of the time]).

Anybody have better ideas than I do?

mycrofth

Bruce --
I am puzzled by this as well. I have not seen anything in the forums since the beginning of the project which would account for a processor rating to drop as you are seeing other than those listed by mycrofth in his post. I begin to wonder if you might not be experiencing some type of hardware problem which is causing the slowdown. That being the case, I decided to look for a free software package that could benchmark a processor and think I have found one that looks pretty good. It has a lot of features to it, but the one I was interested in was the processor benchmark. I downloaded it and ran it on my system and found that it reported my 2.4 GHz Pentium 4 (normally it gets from about 115 to 130 on the overall rating by the WCG Agent -- the most recent processor rating for it was 146) and compared it to other processors pretty accurately (it gives both a graphical and text report).

My processor tested out as follows:

CPU Bench Result
WhetStone FPU 2,372 MWIDPS
DhryStone ALU 5,558 MDIPS
Speed 2390 MHz

If you wish to try it, it is called "Fresh Diagnose" and it is available for free download at http://www.freshdevices.com/freshdiag.html. You have to provide them with your e-mail address to receive a pointer to a download site and to obtain a registration key. Hopefully I have not opened my e-mail to a lot of spam. I cannot vouch for this aspect of it yet.

Let us know the results if you try this. Other than that, I am not sure if I can be of much help. I wonder if anyone else might have any ideas on this.

I downloaded Fresh diagnose and ran the benchmarks over 2 days with all process using cpu turned off.

The speed for all trials, about 20, remained constant at 3408mhz.

The MWIPS varied between 794 and 1878. No rhyme or reason to the variation either.

The MDIPS varied between 2879 and 3335.

Fresh diagnose exhibited the same behavior as the processor score section of the WCG agent benchmark.

I am accessing the processor score and not overall benchmark on the agent.

Those are good points you brought up on the watercooling.

The prescott core especially at 3.4 ghz runs extremely hot. This system because of 24/7 rosetta work has 3 fans on the radiator two on the case and I have a 24 inch box fan blowing into the front of the case.

Because during the 1980's and 1990's the only processsors around were intel based systems. I got real use to them. But this prescott experiance has definiely converted me to amd. Just for the cooling aspects alone.

I am real nervous about breaking into that watercooling system. It was one hell of a nightmare to install.

I am reviewing all my results since this system went online in d
December to see if I have a degrading in work accomplished.

I'm not sure what is happening on your particular system. However, the agent runs some sort of quick benchmark, each time it communicates with the server to get new work. It might be that right at this time, something else is triggered to run for a short time. For example, if you have a virus checker which scans every file written, it might be invoked and be checking the new downloaded work data or newly assembled data file to be returned at the same time that the quick benchmark is running. Or, the software firewall might be doing something similar. Also, a spyware checker might come alive at that point. So, this is just speculation, but might explain some of this if something else runs while the benchmark runs.