OK fine, here is the data you requested:
1) Start up log:


2) Task Manager Screen shot: the first two (from left to right) idle valleys are of true idle time, the last valley I was using the PC (loaded a Spreadsheet, Word processor, web browsing, task switching, etc).

http://www.imagebam.com/image/4c8f1049208579

I know what are you going to say: this proves me right and you wrong, well that is not a correct assumption, the process window which shows a real-time CPU use, says exactly how much the CPU is used and in the 99 secs interval often both tasks show les than 50% CPU usage (1 CPU: 2 Cores, each is 50%).

And you do have an "attitude" dude, face it, and it is: Saul what you think is just crap if it works for you is OK, but for the rest of the world it just sucks. Instead of thinking that I might have a good idea and improved the WCG performance, but no, that can't be given the credit to the little me.

Here writes:

So I did.

And since you don't believe me at all I suggested you tried yourself and you took it as an insult, because you have to let it "slide", really? you're so proud that you can't take a suggestion as it is? is insulting? I must do all the proofing and you have to do nothing? You won't try to help the WCG Network, since for sure you're not trying to help me, and try to see if what I've thought could be good and worth to try it out to actually have a first hand experience and judge for yourself?

If you're tired or not in the mood for this kind of posts I can understand that, but if this is the way you support people, is really discouraging.

Regards.

Well your assessment is incorrect, because maybe the WCG Client was designed in one way but it works this way: the less CPU workload you assign to the CPU(s) the "lighter" the PC becomes; is a huge mistake to take it to 100%, the PC WILL LAG, the WCG Client will have trouble giving back control to Windows so it can perform its tasks. I have tried several workloads and 99% works fine and really doesn't get noticeable lag.

The WCG Client seems to also release the thread to give control back to Windows in a more effective way than when 100% workload is used, so is not only the amount of time spent on idle what makes the PC responsive but the percentage (%) that makes the threads running "lighter".

Also you don't need to have 2GB of RAM like I do, you can run 2 projects in a P4 HT with 784(512+256) MB (0.75 GB) of RAM and still have a spare for applications as is demonstrated in the Task Manager graphics I posted, hence I suggest the following:

Once the client is installed and runs for the firs time it could check the amount of RAM installed plus the CPU (minimum a P4) and then try to configure a 90-99% workload, depending on a small & fast benchmark to test the System responsiveness, then assign a suitable profile for the percentage calculated in the benchmark.

I really think you are wrong when you say most P4 HT won't run well on this configuration, as i wrote if they have at least 768 MB of RAM they'll probably do OK at 99%, and of course at any percentage above, the extremely low, 60%.

Is really a shame that many ideas that could improve WCG Network are actually minimized and stuck here in this forums just because the moderator thinks it won't work well.

This is my formal protest against this absurd behavior. Ideas like this one should be case study on a small network with actual users to see how good they really work.

Regards.

Saul,
Following is the result of my testing your settings in my P4 HT.
This machine has 1 GB of RAM and is normally set to 100 % both for CPU time usage and for the number of processors to use, therefore it runs two tasks without any problem. The temperature of the processor reported by Speedfan 4.32 is 56 °C today.
The BOINC client is WCG's 6.2.28, same as yours.

I have changed the settings in local preferences, both because it is the way I use to do it and because it is the only way to be certain of what is really used.

Setting the "Use no more than: xxx % of processor time" parameter to 99 %.
- Task manager's graph shows valleys at the same frequency as yours, i.e. every 99 seconds (measured with a stopwatch).
- No noticeable change for the temperature of the processor (56 °C).
- CPU use reported by TaskManager is 100 %

Setting the "On multiprocessors, use: xxx % of processors" parameter to 99 %.
- BOINC immediately switches to one task only and starts running the CPU benchmark
- CPU use reported by TaskManager falls to 51-52 %
- Temperature falls to 54 °C since the processor is used a little less efficiently
- TaskManager's graph shows that the work is split between both threads, about one third on one and two thirds on the other one.
- Valleys are still occuring every 99 seconds but are less clearly visible since the split of work between both threads is not straight.

Incidentally one short deadline task which was waiting normally in the queue became "high priotity" and started after the benchmark while the two other tasks running before the change became "Waiting to run".

Edit: Typos and problems with special characters

Although we would love to set the default CPU usage above 60% the reality is that there are some laptops and PC's that were not designed to dissipate the heat generated from running at a higher rate. It is really important to us to be able to say that the software will run on your computer without interfering with your normal usage. If we set the throttle higher, then we know that some computers will have issues (even if it only a few - I personally run the throttle at 100% on all my computers without issue). This would cause many potential long term contributors to never try it out in the first place.

That is a good idea. Problem is that WCG uses BOINC and so far efforts to get the processor usage logic in BOINC changed have failed (the developers rather tinker with GPU support I think )

Therefore someone has developed an external application that smoothly throttles the BOINC sciences when the cpu temperature goes above a certain temperature. This is the TThrottle application Sekerob was referring to. It works great on my laptop, smoothly throttling my Boinc processes to keep the cpu temp under 65 °C. Have a look at it.

http://boinc.berkeley.edu/addon_item.php?plat...%2F%2Fefmer.eu%2Fboinc%2F