Given the slower w98 machine's relatively fault-free performance, and in light of the built-in suspend features of the BOINC agent, without screen-saver activity (of which I wasn't originally aware) , I've reinstalled the same BOINC rev in the faster w98 machine.

I'm sort of satisfied that previous problems occurred due to differences in throttling methods between UD(Win only) and BOINC, that make user operations inadvisable at any throttle setting short of 'off'.

The only fear I have is that there will be issues when the program suspends and resumes - that ominous pause in keyboard response, when BOINC is deciding " Is the machine really in use?" and I'm deciding "I guess I really should add the last item and save this 60,000 line spreadsheet.........."

As is the past, no news will be good news.

RL

re GUI, each SW application has it's own, though it may borrow heavily from elements supplied by the OS. There are as many GUIs running as there are open programs.

With regard to stable W98 OS, this has always been an oxymoron. I am currently letting the slow machine run and am experimenting with throttling and suspension on the fast one.

I have some ideas about the BOINC throttling method that have hardware implications that you likely haven't considered - as your solutions generally seem to hover around OS, sw revisions and hardware drivers. It requires the use of a scope to confirm or negate my suspicions, so it may be some time before I've anything to report.

Basically, it's possible that all motherboard/PSU combinations may not be up to snuff when it comes to 100% load transients on the processor supply rail, possibly producing flakey performance on hardware using the other rails. Just a hunch, but backed up by a few decades of product design and by my own poor estimation of current commodity PC marketing practice. Without a few pics, its just speculation.

RL

The Thermal Throttling option in BOINC is not OS or hardware dependent. Specifically done so it works the same in all the (OS) platforms out there. BOINC will pause and resume so many times per 10 seconds as is set in the percentage you want to throttle. meaning that at 50%, you run 1 second, pause 1 second. At 90%, you run 9 seconds, pause 1 second.

It's a thermal throttle only, nothing more nothing less, while it had to be implemented this way as there is only one source code available between all platforms. You download the BOINC source code and can go compile it for Windows, Linux, Mac OSX and a lot more where it's been ported to by now.

So any operating system dependent method of throttling cannot be used as it will not be the same across platforms. Then the developers would need to make a separate source code for all the separate operating systems, all with their own dependencies and hook-ins. Perhaps doable by a commercial company with plenty of programmers, but not by the 3 to 5 people who program BOINC in the not commercial outfit that BOINC is.

Want them to do it your way? It is Open Source. Go make your own working version, send it to the developers, send your source code with changes to them, show them what you changed.

If not, wait for a future 6 version of BOINC which will have a refined version of the thermal throttling system included. The version as is works on a plethora of systems out there, if not on yours it is most probably your system that's the problem.

3rd party throttling software as BES and Threadmaster need higher OSes. While BES may work with Windows Millennium but definitely not on Windows 98, Threadmaster needs a minimum of Windows 2000 to be compatible with the API used.

It won't matter anyway as it won't work in Windows 98, as I said above as well.


Your guess is as good as mine, perhaps 6.4
It's definitely not in the upcoming 6.2 release, that one uses the same method as 5.8 and 5.10

I managed to scope the auxiliaries in both machines and see no abnormal regulation effects from the throttled prcessor load.

http://www.magma.ca/~legg/TVS/unit1_throttle.gif
http://www.magma.ca/~legg/TVS/unit2_throttle.gif

The recognisable regulation pattern had margin on the 2% variation that might be expected in a well regulated system, on either the 12 or 5v rails.

If the throttling method has any effect, it's not through the supply rails.

RL