Johan wrote:
Well, what I just came to think of is this: Would you like me to zip and mail all my NHC-stuff from my T42p 2373-Q1U (the program and all settings, XML and CSS files etc); I mean you could then simply temporarily rename your present NHC-directory, and try see if mine works better on your ThinkPad? If you want this, please let me have an e-mail address (you may send it via a PM). The whole NHC (zipped) incl. exe's is about 7 MB - that's not too bad for mailing, I think.
Well, I've finally had time to sit down and systematically play with NHC to see if I could address my GPU heating problems. Work has been busy the last few weeks, and I have been mildly distracted by this tall blonde Dutch student who will be working in our lab for the next half-year.
I first tried switching the IBM.css, IBM.xml, and IBM.dll files in the "ACPI" folder of my NHC directory with those of
Johan, testing all combinations of my versions with his. I observed no difference in reported fan speeds (always maxing out at ~2900 rpm). GPU temperatures were assessed briefly with no improvement noted.
I then decided to switch my settings.xml file in NHC's main directory with that of
Johan. I immediately noticed the GPU not heating as rapidly when subjected to the Video Card Stability Test program, so I investigated further. The only differences between the two files were more conservative CPU undervolting in
Johan's version along with the "Enable Custom Dynamic Switching" option being checked under the CPU speed tab. I modified these settings to contain the more aggressive undervolting that I use but left the "Enable Custom Dynamic Switching" option checked.
With these settings, the situation seems improved, as Video Card Stability Test has not yet pushed temperatures over 100 C, which had been very easy to do before. The temperatures are slowly creeping into the upper 90s after ~15 minutes as I write this, so perhaps only the time constant for the rise has been increased. Interestingly, the reported fan speed isn't much different, maxing out at ~2990 rpm.
So, in short, I'm not sure what's going on, but
Johan's settings.xml file
seems to have improved the situation. Perhaps my old settings.xml file got corrupted somehow.
I would still like to see GPU undervolting added to a future release of NHC as I think it would be a useful feature.
EDIT
I'm clearly daft; perhaps that Dutch student is getting to me.

The reason the temps were lower was because
Johan's settings had ATI Clock Control enabled with a setting of 175 MHz core / 157 MHz memory. If I disabled the clock control so that the default 320/210 was available (but left PowerPlay enabled), Video Card Stability Test would heat the GPU until I got warning beeps from the machine.

A minor underclock to 300/200 slightly improves the problem, but the warning beeps are not something I was getting before (though I also hadn't torture tested the GPU to this extent before). So, there is no improvement using different NHC scripts; the issue is with heat dissipation, perhaps from my removing and replacing the GPU's heat spreader during a CPU upgrade and AS-5 application process (there was no other way to get the fan off). Again, an easy way to undervolt the GPU (i.e. without having to write to the video card's BIOS) would be a big help.
FYI, the machine has a MR 9600 64 MB with the latest release of Omega Drivers (v. 4.8.442).