Just a few notes.
First of all, I am really happy to read about the first success stories, so thanks to Everybody.
Then one note about the different extent of speedup different people are getting. Basically, it all depends on the hard drive. The newer, bigger and faster the hard drive is advertised to be, the more this fix will speed it up. Or I could put it this way: the faster your hard drive is the more the IO holds it back, so the more you need this fix.
This note mostly for jbarr. I see your hard drive already gives you its best at X=21. In this - and any similar case - I would actually recommend against using X=10. In these cases it is safer to use the more relaxed setting, and it may even give you better performance. If the chipset is just a bit too fast for the HDD it may have to repeat its requests to the HDD from time to time. (If the timing is really off, that may cause data loss...). I've actually seen this fix with X=10 slow down a 4GB IBM Travelmate. I could only experience it when getting to a particular point in Win98 install. I actually had to take it back to X=32 for it to work the best. It was the correct value, proven by the fact that Corespeed already showed the speed improvement with X=32 that it did with X=10. In conclusion: if you don't have the latest and greatest of HDD technology, it may not be a bad idea to only push down the value of X as long as it really helps HDD speed in Corespeed. Hope this helps.
To vailr. The fix is not OS-specific. Only the tools I used to present it. If you have a software tool to manipulate PCI registers in XP or Linux, that should work equally well. But you really shouldn't forget about the importance of testing with this fix. Otherwise you won't know there is a problem until it's too late. I would definitely recommend testing what you can do in DOS, and once you have the value which works best for you, you could use it in your OS of choice. Ideally this fix should be - and I am sure it will be in no time - part of the BIOS.