Reader warns of Intel-NT overclock shock
Cunning Wintel plan puts spanner in works?
Reader's Letter Chip giant Intel has warned since the beginning of this year that to overclock its processors is verboten. A reader describing himself as 'Firestormer' wrote to us with the following tale: "You might find this interesting, but if I'm correct, Intel and MS have finally come up with a way to defeat overclocking totally, at least under NT4 Service Pack 5 and with the Pentium III. The sorry tale is as follows. "I've dabbled from time to time with FSB adjustments on my Tyan Tsunami-based system, originally tweaking my Pentium II 350 to see if I could get another 50MHz or so out of it. My NT4 SP5 barfed at 112MHz bus, but I was able to get it running at 103MHz bus, and the speed difference showed up accordingly under NT Diagnostics as well as H-Oda's WCPUclk util. The performance boost was there, but at the time another 10MHz didn't seem to give enough of a boost to be worth while. "Recently, however, I upgraded this system from the P2-350 to a P3-450, and then tried to tinker with overclocking via the FSB yet again (enough is never enough where speed is concerned :P). BIOS, check; post screen, check; DOS level utils, check... all reported ~500MHz correctly (I'd tried 112MHz bus again). "However, upon getting into NT, I ran my benchmarks and found that the system performed identically to before the overclock. Upon further analysis, I found that all of my speed reporting utils still said that the CPU was running AT EXACTLY STOCK. Nothing I tried could get the chip to run under NT4 SP5 at anything other than stock and only stock speed. "This changed, yet again, when I manually replaced HAL.DLL (you know, the lowest level part of NT there is) with the HAL.DLL.SOFTEX that is included with SP5 so I could make the system shut off at NT shutdown. Earlier this evening, I decided to experiment with the FSB overclocking yet again, determined to find out why the overclock failed, only to find that it worked this time. "This has lead me to one and only one conclusion: the HAL.DLL in NT4 SP5 is capable of reading the processor serial number/ID tag and is using that information to ensure that the CPU will only run at stock speed under NT. And considering that there are utilities available which do similar things (see H-Oda's SoftFSB), this seems the most likely, if somewhat disturbing, explanation. "Make of it what you will, but this is one tidbit of information I'd like very much to see verified (if possible) and if found correct spread far and wide across the Net." ® NB We have received an unprecedented flood of emails about this reader's letter, many saying his points are factually wrong. If you want to comment, please post your replies to our bulletin board.
Sponsored: Benefits from the lessons learned in HPC