This might not work for everyone, but try going into your BIOS, and DISABLE L2 cache ECC checking. (please do not disable the L2 cache itself). This should definitely affect Tualatin systems, but may also affect P4's and AXP's.
Also note that the ECC, according to Intel, can not be disabled on P3 and higher CPU's, but....
On my P3 1.4-S @ 1575 mhz (FSB 150), before disabling L2 ECC, the following things happened: (Note: Abit BH6 1.1, Win98 SE, 256 MB PC133 cas 2-2-2, Gf4 Ti 4600).
Games: stable (some OGL games would crash occasionally on alt tabs)
3dmark: randomly CTD, usually in less than 3 loops. (note:3dmark is totally stable at 144 FSB). 3dmark score about 10,500. UT2003 benchmark 140/53.1
Icon corruption in some programs, starts at 144 FSB, but very uncommon. Happens much more at 150 FSB.
Occasionally might get a CRC error when unraring something (very rare).
SuperPI and Prime95 are stable, though.
Unexplained, very rare lockups sometimes at 150 FSB (like, sometimes trying to close a window, can freeze the entire computer hard).
-------
Ok, after disabling L2 cache ECC checking: FSB at 150:
1) CPUZ diagnostic still reports it as enabled (of course, Intel says you can't disable ECC on P3's and higher).
2) 3dmark ran for 24 loops before I got sick of it.
tried it again later, and it ran for 11 loops before I got bored. No CTD's whatsoever. Score is roughly 10,500.
3) No program icon corruption in 3 days, so far. *knock on wood*. None at FSB 144, none at FSB 150.
4) Ut2003 benchmark 140/52.97
5) Happily completed Super PI 32M test (though I never did this test before).
Good old IE will still occasionally, but very rarely crash in user.exe, but it's always done that. Doesn't seem to be related to the overclock. (seems more common when other ie windows are opened) might be caused by a cacheman setting; happens at 144 fsb too (never tried lower)
No problems unraring files.
Another user on the overclockers forums brought this to my attention. he was only able to overclock his Tualatin (cel-T) to 1680 mhz originally. After disabling L2 cache ECC testing, even though all diagnostic programs still show it as enabled, he could then reach the 1.7 ghz barrier, and run at 1715 mhz. Enabling L2 cache ECC checking caused random errors in all programs, at 1715 mhz. (and 3dmark CTD's at 1700 mhz, but all other programs stable).
I didn't believe it when I saw his post, but not getting annoying icon corruption and no CTD's in 3dmark is enough to turn the eye of the most jaded overclocker.
So, give it a try (especially for people getting CTD's in your benchies). Won't hurt. This apparently works on even P4's and AXPs.
Now the magic question is: Why?
One theory is, since the ECC can't be disabled on P3's and higher, that enabling it in the BIOS is somehow screwing things up and stopping it from working properly (even though it's enabled), and disabling it in the BIOS actually makes ECC work as it should.
Another theory is that disabling in the BIOS is somehow reducing motherboard load, which is rather important at high FSB speeds.
Also note that the ECC, according to Intel, can not be disabled on P3 and higher CPU's, but....
On my P3 1.4-S @ 1575 mhz (FSB 150), before disabling L2 ECC, the following things happened: (Note: Abit BH6 1.1, Win98 SE, 256 MB PC133 cas 2-2-2, Gf4 Ti 4600).
Games: stable (some OGL games would crash occasionally on alt tabs)
3dmark: randomly CTD, usually in less than 3 loops. (note:3dmark is totally stable at 144 FSB). 3dmark score about 10,500. UT2003 benchmark 140/53.1
Icon corruption in some programs, starts at 144 FSB, but very uncommon. Happens much more at 150 FSB.
Occasionally might get a CRC error when unraring something (very rare).
SuperPI and Prime95 are stable, though.
Unexplained, very rare lockups sometimes at 150 FSB (like, sometimes trying to close a window, can freeze the entire computer hard).
-------
Ok, after disabling L2 cache ECC checking: FSB at 150:
1) CPUZ diagnostic still reports it as enabled (of course, Intel says you can't disable ECC on P3's and higher).
2) 3dmark ran for 24 loops before I got sick of it.
tried it again later, and it ran for 11 loops before I got bored. No CTD's whatsoever. Score is roughly 10,500.
3) No program icon corruption in 3 days, so far. *knock on wood*. None at FSB 144, none at FSB 150.
4) Ut2003 benchmark 140/52.97
5) Happily completed Super PI 32M test (though I never did this test before).
Good old IE will still occasionally, but very rarely crash in user.exe, but it's always done that. Doesn't seem to be related to the overclock. (seems more common when other ie windows are opened) might be caused by a cacheman setting; happens at 144 fsb too (never tried lower)
No problems unraring files.
Another user on the overclockers forums brought this to my attention. he was only able to overclock his Tualatin (cel-T) to 1680 mhz originally. After disabling L2 cache ECC testing, even though all diagnostic programs still show it as enabled, he could then reach the 1.7 ghz barrier, and run at 1715 mhz. Enabling L2 cache ECC checking caused random errors in all programs, at 1715 mhz. (and 3dmark CTD's at 1700 mhz, but all other programs stable).
I didn't believe it when I saw his post, but not getting annoying icon corruption and no CTD's in 3dmark is enough to turn the eye of the most jaded overclocker.
So, give it a try (especially for people getting CTD's in your benchies). Won't hurt. This apparently works on even P4's and AXPs.
Now the magic question is: Why?
One theory is, since the ECC can't be disabled on P3's and higher, that enabling it in the BIOS is somehow screwing things up and stopping it from working properly (even though it's enabled), and disabling it in the BIOS actually makes ECC work as it should.
Another theory is that disabling in the BIOS is somehow reducing motherboard load, which is rather important at high FSB speeds.
