geez, I'm impressed! I don't know if I would've had the guts to try out
those combinations... Height-wise, things must've been getting a bit
cramped, though?
I don't even remember what heatsink I used, but one of the nice things
about Tualatins is their relatively low heat, so older heatsinks can
be reused or reshaped... It's a lot easier to play around, try
different things when it's not your primary or important system in
jeopardy.
Thanks, in light of your comments here, I've narrowed my purchase options
down to Slot1-to-Tualatin convertors which seem to have this safety
mechanism.
In the worst case scenario of frequent shut-downs, I guess I'll just go back
to the P3-800. When I have more time to rip my system apart, I could then
get a Tualatin supportive motherboard and use the Cel1.4 CPU that way, and
sell the adapter...
If the adapter has onboard voltage regulation there should be no
chance of the shutdown problem, it will actually reduce the power the
motherbaord supplies.
Hmm. My CPU is currently running on 1.65V, and I recall that my Gigabyte
BX-2000 mainboard automatically adjusted the voltage for me when I switched
between Cel300A and P3-800 CPUs. Plus, my RAM is 4 bars of Kingston 256MB
PC-133. So the chances of a stable OC seem good with a Cel1.0/1.1?
The motherboard has a switching regulator controller which senses
logical high/low state of the CPU pins... that is, it ran at 1.65V
only because that's what the CPU "told" it to do. A Tualatin defaults
at a lower voltage, so if the motherbaord were suppling power for the
CPU, it'd start out lower than 1.65V, and you'd need either a slotket
with jumper settings that override, mimic the CPU's pin high/low state
per desired voltage, or to do one of a few different mods... There
are three or four ways to achieve higher voltage:
Wrap wire around appropriate CPU pins.
Solder to connect slotket pins.
Solder to connect motherboard slot pins.
Solder voltage regulator controller VID pins.
Generally most people prefer wrapping wire around the CPU pins, is
quicker and easier but leaves the CPU sitting barely at an angle which
isn't too much of a problem with regards to heatsink contact since the
Tualatin's heat spreader is more forgiving than a tiny bare core would
be.
But, the only thing is that locally here, the price of the Cel1.4 and Cel1.0
are almost the same, so I'm leaning toward just getting the Cel1.4 and not
bothering with overclocking.
That's what I would do since you have a BX chipset. If you'd had a
Via 694X or 694T, Intel 815, I would've chosen the 1.1GHz (as I did).
I also have a 1.2GHz Tualatin Celery somewhere, which won't run on
133MHz FSB stabily no matter what the voltage is increased to.
(as a side comment, I OC'd a P2-266 system to 350MHz a while back; general
use was stable, EXCEPT whenever I tried to defragment the HD, the system
consistently got "wiped" out and I had to re-install the OS/apps. So once I
got my P3-800 stable at 896MHz, I've actually been afraid to defragment my
HDs, over a year now! anyone else have this problem?)
The key is to look at what you're overclocking. By running the P2-266
@ 350, you had a resulting 44Mhz PCI bus speed. Generally to play it
safe you should shoot for no more than 36MHz PCI, which you've barely
exceeded. The odds of corruption are far less than with your P2 o'c,
but still a slight chance. By overclocking to 133MHz FSB you would
have a 1/4 PCI divider, so the PCI bus is back exactly in-spec, at
33MHz, should pose no greater risk for HDD corruption than on 100MHz
FSB (providing your memory is error-free at 133MHz, and CPU has no
errors overclocked to the resultant operating speed). So of course if
your P2 had been able to o'c to 400MHz, you wouldn't have seen any HDD
corruption due to exces PCI bus speed.
Another interesting point. I never considered that I'd have a better chance
of hitting high memory bus speed with less RAM. I can live fine with
2x256MB instead of the current 1GB, so I might keep this in mind when
optimizing the Cel1.4 setup.
It's often overlooked... many boards will run quite stabily at default
speeds with a memory module or two, but not with three or four modules
at default speeds... depends on the specific memory modules too, but
the highest o'c would always be with one module, then "probably" with
2 single-sided modules. Actually I think the BX chipset is spec'd to
require registered memory for more than 512MB anyway, even though many
people were able to run more without problems. Memtest86 is very
helpful to test memory stability:
http://www.memtest86.com
My BX2000 mainboard has jumper settings described in manual addendum for
133FSB, so it looks like I have a lot of options open, and a lot of things
to try out.
With the BX chipset your AGP bus is out of spec at 133MHz FSB, at
89MHz. Many video cards will run fine on 89MHz AGP port but some
won't and some need side-band addressing (SBA) disabled to retain
stability.
If your board has Gigabyte's questionable northbridge mounting, which
uses a tiny heatsink attached by foam tape (may not even be thermal
tape), you may find greater stability at high FSB speeds by replacing
the 'sink, with one using thermal compound and through-board mounting
if possible, or otherwise attached with epoxy. It need not
necessarily be a loud fanned 'sink, but almost anything would be
better than Gigabyte's stock sink.
Dave
