David Maynard wrote:
~misfit~ wrote:
I'm running a Tualatin Celeron 1.3Ghz in a Gigabyte GA-6VEM-L integrated
mobo. It doesn't have BOIS setting to change between 66/100/133, or
switches/jumpers from what I can see. It seems to get the default FSB
speed
from the CPU and it then allows you a certain overclock (in BIOS) from
there, in pre-determined steps unfortunately, I can't go one Mhz at a
time.
Well, my 1.3 is running 24/7, Prime95-stable at 124Mhz FSB for
1.612Ghz. I'm
pleased with that, it has a copper-bottomed HSF on it that came with a
Barton and stays cool, Also I removed the IHS so the die is in direct
contact with the HS. There is no way in BIOS to adjust vcore, however
this
CPU has always ran a little high, 1.53 average according to MBM5, ranging
between 1.51v and 1.54v. (Strange because I had a Celly 1.4Ghz in this
board
last week and it only ran at 1.35vcore. It ran fine though, however I
didn't
try to OC it at that low a vcore. Both have 1.5v written on them)
I've noticed odd Vcore values with tualatins on non-tualatin
motherboards as well but it's not a problem here as I always have
something else setting it.
IME Tualatins don't respond to Vcore increases anyway - at least not on
BX boards. Every Tually I've tested tops out at around 1650Mhz, and
won't go any further even if you increase Vcore to dangerous levels.
OTOH, most of my Tuallies will do over 1600 at 1.35v, and would probably
go even lower except Vcore tends to exhibit increasing ripple below 1.4v
on BX boards.
The tualatins using an open drain FET to set the Vcore identification
pins whereas the earlier celerons used straight wire shorts to ground.
And the specs for earlier tualatins specify the motherboard pull up to
be 10K, or more, while the tualatin says 1K. I could imagine that the
10K "or more" pull-up on the non tualatin motherboards is not always
sufficient to drive the pin high when it's not being pulled low by the
on-board FET.
Which would be a great theory except for not explaining why yours seem
to be 'high' instead of low, as that theory would expect. Unless the
on-die FET pull downs depend on some Vss pin that's 'unused' on earlier
motherboards.
Whatever the reason, I've seen it too.
Now, 1.61Ghz is pretty damn good I figure. (Even though it's a dreaded
Via
chipset board I get 107 marks with CPUMark99, up from 84.9 at 100Mhz
FSB).
From what I can tell I can't monitor die-temp with this board. However
socket temp sits at around 29°C, case temp 19°C, crunching SETI
full-time,
about the same as the 1.4 did using it's stock HSF. It's winter here.
I'm never satisfied with an OC until I've reached the point where I get
errors in Prime, then backed off to my last stable setting. I can't do
that
with this set up as-is. Also the PCI bus is 41Mhz, a setting I'm not
exactly
comfortable with although all seems fine. The only card I have in it
is an
Adaptec AHA-2940-AU for my old scanner and it's always been a
forgiving card
bus-speed-wise, everything else is on-board. (I haven't tried my
scanner at
this speed though). The HDD is an old 8.4GB Fujitsu. The board
actually has
three bulging capacitors, with two of them leaking a little on top. I
have
replacements ordered and they should be here in a few days. I was
going to
replace them as soon as they arrived, however, with it running this
well I
might just leave them as they are until I *have* to replace them
(unless I'm
risking my CPU?)
So, the question: Can I easilly (and non-destructively in case I need to
revert back) fool my mobo into thinking it's a 133Mhz FSB CPU? It does
run
133FSB CPUs. That way I could try for 1.73Ghz and have the PCI back in
spec.
I realise that there isn't a very good chance of me getting it to run at
that speed (especially with no vcore adjustment) but this *does* seem
to be
a good bit of silicon, I'd hate to not try it. I'm sure most of you
understand.
Unlikely you're going to get there, especially without a Vcore increase,
but who knows? It's already doing better than average to be stable at
1.6 Gig at stock Vcore.
Can it be easilly done?
Unfortunately, no. The bus select pins are pulled high on the
motherboard with pull downs on the processor to select which speed it
wants and to get 133MHz as the 'requested' FSB you need to break the
pull-down on BSEL1 (AJ31) so that it goes high (on the VRM side). The
obvious way is to break the pin, but that isn't replaceable (for most
mere mortals anyway). The alternate is to insulate it somehow.
Vcore, btw, can be increased by jumpering processor pins to Vss. You
don't have the choice of just any voltage because you can't make a pin
go 'high' with that method but, with a 1.5 volt default Vcore, you can
get 1.55, 1.6 and 1.65. The next higher jump, by only jumpering to Vss,
is to 1.9.
The really ambitious have been known to break the motherboard traces to
the socket and wire in their own DIP switches for the VID and FSB
signals to make both adjustable.
That's why I prefer Slot-1 BX boards for Tualatins - the adapter gives
you adjustable VID and FSB without resorting to 'ambitious' modifications.
This is my second machine, mainly used for SETI
(BOINC actually) and a bit of 'muling' in multiplayer Diablo2 games.
I have seen a web page with other people's OCs on it, some overclocking
site. Does anyone have the URL? I'd be interested to see how far
others have
managed to push their Tui Cellys.
Also, Spajky, what speed do you have your Tui at and how many marks
does it
get with CPUMark99 on a BX?
Thanks in advance for the replies folks.
--
~misfit~
|