On Thu, 19 Nov 1998, Seth Leigh wrote:
> I am brand new to this list as of about a week ago. With my first batch of
> emails from this list, I saw a post from a German guy whose name I think
> was Matthias,
Well, if that was me, then it is Austrian ;-) An Austrian ALWAYS has to
point out the difference (at least, since WW2). ;-)
> who has modified two Celeron 300as as per the instructions from those
> Japanese guys on cpu-central.
To be precise: The cpu-central-guys are just mirroring the Japanese page.
> A colleague of mine at work wants to do this, too, but we don't know
> exactly which motherboard is actually capable of overclocking Celeron
> 300as to 450 MHz, and can run TWO of them this way.
First, look at the original Japanese page, they have a "dual
celeron success" list (well, not in English), but one can see which boards
are likely to work and to support Dual-Celeron or mixed Pentium-II and
Celeron.
Points to consider:
(a) The motherboard should be _good_.
I read some tests about dual-Pentium-II-boards in the German c't magazine,
and the ASUS-board tested there (well, it was not the P2B-DS which I use)
did it very well.
In addition, the "Japanese Guys" use Asus P2B-DS and said, it worked - so
I took the same.
Remember: My two Celerons plus the Motherboard costed about 11000 ATS
(which is below USD 1000) - that was the price for one Pentium-II-450 at
that time!
(b) The motherboard must be capable of driving Celeron-300A correctly
Well, that is a little bit risky. I don't know if one gets a P2B-DS
nowadays with a BIOS so old not to support the Celeron-300A, but you have
to consider that if it doesn't, you have to either contact your dealer to
update your BIOS or you've got a friend who can borrow you some P-II to
just boot up and update the BIOS.
(c) The motherboard must support two CPUs
Well, that shouldn't be too hard to find out ;-)
(d) The motherboard should be OCable in a broad range
Well, this is a "nice to have", since the Dual-option is the main goal,
but ... when the warranty is already gone due to drilling and soldering,
why not OC, too? :-)
Since I didn't know whether my Celerons run at 450 MHz (100 MHz FSB)
correctly (which was my goal), I thought it would be better to take a
board supporting 75 and 83 MHz FSB clock, too. There are not that much
BX-boards capable of more than 66/100 MHz FSB, so again the choice was the
ASUS P2B-DS. It supports 103 and 112 MHz, too, but 112 MHz didn't work for
me @2.2V (see below).
Remember: Celerons have a fixed clock ratio! The 300A has 4.5x, which
means, the FSB is always multiplied with that value to get the CPU clock.
If your board only supports 66/100 MHz, the CPU will not make it other
than @300/450 MHz.
(e) The motherboard should support voltage tuning
Well, OCing my Celerons didn't work for long (NT and Linux setup hang
immediately after boot), so you need to change the core voltage, too.
At first I thought of an Abit board, because they have their "SoftMenu"
called BIOS-level setup of voltage and clocks. Very fine :-). But first,
they don't have a dual board and second, the performance of their boards
is said to wipe out all OCing effect, means, memory and PCI performance is
_very_ bad.
The Asus P2B-DS does _not_ have such a (or any other) way of changing
the core voltage, neither in setup nor with shunts on board. Again,
one has to modify the CPUs, but (look at (d)), the warranty is gone
already, so ... don't care. Some guys do this with covering some pins of
the CPU cartridges - I did it the hard way and used a knife to cut some
pins in the voltage selection section on the Celeron (look at yahoo.com or
dejanews.com for "Overclocking" and "A119 B119 A120 B120" or so.)
2.2V worked just fine for me - don't use more than 2.6V! Always remember,
Power=Voltage^2/Resistance, means, 10% more voltage is 21% more power! 20%
more voltage is 44% more power! This brings us directly to point (f).
(f) You need good cooling for the CPUs
I ran them (both Celeron-300A BOXED) @463 MHz (925 BogoMIPs :-) ) for
about 1.5 days rc5des (RC5-64) @ 2.6 MKeys/s and 2x100% usage (Kernel
2.1.124), but after that, the PC hang. I did not try to reproduce the
problem, but I think, it is heat-related.
Running them in normal workstation usage @2x10% or so keeps them cooler
than the Coke in my fridge :-)
Since I have to do a lot of work for my studies, I have not much time for
my little "pair" (as I call them), so the new ASS (air supply system) will
have to wait ... :-)
(g) Watch out for good CPUs
Somewhere, I read about Celerons which are better to OC if they are BOXED,
i.e. equipped with the original FAN from Intel and sold in a carton. So I
bought two of them (at different dealers, different dates of production,
both MALAYSIA), and both work perfectly @463 MHz (103 MHz FSB). I don't
know how good non-boxed Celerons are, but the original fan is good enough,
even for 463 MHz. Well, in principle, but they also make a lot of noise.
But always consider this: The fan is not the problem for overheating, the
problem is _fresh_ air. So watch out for a solution to put as much of it
to the CPUs as you can.
> I need advice on what motherboard would be best. Can Matthias or anyone
> else on this list give such advice? My friend understands fully the risks
> involved. He is the type of guy who takes apart common consumer
> electronics and modifies them, so he is unafraid. In fact, he gets an evil
> grin and says "that would be NEAT!"
Remember: Win95/98 does not take _any_ advantage from having a dual CPU.
OS/2 SMP, NT and Linux work perfectly here (well, ignoring the
task hopping between the CPUs which seems to occur on every OS here ...).
Need to test BeOS ... but I have to spare some happy hours for my elder
days :-)
Good Luck!
Winschdawos,
- Matthias
--
Der Wein mit der Pille ist in dem Becher mit dem F�cher.
-----------------------------------------------------------------------------
