P2B and > 1ghz Celeron ?

[Bob]

Assuming I get a slot-1 slocket, will a P2b (>1.10 & correct
regulator) actually run a Celeron at 1.1 or maybe 1.4ghz at 100fsb ?
That is, does it really work ? Is memory really critical or is
it not that crucial since you are only running 100fsb ? Do you end up
playing with the voltages and cooling to get stability ? Can I just
plug-n-play ? If I am looking for stability, would I be better off at
1.0ghz or lower ?

The price difference is meaningless in the CPU's, so stability is the
issue for me. I don't mind tweaking voltages as long as there is one
that will eventually work. FWIW, I'm upgrading several of these all at
once and I don't want to make a volume mistake.

Thanks,
Bob

 

[Barry Watzman]

Yes and no.

You need a very expensive slocket to make a Tualitin-family CPU chip
(which is what you are talking about) work in this motherboard, but it
can be done, at least on some versions of the motherboard. However, in
my opinion, while it may be physically possible, it's not worth it, you
are better off upgrading to a new motherboard.

 

[Stephan Grossklass]

Assuming I get a slot-1 slocket, will a P2b (>1.10 & correct
regulator) actually run a Celeron at 1.1 or maybe 1.4ghz at 100fsb ?

http://homepage.hispeed.ch/rscheidegger/p2b_procupgrade_faq.html
With the correct Tualatin compatible slot adapter, yes.

The price difference is meaningless in the CPU's, so stability is the
issue for me.

Since at least on the P2B-S the voltage regulator seems to be stressed
much more with a 1.4 GHz chip, I'd recommend the 1.2 GHz Celeron. You
might even be able to reduce the core voltage a bit.

Stephan

 

[Dave]

Yes and no.

You need a very expensive slocket to make a Tualitin-family CPU chip
(which is what you are talking about) work in this motherboard, but it
can be done, at least on some versions of the motherboard.

If it's the right PWM IC, you can use a ~$25 Slot T Upgradeware slotket. Not
very expensive at all. In fact, he can upgrade the CPU for around $100, all
told.

However, in

my opinion, while it may be physically possible, it's not worth it, you
are better off upgrading to a new motherboard.

Eh, perhaps. But this will certainly be a good deal more expensive, and
he'll have to buy memory, processor, mobo, bigger PSU, maybe even a new
video card...

Ballpark figures:

Barton 2500: $75
Good, cheap Socket A mobo: $75
Obligatory minimum 512-meg memory stick: Around the same
Decent PSU: $50
Cheezy and hopefully 128-bit FX5200: $75

So far, we're at more than 3x. And here comes the Monkey Wrench (TM):
Powerleap dropped their prices recently, they have listed:

PL-iP3/T 1.4 GHz Intel Celeron $119.95

Now THAT's more LIKE IT, Powerleap! Their adaptor is head-and-shoulders
above the Slot T in functionality, facility (and expense!). Their heatsink
sucks, tho.
Still, if the regulator is at all a point of concern, this would be the way
to go, now that it's only a little more expensive than the barebones
approach (which should work fine in this case, nonetheless), and certainly
less expensive than before.

You were saying, Barry... ;-) Of course, then we get into the ATA-133
controller (Highpoint Rocket 133 dual-channel, *can be flashed into RAID
support* <hehe>, only $30...) If you're doing this bulk (upgrading the
office boxen on the cheap), the Slot T/Tully combo is even more a
cost-effective choice than a whole system upgrade.

Yep. Should.

Is memory really critical or is
It is a really critical bottleneck, but other than that, normally CPU-bound
apps fly. If you're leaving the FSB alone, your existing memory will work
fine. Running 133 MHz FSB will wake things up a bit, but YMMV with stability
here.


Do you end up
Shouldn't have to...

Can I just
You need to flash the latest BIOS beforehand (v. 1014xxx, IIRC), and turn
off CPU voltage monitoring in BIOS, other than this, it's a shoo-in.
Multipliers are auto-set.


If I am looking for stability, would I be better off at
I'm not sure what draws more, a 1.0 G CuMine, or a 1.4G Tully Celly. Both
seemed to tax the power regulator circuit a wee bit on this here P2B-F. I
have a 1.1 Celly in a $20 Slot T running 133 MHz FSB. The biggest hurdle was
finding a video card to be 24-7 stable with an 89 MHz AGP 2x slot. 2 GF3's,
1 GF4 4200 later, I have a Radeon 8500 in here and it works fine. The Nvidia
cards were rather power-hungry, so they caused occasional hard locks. Not
Acceptable, Sir! The ATI card works without complaint and this box has been
completely stable for many months. Aside from the video thang, it was all
PnP.

P2B and Wes Newell would likely be able to fill in on detailed specifics.

 

[P2B]

Assuming I get a slot-1 slocket, will a P2b (>1.10 & correct
regulator) actually run a Celeron at 1.1 or maybe 1.4ghz at 100fsb ?
That is, does it really work ? Is memory really critical or is
it not that crucial since you are only running 100fsb ? Do you end up
playing with the voltages and cooling to get stability ? Can I just
plug-n-play ? If I am looking for stability, would I be better off at
1.0ghz or lower ?

I have upgraded numerous P2Bs to Tualatin Celeron and P3 processors.
Provided the board has the correct voltage regulator (as you noted), the
latest BIOS (1014beta3), and you use an Upgradeware Slot-T adapter, it's
a very stable plug'n'play upgrade. The Slot-T jumpers are set to 1.5v by
default, this can usually be reduced to 1.45v or even 1.4v without loss
of stability. The Intel retail heatsink and fan provides more than
adequate cooling.

Memory is the performance bottleneck on P2B systems with Tualatin
processors, so it's advantageous to maximise FSB speed. If you have
PC133 memory, a 1.0 or 1.1Ghz Tualatin overclocked to 133Mhz FSB will
perform much better than a 1.4Ghz on 100Mhz FSB - with no loss of
stability. Running the P2B at 133Mhz does overclock the AGP bus, but
most reasonably recent video cards will tolerate this without problems.

The price difference is meaningless in the CPU's, so stability is the
issue for me. I don't mind tweaking voltages as long as there is one
that will eventually work. FWIW, I'm upgrading several of these all at
once and I don't want to make a volume mistake.

I would choose processors based on the existing memory (assuming you
don't wan't to replace it). Systems with PC100 memory get a 1.4Ghz and
run at 100Mhz FSB. Systems with PC133 get a 1.0Ghz and run at 133Mhz
FSB. The 1.0Ghz Tualatin Celerons are virtually guaranteed to be stable
at 1.33Ghz and (at most) default voltage. 1.1Ghz processors are a fairly
safe bet too, but you might be unlucky and get one that isn't stable at
1.46Ghz.

HTH

P2B

 

[P2B]

Yes and no.

You need a very expensive slocket to make a Tualitin-family CPU chip
(which is what you are talking about) work in this motherboard, but it
can be done, at least on some versions of the motherboard. However, in
my opinion, while it may be physically possible, it's not worth it, you
are better off upgrading to a new motherboard.

Very expensive slocket? Slot-T adapters are $20, and will work on any
P2B board with the newer voltage regulator mentioned by the OP.

 

[joel]

...
The 1.0Ghz Tualatin Celerons are virtually guaranteed to be stable
at 1.33Ghz and (at most) default voltage. 1.1Ghz processors are a fairly
safe bet too, but you might be unlucky and get one that isn't stable at
1.46Ghz.

P2B

I've noticed Powerleap sells the 1.1Ghz tualatin celery in their $99
special. I guess they want to minimize your chances of getting
anything to compete with their 1.4Ghz model.

Or maybe they ran out of the 1Ghz models.

I think I'd just buy the old 1Ghz coppermine core celeron and forget
the adapter and tualatin. You lose some cache, but save some too

(duck and cover)

 

[Stephan Grossklass]

I think I'd just buy the old 1Ghz coppermine core celeron and forget
the adapter and tualatin. You lose some cache, but save some too

Well, the differences between a CuMine Celeron and a Tualatin Celeron
(or a CuMine PIII, for that matter) aren't limited to the L2 cache
*size* alone:

* L2 cache size: 128K vs. 256K
* L2 cache associativity: 4-way vs. 8-way
* L2 cache latency: 11 cycles vs. 7 cycles

That's why a PIII-700 is about as fast as a Celeron 850; a Celeron 1.1
should be about as fast as a PIII-900 then. A Tualatin Celeron, however,
is pretty much exactly as fast as a Coppermine PIII at the same clock
speed, which is not too shabby.

Stephan

 

[P2B]

Well, the differences between a CuMine Celeron and a Tualatin Celeron
(or a CuMine PIII, for that matter) aren't limited to the L2 cache
*size* alone:

* L2 cache size: 128K vs. 256K
* L2 cache associativity: 4-way vs. 8-way
* L2 cache latency: 11 cycles vs. 7 cycles

That's why a PIII-700 is about as fast as a Celeron 850; a Celeron 1.1
should be about as fast as a PIII-900 then. A Tualatin Celeron, however,
is pretty much exactly as fast as a Coppermine PIII at the same clock
speed, which is not too shabby.

Exactly - and worth noting a Coppermine core will rarely clock much over
1.2Ghz, while a Tualatin core rarely fails to make 1.5Ghz.

 

[Bob]

Memory is the performance bottleneck on P2B systems with Tualatin
processors, so it's advantageous to maximise FSB speed. If you have
PC133 memory, a 1.0 or 1.1Ghz Tualatin overclocked to 133Mhz FSB will
perform much better than a 1.4Ghz on 100Mhz FSB - with no loss of
stability. Running the P2B at 133Mhz does overclock the AGP bus, but
most reasonably recent video cards will tolerate this without problems.

Will setting the AGP jumper to 2/3 solve this potential problem
if it occurs ? Or will that slow the graphics and negatively impact
performance ? These boards are running ATI 3d Rage Pro 2x cards.
I'd prefer not to upgrade them since whatever I buy will no doubt
also be out-of-date when I replace these mobo's. This is just to
tie over these (several) systems until I can spring to rebuild
them from the ground up.

Thanks,

 

[P2B]

Will setting the AGP jumper to 2/3 solve this potential problem
if it occurs ? Or will that slow the graphics and negatively impact
performance ? These boards are running ATI 3d Rage Pro 2x cards.
I'd prefer not to upgrade them since whatever I buy will no doubt
also be out-of-date when I replace these mobo's. This is just to
tie over these (several) systems until I can spring to rebuild
them from the ground up.

Thanks,

The AGP jumper should be set to 2/3 when FSB is 100Mhz or higher, but
2/3 of 133 is still almost 89Mhz. Unfortunately the BX chipset does not
support a 1/2 divider, which is 'correct' for 133Mhz FSB.

I've had mixed results with ATI 3d Rage Pro video cards at 89Mhz - most
are OK, some misbehave. IME nVidia-based cards never have a problem, but
other chipset brands are less predictable.

P2B

 

[Bob]

The AGP jumper should be set to 2/3 when FSB is 100Mhz or higher, but
2/3 of 133 is still almost 89Mhz. Unfortunately the BX chipset does not
support a 1/2 divider, which is 'correct' for 133Mhz FSB.

I've had mixed results with ATI 3d Rage Pro video cards at 89Mhz - most
are OK, some misbehave. IME nVidia-based cards never have a problem, but
other chipset brands are less predictable.

Oh... I didn't know that. I thought you only had to set it for speeds
over 100mhz. I realize now that it should be 66. I've got it running
at 100mhz with my ATI Rage Pro... so I guess it will do 89 OK. But, I
don't know if all my ATI cards will based on what you said. I'll have
to swap a few cards and see how they do.

What's the usual symptom of overruning the video ?

 

[P2B]

Oh... I didn't know that. I thought you only had to set it for speeds
over 100mhz. I realize now that it should be 66. I've got it running
at 100mhz with my ATI Rage Pro... so I guess it will do 89 OK.

Your wording suggests (to me anyway) there may still be some confusion.

The AGP jumper can be set to 1/1 (intended for 66Mhz FSB), or 2/3
(intended for 100Mhz FSB). The AGP bus will run at the standard 66Mhz in
either case.

Are you saying you are running 100Mhz FSB with the AGP jumper set to
1/1? If so, I suspect that the 1/1 setting is really an "auto" setting,
and 2/3 is "force 2/3" - because I doubt any ATI Rage Pro would work at
all on a 100Mhz AGP bus. Most P2B boards (including all those I have on
hand) don't have the AGP jumper, so I can't test it's exact behavior
myself. Anyone?

But, I don't know if all my ATI cards will based on what you said. I'll have
to swap a few cards and see how they do.

What's the usual symptom of overruning the video ?

Many and varied - corruption on screen or frozen screen is most
common. When the screen freezes, the OS is usually still running and can
be shut down cleanly if you know the keystrokes.

A friend who recently did a Tualatin-at-133 upgrade on his P2B is
tolerating his ATI Rage Pro while he looks for a cheap nVidia-based card
because it mostly works - rapid scrolling through a large PDF document
will freeze it every time, but problems are rare in normal use.

I have a P2B-DS here running dual 1.4Ghz Tualatins and Solaris X86. It's
got an ATI Rage Pro because the OS didn't recognise anything else I had
on hand. There was some bizarre behavior when I first set it up - mouse
cursor would suddenly turn into a big grey square - which moved with the
mouse, but the system no longer responded to mouse clicks. Maybe the
card got used to the overclock - it's behaved itself perfectly for the
last few months

P2B

 

[joel]

Well, the differences between a CuMine Celeron and a Tualatin Celeron
(or a CuMine PIII, for that matter) aren't limited to the L2 cache
*size* alone:

* L2 cache size: 128K vs. 256K
* L2 cache associativity: 4-way vs. 8-way
* L2 cache latency: 11 cycles vs. 7 cycles

That's why a PIII-700 is about as fast as a Celeron 850; a Celeron 1.1
should be about as fast as a PIII-900 then. A Tualatin Celeron, however,
is pretty much exactly as fast as a Coppermine PIII at the same clock
speed, which is not too shabby.

Stephan

I recall that cache associativity stuff, they really hacked the thing
in half. I was wondering if having 512M ram memory and 8M hard disk
cache these days makes up for some of that latency.

 

[Bob]

Are you saying you are running 100Mhz FSB with the AGP jumper set to
1/1? If so, I suspect that the 1/1 setting is really an "auto" setting,
and 2/3 is "force 2/3" - because I doubt any ATI Rage Pro would work at
all on a 100Mhz AGP bus. Most P2B boards (including all those I have on
hand) don't have the AGP jumper, so I can't test it's exact behavior
myself. Anyone?

My mistake. It looks like my boards are all set for 2/3. I didn't do
that, so they either came that way from the factory or the folks who
originally built these set them to 2/3. I spec'ed these systems but
bought them from an integrator who gave me as good a package price
as I could get for the components and _they_ did the troubleshooting.
But, I never checked the AGP setting before now. My bad.

Many and varied - corruption on screen or frozen screen is most
common. When the screen freezes, the OS is usually still running and can
be shut down cleanly if you know the keystrokes.

<snip details>

Thanks,

 

[Roland Scheidegger]

I recall that cache associativity stuff, they really hacked the thing
in half. I was wondering if having 512M ram memory and 8M hard disk
cache these days makes up for some of that latency.

No. Cache reduces access times for ram, it doesn't matter how much ram
you have (having some uber-fast ram comparable with L2 cache
latency/bandwidth would help, but that just doesn't exist - L2 cache is
much faster, and in fact the performance difference between ram and
cache on a Piii platform will only get bigger with higher cpu frequencies).
And the hd cache only speeds up i/o access (which is orders of
magnitudes slower than memory access), nothing else.

Roland

 

[P2B]

My mistake. It looks like my boards are all set for 2/3. I didn't do
that, so they either came that way from the factory or the folks who
originally built these set them to 2/3. I spec'ed these systems but
bought them from an integrator who gave me as good a package price
as I could get for the components and _they_ did the troubleshooting.
But, I never checked the AGP setting before now. My bad.

Glad we straightened that out

One more thought - if your P2Bs are the versions with 4 FSB jumpers,
there is a 124Mhz FSB setting available which won't overclock the PCI
bus but will reduce AGP to 83Mhz. This might be a reasonable compromise
if you find the video card is preventing stable operation at 133Mhz FSB.

P2B

 

[Bob]

One more thought - if your P2Bs are the versions with 4 FSB jumpers,
there is a 124Mhz FSB setting available which won't overclock the PCI
bus but will reduce AGP to 83Mhz. This might be a reasonable compromise
if you find the video card is preventing stable operation at 133Mhz FSB.

P2B

Yeah... I see that. There's actually a 124 setting with the PCI at
31mhz or at 41mhz. I'll try it both ways. There are no SCSI drives on
these machines and I've had good luck with an overclocked PCI on my
machines as long as there were no SCSI components.

Is the 1.4g Tat-Cel known to run at higher bus speeds of 112, 115,
124, etc ? An sort of improved cooling needed ?

 

[P2B]

Yeah... I see that. There's actually a 124 setting with the PCI at
31mhz or at 41mhz. I'll try it both ways. There are no SCSI drives on
these machines and I've had good luck with an overclocked PCI on my
machines as long as there were no SCSI components.

Overclocking the PCI above about 37-38Mhz is not recommended due to the
high risk of corrupting data on IDE drives. Overclocked PCI doesn't
really improve performance, it's usually considered an undesirable side
effect of increasing FSB.

Is the 1.4g Tat-Cel known to run at higher bus speeds of 112, 115,
124, etc ? An sort of improved cooling needed ?

I haven't personally overclocked a 1.4Ghz Tualatin Celeron, but based on
my experience with other Tualatins and the results posted at
cpudatabase.com you could reasonably expect to run them at 1570Mhz
(112Mhz FSB) on default voltage and standard cooling. In my experience
Tualatins don't really respond to voltage increases, but it's worth
trying 1.55v if the processor is almost stable at a given speed.

P2B