northernosprey02 wrote:

... First, is it support OC well? I had OC my PII 350 to 466 but it was unstable (freeze when startup almost done and won't boot) ...

The most I managed from a PII Deschutes 350/100 was 434/124:

But a PIII Katmai 450 went to 600/133:

... Second, is this board still working on PC133 SDRAM? ...

Mine at least does 133, and then more:

... Third, can I forcing using PIII 800EB (133FSB) on this board? ...

I think there are good chances, perhaps a BIOS update would be required.

... Last, can it still running properly when I losted chipset heatsink? ...

Not a good idea if you're overclocking.

... BONUS: Why 440BX was superior chipset than i815 and i820? VIA Apollo Pro 133 was suck

The i820 was an epic fail (read here). Then Intel purposely handicapped the i815 to try to sell i820s. The end result was that good 'ol i440BX kept creaming them both.

Deschutes were not that good at overclocking. Only the later 333MHz models;

Towards the end of its design life, Deschutes chips capable of 500 MHz within Intel cooling and design specifications were produced. However, these were not marketed. Rather than destroy already multiplier-locked units, those Deschutes units that had been tested and locked with a multiplier of 5 were sold as being 333 MHz. This was accomplished by disabling the 100 MHz bus option. Overclockers, upon learning of this, purchased the units in question and ran them well over 500 MHz; most notably, when overclocking, the final batch of "333MHz" CPUs were capable of speeds much higher than CPUs sold at 350, 400, or 450 MHz.

From Wikipedia

"The i820 was an epic fail (read here). Then Intel purposely handicapped the i815 to try to sell i820s. The end result was that good 'ol i440BX kept creaming them both."

The big problem with i820 was that RDRAM and Pentium III were not a good fit and RDRAM was just too expensive. It wouldn't be until the Pentium 4 that RDRAM started to make more sense but by then DDR RAM was out offering higher performance than SDRAM and a lower price than RDRAM. RDRAM had more potential bandwith than DDR, but in the real world performance was not much better with RDRAM and not worth the extra cost.

There's some benchmarks in this review that demonstrate that in spite of having up to 50% more bandwith potential over DDR, RDRAM never outperformed DDR by that much.

http://www.pcstats.com/articleview.cfm?articleID=1096

NitroX infinity wrote:

Deschutes were not that good at overclocking. Only the later 333MHz models;

Towards the end of its design life, Deschutes chips capable of 500 MHz within Intel cooling and design specifications were produced. However, these were not marketed. Rather than destroy already multiplier-locked units, those Deschutes units that had been tested and locked with a multiplier of 5 were sold as being 333 MHz. This was accomplished by disabling the 100 MHz bus option. Overclockers, upon learning of this, purchased the units in question and ran them well over 500 MHz; most notably, when overclocking, the final batch of "333MHz" CPUs were capable of speeds much higher than CPUs sold at 350, 400, or 450 MHz.

From Wikipedia

What about Klamath? I also own PII 233 Klamath, is it was unlocked multiplier? But overclocking was no sense because it eating lot power 🤣

The P2B-F is a good 440bx board, just don't expect to run Tualatin's on it unless you have the Powerleap adapter. I personally think the successor, the P3B-F is superior to the P2B-F with its jumperfree BIOS, Tualatin voltage support and better (unofficial) support for 133mhz FSB.

More on the Rambus debacle from the superb Red Hill website.

"... Intel's ill-advised commitment to shoving Rambus RAM down the industry's throat, willing or not, had two direct results: it handed leadership in the CPU market to AMD, and number one spot in the chipset market to VIA."

It's interesting to draw a parallel with what is now (2013) happening with a certain OS ...

In theory, you can cram twice as much RAM into a 440BX board than you can in an i815 board – but then, depending on what you're using it for, you might not really need more than 512 MB anyway.

If you're gonna go P2/P3, you might as well make sure you have ISA slot(s). They can come in handy for what is perhaps the best SVGA DOS game platform.

NitroX infinity wrote:

Deschutes were not that good at overclocking.

SL2W8. 😳 300->504 baby.

As long as you are running PC133 RAM, the 440BX runs fine at 133FSB. Keep in mind that you will be overclocking the AGP bus to ~87Mhz due to it having a fixed 2/3 bus divider. Most decent video cards will handle this without complaining, but some will get pretty grumpy.

It won't be "most" cards that work ok at 90 MHz AGP. Some cards might work ok. Or they may seem ok but lock up once every half hour for example. It's just another way to make old junk more flaky.

I wonder if the difference between AGP and PCI with a processor in the 300MHz to 600MHz range makes a large difference. If not, you could go with a PCI videocard and not worry about the AGP-bus speed.

Sure. Going with PCI video with a 440BX @ 133 was something people did. AGP 2x really is of limited practical value, especially for 3D cards of the time.

I can vouch for some of the GeForce cards to be able to cope with the 87MHz AGP clock. I used to run my GeForce 2 Ti (don't remember the brand unfortunately) and my Creative GeForce 4 Ti 4200 on a 440BX board using a PIII 1GHz.
I think both setups were quite stable, as stable as Win98SE can be anyway 😜.

swaaye wrote:

It won't be "most" cards that work ok at 90 MHz AGP. Some cards might work ok. Or they may seem ok but lock up once every half hour for example. It's just another way to make old junk more flaky.

Yeah, I was got locked when I using S3 Trio3D/2X AGP.

TELVM wrote:

shoving Rambus RAM down the industry's throat

You really have a morbid fascination with the "shoving things down throats" metaphor...

...kinky!

I had a bunch of P2B, P2B-F, and P3B-F boards at one time. I sold most of them off. I wanted to keep 2 of them, so I tested the P2B-F and P3B-F boards for their overclocking. The winners were both P2B-Fs. That's maybe just luck of the draw but that's how it ended up for me.

My desktop 2 PCs ago was a P2B-F, 4x PC133 CL2 256MB Micron memory modules, Geforce4 Ti4200 128MB, P3-600E cB0 stepping at 800/133, an ATI video capture card, and of course a Sound Blaster AWE32. 😀 It was a good system.
I built a similar setup for a relative using a Geforce2 MX, same CPU type, RAM and overclock. Both tested stable at 840/140.
I filled in missing capacitors and replaced the chipset heatsinks with larger ones, but I don't think that really did much.

First thing you want to check is whether your board supports Coppermine voltages. You can tell that by checking the part number on the VRM chip, which is somewhere between the CPU slot and the peripheral ports. It's about 20 pins or so. If you don't know how to look up the info, post the part number here and you can get an answer.
If it's voltage compatible, then you just need to update the BIOS and you're good.
I wouldn't go over 100FSB without a heatsink on the chipset.

Some P2B board revisions had their FSB jumpers marked badly, but I don't remember what version that was. The problem was they marked how to get 133FSB with a 1/3 PCI ratio - so PCI ran at 44MHz. To get a 1/4 ratio requires a different jumper arrangement which wasn't marked on some boards. *most* are marked correctly though.

Remember that the AGP ratio is controlled by another jumper - make sure it's set to 2/3 if you run 100-133FSB.

89MHz AGP was 100% completely reliable for me with a Geforce2 MX, Geforce3 (different motherboard), and a Geforce4 Ti4200 - all 3 of those were in use at 89MHz simultaneously for 1-3 years on different BX machines. I've read that it can sometimes be a problem on some TI4xxxx cards, but mine never had a hint of trouble.
I've read that ATI cards of the time period were more problematic, but I never tried. nVidia FX5000 and later cards also reportedly don't like it.
I think ATI Radeon 8500 was known for handling 89MHz well, but like I said I never had an ATI on these.

You aren't guaranteed to hit 133FSB - some only pass stress testing at 124MHz reliably. Some can do it at 140MHz, maybe more, those are the boards I used for 133 operation. It just depends how lucky you are with your particular board. Later boards generally get faster.

Most brand-name PC100 CL2 memory is actually populated with 133CL3 RAM chips. I've even seen some that had PC133CL2 chips on them, but that's rare. Mostly just early or cheap PC100 modules are truly PC100 only (marked 8ns).
256MB modules need to have the right chip type (16Mx8) to be compatible, but the speed doesn't matter. PC133 itself doesn't make them incompatible, it's just that PC133 modules frequently don't have the right chip type so you have to be careful. I can explain in another post if you need more info.

These boards have trouble POSTing with a full 1GB of unbuffered PC133 memory. Intel wanted you to use registered memory with that amount. When I used 1GB PC133 CL2 on a P2B-F, I found that I needed to reset a few times until I saw the full 1GB appear at POST. Whatever RAM it saw at POST, was completely stable in operation - it was just a POSTing issue. This was the only nuisance I had with the machine.
Dropping to PC100, or removing 1 module, would eliminate the problem. I added some capacitors at empty spots on the board, and it seemed to help the POST only slightly, but it was far from a fix.

My favorite 440BX board for overclocking is probably the ABit BX133-RAID. It seems to be rare though, and I think every single one of them died of bad capacitors except those few that were recapped. But it's a socket-370, not a slot, and it only has 3 DIMM slots. It's a classic example of an ABit overclocking board, at a time when hotrodding the 440BX really made sense vs the chipsets that were meant to replace it. This was the board I put a Geforce3 on.
I think the BE6-II-RAID is very similar, but in Slot-1 format, and it's more common.

I have a P3B-F rev. 1.04 (no chipset heatsink), it runs the Pentium3-933 SL4BT at 133Mhz bus clock, with Radeon 8500 at 89Mhz, and everything works this way.

I had installing bolder heatsink, but I got some problem to installing Slot 1 CPU.

How about P2B-F multiplier limited to x8.0?

So some people said not all PC133 SDRAM support this, it is true?

Is PII Klamath unlocked multiplier?

_All_ Pentium II and III processors are multiplier-locked (as far as I know).