The relevant point here isn't some magical 'server class' stuff, it's the basics:

- this CPU has an FC-PGA2 pinout. FC-PGA2 sockets are backwards compatible with FC-PGA CPUs, but FC-PGA2 CPUs aren't so you need a motherboard with an FC-PGA2 socket, a slocket with FC-PGA2 or you need to hack CPU or motherboard.
- this CPU has (like late regular P3 and indeed Celeron CPUs) a Tualatin core. The system's BIOS needs to be able to work with that.
- Tualatin CPUs nominally run at 1.45V, as opposed to 1.60-1.75V for older Coppermine CPUs. Your motherboard needs to be able to provide this voltage and do so in a stable way with sufficient current for the CPU to work.

In addition, to run at full speed the motherboard needs to support 133MHz FSB, but anything with an FC-PGA2 socket would be able to do that. That motherboard may be "PCI only" (almost certainly it's not, it only has PCI slots implemented...) too. It would have been helpful if you had shared with us exactly which motherboards you have, then we could have been able to say if they could out-of-the-box and possibly whether modding would be an option.

All of this would apply just as much to say a Celeron 1200, so again it's nothing to do with "server". There's also no such thing as a "standard" So370 board, as there were three different pinouts in So370 (PPGA, FC-PGA and FC-PGA2), three different CPU core versions, three different FSB speeds and three different voltage ranges.

This is why I ask. Thanks for the information! 😀

The board is an HP 216906-001 (Compaq) 801E Chipset. Yes, it has on-board AGP graphics, but that's all I know.

The closest thing I can find on Retro Web is... https://theretroweb.com/motherboards/s/lite-on-tr101

But, it's not an exact match. It came out of a Compaq desktop. Perhaps these were made for them by Lite On?

Not a lot of info on it that I can find. The chipset seems to support Tualatin and 133MHz and the manual for the TR101 says it supports Celeron / Pentium III / Tualatin. However, it also says "Processor integrated on-die 32KB level-one (L1) and 128KB/256KB level-two (L2) cache.". Sadly, nothing on voltages.

It looks like a variant of this board:
https://theretroweb.com/motherboards/s/bcm-in810ep

smtkr wrote on 2025-02-02, 23:15:

It looks like a variant of this board:
https://theretroweb.com/motherboards/s/bcm-in810ep

Much closer, you're probably right. Just lacking the AGP expansion.

That being said, it doesn't seem to support the CPU in question.

Guess I'll just hold on to it for now. 😀

StriderTR wrote on 2025-02-03, 00:02:

smtkr wrote on 2025-02-02, 23:15:

It looks like a variant of this board:
https://theretroweb.com/motherboards/s/bcm-in810ep

Much closer, you're probably right. Just lacking the AGP expansion.

What AGP expansion? This board just has PCI and AMR. If your board has AGP, it's not the board in the pic you posted earlier.

The only differences I see vs the earlier pic is that on the first one the AMR slot isn't populated but the 4MB display cache next to the motherboard is. So these are the same PCB even if a few components differ. It is possible that the socket pinout is also different, but I wouldn't bet on it.

That being said, it doesn't seem to support the CPU in question.

Guess I'll just hold on to it for now. 😀

Tualatin support is rare, as it was only introduced at the very end of the P3 lifecycle, after the introduction of the P4. Compaq definitely did use uATX boards that support it as I have one of them, a late revision Compaq Garcia (P/N 240762-001)

There were certainly late i810E boards that supported Tualatin (although I've never seen one with the 4MB display cache on it - by this time this was a really low-end solution and any corners that could be cut were). It wouldn't hurt to try the CPU in this board. Worst-case it just doesn't boot. That said, i810 boards are deliberately castrated low-end solutions (100MHz memory bus even if CPU is running at 133MHz...) that wouldn't come close to giving a P3-S a decent showing. I'd recommend looking for a better board for the CPU even if it does run on here. For Win98SE, you can't beat a late i815 board, for Win2k/XP a Via ApolloPro133T board would be better (slower clock-for-clock than i815, but able to install three times as much RAM).

That's my bad, my bad eyes mistook the AMR for AGP. Thanks for pointing that out.

Yeah, I think I'm going to look for a different setup for it. I'm in no real hurry, was just going through what I have here. I'll piece something together eventually. 😀

On a side note...

What sucks is I have a Radeon X700 Pro 256MB card, but it's PCIe. I have a PCIe board with an Athlon 64 3200 on it. I've just never tried to use PCIe with Win98SE. I read it can be done, just not sure I want to fudge with it.

Right now I'm trying to use a Sempron 2500+ and Radeon 9250 AGP, not having much luck. 9250 may be flaky. Kicked my backside all day today.

dionb wrote on 2025-02-03, 07:42:

That said, i810 boards are deliberately castrated low-end solutions (100MHz memory bus even if CPU is running at 133MHz...) that wouldn't come close to giving a P3-S a decent showing.

You could argue that the extra cache on the P3-S would help mitigate the low memory bandwidth, though 😉

H3nrik V! wrote on 2025-02-03, 09:05:

dionb wrote on 2025-02-03, 07:42:

That said, i810 boards are deliberately castrated low-end solutions (100MHz memory bus even if CPU is running at 133MHz...) that wouldn't come close to giving a P3-S a decent showing.

You could argue that the extra cache on the P3-S would help mitigate the low memory bandwidth, though 😉

True, it would be just that little bit less awful than the regular P3 at same speed 😜

Actually the 4MB display cache probably has a bigger impact, as without it that 800MB/s memory interface would need to be shared fully between CPU and IGP, so the CPU would only get about 400MB/s of the 1066MB/s it wants.

dionb wrote on 2025-02-03, 07:42:

StriderTR wrote on 2025-02-03, 00:02:

Tualatin support is rare, as it was only introduced at the very end of the P3 lifecycle, after the introduction of the P4. Compaq definitely did use uATX boards that support it as I have one of them, a late revision Compaq Garcia (P/N 240762-001)

Random fact: there were s479 tualatins. Something I only learned recently. Compatibility is a question I haven't yet figured out.

myne wrote on 2025-02-03, 10:20:

[...]

Random fact: there were s479 tualatins. Something I only learned recently. Compatibility is a question I haven't yet figured out.

According to CPU World: not.
https://www.cpu-world.com/Sockets/Socke ... 479M).html

The Pentium III-M CPUs were manufactured 478-pin package, but the socket itself had 479 pin holes, which made it easier for Intel to re-use the socket for other generations of microprocessors. Over the course of the next 5 years Intel introduced two other generations of mobile microprocessors that utilized the same socket, but were not compatible with each other or with Pentium III-M motherboards.

The question is how big the delta is. PPGA -> FC-PGA and FC-PGA->FC-PGA2 are so similar they can be modded easily (which begs the question why Intel didn't just stick with the same PPGA layout...), but the Pentium M shared the quad-pumped bus of the Pentium 4, so I doubt it would be easy to connect a Pentium 3 to that. I have an So479 motherboard (for Pentium M) so could try, but without any expectation of success...

StriderTR wrote on 2025-02-02, 20:14:

So, going through stuff from a previous haul, I found a Pentium III-S 1333/512/133 processor. […]
Show full quote

So, going through stuff from a previous haul, I found a Pentium III-S 1333/512/133 processor.

https://www.cpu-world.com/CPUs/Pentium- ... 3512).html

I have a couple PCI only Socket 370 boards, both have slower Celeron processors in them.

The question is, will the "Server" class P3 work in a standard 370 board that supports a standard P3 1133? From what I recall, and read, the only difference was 512K L2 and support for dual processor boards? Or, am I mistaken?

I don't want to drop it in a board and risk damage due to my lack of knowledge of these processors.

Tualatin CPUs use a slightly different pinout (FPGA2) than normal Pentium II CPUs (FPGA), and lower voltages.
Besides that, there's no special BIOS or chipset support required: Back in those days there were adapters sold for running a Tualatin in a Slot 1 board. The web was full of instructions on how to modify FPGA socket to FPGA2 - all you need are some thin wires and a steady hand.
The "S" Tualatins run with 133 MHz FSB. Your board might max out at 100 MHz, so you loose 25% performance.

If in doubt simply put the CPU in and try if the machine runs. Either the CPU gets detected or not. Nothing will burn.

dionb wrote on 2025-02-03, 12:42:

According to CPU World: not. https://www.cpu-world.com/Sockets/Socket%20479%20(mPGA479M).html […]
Show full quote

myne wrote on 2025-02-03, 10:20:

[...]

Random fact: there were s479 tualatins. Something I only learned recently. Compatibility is a question I haven't yet figured out.

According to CPU World: not.
https://www.cpu-world.com/Sockets/Socke ... 479M).html

The Pentium III-M CPUs were manufactured 478-pin package, but the socket itself had 479 pin holes, which made it easier for Intel to re-use the socket for other generations of microprocessors. Over the course of the next 5 years Intel introduced two other generations of mobile microprocessors that utilized the same socket, but were not compatible with each other or with Pentium III-M motherboards.

The question is how big the delta is. PPGA -> FC-PGA and FC-PGA->FC-PGA2 are so similar they can be modded easily (which begs the question why Intel didn't just stick with the same PPGA layout...), but the Pentium M shared the quad-pumped bus of the Pentium 4, so I doubt it would be easy to connect a Pentium 3 to that. I have an So479 motherboard (for Pentium M) so could try, but without any expectation of success...

Yeah, I know what the official answer is. I just haven't spent the time looking at datasheets to figure out to what extent it's true - or where compatibility starts and ends.
Maybe it'll work as far back as LX with a slotket, maybe there's some signaling difference that will limit it to 810 without excessive effort.
Or maybe there's some undocumented "power/ground" pin that's pulled to VCC/VSS on the p4 that says "use quad pumped" that's not on the Tualatin.

I'll probably look one day.

"S" stands for 512KB L2 cache.
Regular PIII Tualatins had 256KB with 133MHz FSB.
Some well-known boards like TUSL2-C do not have microcode for "S" processors with latest BIOS, although CPUs work. You do get error during POST.
On P3B-F with latest beta BIOS it is recognized as Pentium II.