bloodem wrote on 2022-06-04, 19:19:

Possibly, but hard to say for sure without testing them myself. :-) Most slotkets (even generic ones that I have) worked fine fo […]
Show full quote

PcBytes wrote on 2022-06-04, 18:56:

So it then boils down to my slotket being trash or my RAM being even more trash I guess?

Possibly, but hard to say for sure without testing them myself. 😀 Most slotkets (even generic ones that I have) worked fine for me with lower power CPUs (like the VIA Ezra-T / Nehemiah).
However, when it comes to running Coppermines on Slot 1 boards, I always go with native Slot 1 CPUs. This has the benefit of eliminating a potential point of failure (the slotket) + it also looks nicer. 😁
If, on the other hand, a socket 370 CPU is all you've got, then it's better to search for a good slotket like the MSI MS6905 "Master" (especially revisions 2.0/2.3).

Regarding RAM, you must use a PC133 module and you should be fine @ 133 MHz FSB (as long as the RAM is not damaged).

PcBytes wrote on 2022-06-04, 18:56:

Mobos I used were as following - ASUS P2B, Soyo 6BA+IV and ABIT BE6-II Rev 1.0. The worst crashes were on the BE6 (which is now polymodded on the CPU side) which went as far as having to do several power cycles to get the 133FSB chips to POST ever again - 100 and 66FSBs recovered pretty fast, a restart and you would be back on track with the 100s, and the 66s were the "it just werkz" chips (both Mendocinos and a bunch of Coppermines between 566 and 700Mhz)

Yeah, I'm pretty sure that all of those boards would work just fine @ 133 MHz FSB with native Slot 1 CPUs and PC133 RAM.

PC133 sticks - check. As for running Slot 1 variants @ 133, that isn't an option either. Both my BE6 and the 6BA+IV choked when I tried to run my SL3XK (Coppermine, 650MHz) at 133FSB. Tried everything from generic PC133 up to heatsinked (yes, you read right, heatsinked) PC133 unbuffered RAM without any success.

PcBytes wrote on 2022-06-04, 19:24:

PC133 sticks - check. As for running Slot 1 variants @ 133, that isn't an option either. Both my BE6 and the 6BA+IV choked when I tried to run my SL3XK (Coppermine, 650MHz) at 133FSB. Tried everything from generic PC133 up to heatsinked (yes, you read right, heatsinked) PC133 unbuffered RAM without any success.

OK, you tried the SL3XK @ 133 MHz and it didn't work. What does that tell you? 😀 Instead of suspecting the motherboards, the more likely culprit is the CPU itself, which is a 100 MHz FSB model and by running it at 133 MHz you are heavily overclocking it by 33%. Some CPUs can handle that overclock without raising the voltage, some CPUs need additional voltage, while others will simply not work no matter how much voltage you feed them.
Try and search for a native 133 MHz FSB Coppermine CPU and see how it goes.

That...would be like finding the needle in a haystack. Most native Slot 1 chips I know of that I could find in my country are 100FSB (mostly Katmai chips) all around, save for two OEM chips (which I think were a dual CPU pair) that would cost me a small fortune. FCPGA chips however, they're absolutely everywhere, hence why it would be easier for me to get a slotket going.

And the thing here isn't that I try to run just the SL3XK at 133FSB. Running a native 133FSB chip through the slotket itself doesn't work either (eg. SL4MB, which is the P3 800EB, or even a SL5DV, which is a P3 1000B) across all three mainboards.

The P2B hangs during RAM count, the 6BA+IV hangs shortly after entering BIOS, and the BE6-II barely manages to show the XP bootscreen before either BSOD-ing or going in a reset cycle. What I can certainly say through this, is that I maintained the voltage always between 1.65v and as of lately, 1.8v. Maybe going with 2.0v would help? IIRC those Coppermine chips would take as high as 2.0v (if not higher, I haven't tried.) vcore.

PcBytes wrote on 2022-06-04, 19:44:

That...would be like finding the needle in a haystack. Most native Slot 1 chips I know of that I could find in my country are 100FSB (mostly Katmai chips) all around, save for two OEM chips (which I think were a dual CPU pair) that would cost me a small fortune. FCPGA chips however, they're absolutely everywhere, hence why it would be easier for me to get a slotket going.

True, I'm from the same country and I know that Slot 1 CPUs are getting very rare "pe plaiurile mioritice" 😁 . However, I'm pretty sure that, with a bit of patience, you can buy one on eBay (for a decent price).
EDIT: Did a 5 minute search and found a Coppermine 667/FSB 133 MHz SL3XL on Electromyne.com. You can use the "PHILSCOMPUTERLAB" discount code and it will only cost you 26,58 € (shipping included), which is pretty good I would say.

PcBytes wrote on 2022-06-04, 19:44:

And the thing here isn't that I try to run just the SL3XK at 133FSB. Running a native 133FSB chip through the slotket itself doesn't work either (eg. SL4MB, which is the P3 800EB, or even a SL5DV, which is a P3 1000B) across all three mainboards.

Yes, but by testing those socket 370 CPUs, you're now introducing a new variable: the slotket. 😀 So the comparison and conclusion are not valid. I even remember that 20 years ago, Intel actually created a list of known good slotkets, because many people had issues with the gazillion no-name models that flooded the market (that list probably still exists somewhere on archive.org).

There's a lot more variables that I do remember now not counting... but if anything, slotkets and Slot 1 CPUs are the hardest to find, at least in my town - the 650MHz was a pure lucky "trash" find no less - I wasn't even sure it was a Coppermine until I mashed the S-spec number in, though I was kinda reluctant to think it was a Katmai either since from what I knew those stopped at the 600MHz mark (or 550MHz was it?)

I wonder if I could get the BE6-II and the WS440BX running from those? I remember that one thing which I had issues with on the ABIT was primarily In-Order Queue Depth, and the chipset voltage. I wonder if 3.4v for the chipset would aid in running it at 133, of if I do need to bump it up ever so slightly. The issues I was having with those two points on the ABIT were that I couldn't run Geforce GPUs - these would cause the same crashes as if I was running the CPU at 133FSB - and the second was that I couldn't run the IOQD higher than 1 unless I set the chipset voltage to either 3.4v or 3.5v - once I set the chipset to 3.4v, I could run Geforce GPUs and run the CPU at its full speed (IOQD of 1 meant half the bandwidth if I recall correctly, which in turn made XP run absolutely crappy - IOQD of 8 means full bandwidth I think, because as soon as I got it to run stable with IOQD set to 8, XP began running decently, bar Firefox eating RAM like it was world hunger lmao), which makes me wonder if a combo of proper CPU voltage + tinkering with the chipset voltage and IOQD setting would equal stable 133FSB operation on the BE6.

PcBytes wrote on 2022-06-04, 20:04:

There's a lot more variables that I do remember now not counting... but if anything, slotkets and Slot 1 CPUs are the hardest to find, at least in my town - the 650MHz was a pure lucky "trash" find no less - I wasn't even sure it was a Coppermine until I mashed the S-spec number in, though I was kinda reluctant to think it was a Katmai either since from what I knew those stopped at the 600MHz mark (or 550MHz was it?)

See my post above, I edited it. Found a pretty cheap CPU that you can buy right now.

PcBytes wrote on 2022-06-04, 20:04:

I wonder if I could get the BE6-II and the WS440BX running from those? I remember that one thing which I had issues with on the ABIT was primarily In-Order Queue Depth, and the chipset voltage. I wonder if 3.4v for the chipset would aid in running it at 133, of if I do need to bump it up ever so slightly. The issues I was having with those two points on the ABIT were that I couldn't run Geforce GPUs - these would cause the same crashes as if I was running the CPU at 133FSB - and the second was that I couldn't run the IOQD higher than 1 unless I set the chipset voltage to either 3.4v or 3.5v - once I set the chipset to 3.4v, I could run Geforce GPUs and run the CPU at its full speed (IOQD of 1 meant half the bandwidth if I recall correctly, which in turn made XP run absolutely crappy - IOQD of 8 means full bandwidth I think, because as soon as I got it to run stable with IOQD set to 8, XP began running decently, bar Firefox eating RAM like it was world hunger lmao), which makes me wonder if a combo of proper CPU voltage + tinkering with the chipset voltage and IOQD setting would equal stable 133FSB operation on the BE6.

You don't need to raise any other voltages, you just need to eliminate unreliable components (first and foremost: the slotket). BTW, do you have a pic of that slotket you're using?

Sure. This is the slotket I use right now (see current slotket w/ 800EB pic). The reason I said there's a lot of variables I forgot counting - I had a previous slotket (much cheaper and had no DIP switches) which I first used to test 133FSB chips on and with which I had trouble.

This new one didn't get tested past the Soyo and ASUS P2B while the old one had been run across both the ABIT board and the Soyo (I didn't have access to the P2B until recently when I took it out from the Gateway GP6-400 case I need to restore). I had them mixed up since they don't really look too much different to me.

I did try this new one on the ABIT once, couldn't get it to POST with any chip, left it aside for time being. Came back to it just now (had to bring most of my stuff including the ABIT machine home) and tested it over on the Soyo (which is for now just a test bed for Coppermine chips, in the event I get the new slotket to properly operate - yes, it's recapped with polymers.) with most chips booting (except a SL5QW that quietly gave up the ghost it seems, as well as a 850MHz 100FSB which I suspect was broken to begin with...) fine - the issue I found with the new one in the Soyo so far, is that 133FSB causes file corruption, and rather frequent BSODs, and this was under 2000 SP4. For some strange reason, the 100FSB 1GHz Celeron (my only 100FSB FCPGA at the moment) exhibited the same issue, though I don't know if the Soyo just legitinately hates running Celerons at anything higher than 66FSB, despite the chip I used being purposed to be natively run at 100FSB - I'll have to scrounge some 100FSB P3s to test.

That, and then there's Tualatin (or at the very least Tualeron) operation which I want to tackle next. Hence why I'm keeping on trying to get the slotket running stable 133FSB - once I can have 133 Copermines running stable on it, Tualatins come next, and I can have a swiss army knife of a slotket that hopefully cab run anything I throw at it, from Mendocinos up to and including Tualatins.

bloodem wrote on 2022-06-04, 19:53:

Yes, but by testing those socket 370 CPUs, you're now introducing a new variable: the slotket. 😀 So the comparison and conclusion are not valid. I even remember that 20 years ago, Intel actually created a list of known good slotkets, because many people had issues with the gazillion no-name models that flooded the market (that list probably still exists somewhere on archive.org).

see attached:

I have Gigabyte GA-6BXC, which is early slot 1 board, originally released with only Pentium II, and I've bought it with Tualatin Celeron 1.4 Ghz inside.
So that's how "unusable" are 1.10 revisions P2B and similiar early Pentium II slot 1 BX boards
every slot 1 board is worth saving, particulary with BX chipset.

PARKE wrote on 2022-06-05, 01:26:

see attached:
Intel.pdf

Thanks! That's the one, yeah! 😀

PcBytes wrote on 2022-06-04, 20:29:

Sure. This is the slotket I use right now (see current slotket w/ 800EB pic). The reason I said there's a lot of variables I forgot counting - I had a previous slotket (much cheaper and had no DIP switches) which I first used to test 133FSB chips on and with which I had trouble.

Yeah, those slotkets seem to be worse than even the ones Intel lists in their document as "not recommended for Pentium 3". 😀

If you want to actually find the source of your issues, you must go with a basic configuration that doesn't involve any third party components.
What I would do:
- choose a known good 440BX motherboard that has Coppermine support (i.e.: supports the Coppermine voltage without any tweaking and has a BIOS revision from the mid - late 2000).
- reset the CMOS defaults
- test with a Slot 1 CPU that runs at 133 MHz FSB natively.
- use the most compatible RAM that I have, one that I know for sure that it works with many other systems.
- use a video card with low power requirements (GeForce 2 MX or GeForce 2 GTS at most).
- don't add any other peripherals!
- do a fresh Win98SE install (NO SERVICE PACKS!). Win98 is the most compatible OS for this type of system, so I would do my initial tests on it (even if I'd want to go with a more modern OS later on).
- install just the bare minimum: DirectX 7.0, 440BX chipset drivers, nVIDIA driver version 7.76.

If you do all of these things, you'll most likely end up with a rock solid PC. Once that happens, you can start to gradually change things and see which change causes problems.

bloodem wrote on 2022-06-05, 10:47:

Yeah, those slotkets seem to be worse than even the ones Intel lists in their document as "not recommended for Pentium 3". 😀

One point worth mentioning is that both PcBytes' slotkets were designed for Mendocino Celerons. They are not necessarily better or worse (although some are undoubtedly worse) than the later crop that was designed for Coppermine but Intel would most likely never approve of this type of hacked product in the first place.
The criteria that Intel applied for their testing are unknown but it is certain that these criteria were harder to meet than via playing a couple of 3D games without hickups. (And when you take it to the next level it is certain that Intel would not approve of running Tualatins on any 440BX boards simply because they cannot supply the correct core voltage - in other words every Tualatin runs outside of Intel specification on 440BX boards.)

I agree with your assessment: if you want to leave the experimental phase behind bite the bullett and get a decent 'top 10' slotket plus a board that is reputed to run everything that is thrown at it in a stable way. Here an example of an ASUS P3B-F with PIII 850/100@1.133/133 on an Iwill Slocket II ver 2.0 slotket.

PARKE wrote on 2022-06-05, 12:08:

One point worth mentioning is that both PcBytes' slotkets were designed for Mendocino Celerons. They are not necessarily better or worse (although some are undoubtedly worse) than the later crop that was designed for Coppermine but Intel would most likely never approve of this type of hacked product in the first place.

Indeed!

PARKE wrote on 2022-06-05, 12:08:

[...]Intel would not approve of running Tualatins on any 440BX boards simply because they cannot supply the correct core voltage - in other words every Tualatin runs outside of Intel specification on 440BX boards.

Well, if a board supports Coppermine voltages, it also suports Tualatin voltages.
Boards that don't support Coppermines can only supply voltages between 1.8 - 3.5, while boards with Coppermine support can provide 1.3 - 3.5V. The Tualatin requires ~ 1.45V, so it falls within this interval.
So the voltage itself isn't a problem for Tualatin CPUs (all Coppermine compatible boards can supply the correct voltage to Tualatin CPUs - which is why CPU's like the VIA C3 Ezra-T and Nehemiah work just fine in FC-PGA boards). The reason why Tualatins are not compatible with older boards is that Intel changed some pins... just because they could (to make the CPUs incompatible with older boards).

bloodem wrote on 2022-06-05, 12:28:

Well, if a board supports Coppermine voltages, it also suports Tualatin voltages. Boards that don't support Coppermines can only […]
Show full quote

PARKE wrote on 2022-06-05, 12:08:

[...]Intel would not approve of running Tualatins on any 440BX boards simply because they cannot supply the correct core voltage - in other words every Tualatin runs outside of Intel specification on 440BX boards.

Well, if a board supports Coppermine voltages, it also suports Tualatin voltages.
Boards that don't support Coppermines can only supply voltages between 1.8 - 3.5, while boards with Coppermine support can provide 1.3 - 3.5V. The Tualatin requires ~ 1.45V, so it falls within this interval.
So the voltage itself isn't a problem for Tualatin CPUs (all Coppermine compatible boards can supply the correct voltage to Tualatin CPUs - which is why CPU's like the VIA C3 Ezra-T and Nehemiah work just fine in FC-PGA boards). The reason why Tualatins are not compatible with older boards is that Intel changed some pins... just because they could (to make the CPUs incompatible with older boards).

'Within this interval' is not what I meant - for specific information on >specification< for Tualatin voltage please read this thread:
https://forums.overclockers.com.au/threads/sl … slotket.164450/

PARKE wrote on 2022-06-05, 13:36:

'Within this interval' is not what I meant - for specific information on >specification< for Tualatin voltage please read this thread:
https://forums.overclockers.com.au/threads/sl … slotket.164450/

Well, I was talking about VCORE, not the AGTL signal, but I get what you meant.

Maybe I focused too much on Intel specification but that is what triggered this part of the conversation. In other words, I am sure that many people successfully run Coppermines on slotkets that were -not- approved by Intel in that article. But there is a difference between 'running' and 'running in spec'.