So far It is not looking good 🙁

I've tested P3 600/100mhz and it is dead same as 866 p3. I expected this one would work for sure as it has supported parameters. Also, It turns out the slotket I borrowed doesn't support coppermine as well... But at least I have another slotket to test. As this new slotket is now a lot less "precious" I've got the owner's permission to modify it. So that's what I'll do next.

I haven't checked Cpus other than p3 600mhz, because I thought there is a small chance my slotket is broken and frying all those cpus. So I didn't want to risk all the cpus I have.

Now I start to question why my Slot1 Celeron 266/66mhz doesn't work too. Previously I assumed it must be broken if slot1 p2 works fine, but now I'm wondering perhaps the Celeron cpu is good, but there is something in this board that stops it from working too.

Another difference between the p2 and all other cpus is that p2 has a factory made plastic frame that holds it in the slot. All the other cpus don't and they are held in the slot just by the pcb. Perhaps there is a contact that fails to connect unless the cpu is "exactly" right? (I tried holding the cpu board with my hand leaning it up or down when powering on, no difference).
I've got some contact cleaners so I'll be using them today.

Also I have to compare slot1 p2 pinout with Celerons to see if there is any difference. If there is any way a MB might work with a p2, but not Celeron? All those "slotket modifications" are based on comparison's between a Celeron and a P3 pinouts. Perhaps that's a clue of some sort?

And finally if all that fails I might start soldering probes to the slotket to check the fsb frequency is generated, reset happens and is released and anything else I can think of.

Also, as mentioned before the typical "coppermine mod" involves two things (connect resets, remove pin ah2), but the list of pin differences between P3-Celeron includes two more. An extra resistor pull down on two pins (rttctrl and slewctr). It is usually not done, because it was found to be unnecessary, but I thought there is no harm in trying so I added it to my slotket (after finding out P3 600 doesn't work). Tom's HW article mentions resistors of 330ohm, but intel's platform guide says 56 ohm and 110ohm. I followed Intel. I doubt it would make a big difference, but I might change it to 330.

So now, we have 3 possibilities :
- MB works only with p2 (weird)
- All my 370 cpus are dead and sellers are bulls****ing!
- My slotket, or the slot1 has something not contacting right/slotket has a design/manufacturing defect that prevents it from working.

If I was really desperate I could trace every single PGA370 pin to every slot1 pad with an ohm meter. That's probably too much effort.

I thought one of cpus you were buying was slot1. Too many unknowns and variables, maybe find someone local with another slot1 board to run some tests.

Please see edit2, I did manage to get it sort of working.

rasz_pl wrote on 2022-11-26, 10:53:

I thought one of cpus you were buying was slot1. Too many unknowns and variables, maybe find someone local with another slot1 board to run some tests.

No, no slot1 cpus. I already have a p2, slot1 and a Celeron 266 the owner claimed "was pulled out from a working machine", but of course only p2 works. P3 slot1 are quite expensive just for a test.

None of P3s pga370 I now have work. The other slotket is a high-er quality gigabyte one. However the one that supports coppermine is GA-6R7+ while this one is GA-6R7 (no plus). All I did to it was a classic mod of disconnected ah2 and connected reset2 and reset. How can anyone capable of soldering screw it up? Here is a photo. Perhaps someone else can see something I can't.

The blue thing in the ah2 pin is a piece of plastic wire isolation. Solder was sucked out from the pin hole and the isolation added to prevent the pin contacting the walls. The mod was tested with a multimeter and is reversible this way (rather than breaking off the pin). One can reverse it by pulling out the blue plastic and soldering the pin again.

At this point I'm willing to accept this chipset will not work with P3 probably no matter what. I never even read about anyone successfully running P3 on via apollo pro plus. Everyone was using via apollo pro 133.

I don't really even need a faster cpu in this build(as I'm planning a WinXP build that's faster) , but there is a "cool factor" in getting the hardware do what you want to exceed the limitations imposed by the designers. Unfortunately on this occasion it looks like a fail.

Or perhaps I'm unlucky for slot1 stuff. I remember back in the day my first proper pc was a Celeron slot1 (probably 266). I remember at some point it got crashy and eventually it wouldn't even post. I ended up having to install a rubber band that held the cpu board in tension in the slot or it wouldn't boot... Then every few months I woukd replace the rubber band when it lost its flex.
That was probably a sign to not touch slot1 in future 😉

Edit: one more thing, on theretroweb.com this MB is listed as only working with pentium 2 (not even Celeron). I thought this was a mistake, but perhaps it is exactly right. Unfortunately the manual is not even specific for this MB. There is one manual for 3 versions of the MB so one cannot know which statements apply to which MB.

Edit2: Something works! 😀

I was wondering if I should remove the above, but I decided to keep it for educational reasons.

The "new" gigabyte slotket has a single jumper that selects 66mhz,100mhz or 133mhz. When the jumper is removed it is supposed to be 100mhz (this was also confirmed by my testing with a multimeter). However, before giving up I tried to set the jumper to 66mhz again (not expecting it to work as the chips next to thd jumper are not populated), but 866/133mhz cpu actually posted and was recognised as a p3(coppermine) 600mhz! (this is a patched bios).

Now the question becomes how do I verify actual fsb... With 386 etc I can just run a simple benchmark to get cpu frequency. Same on P3 will just show huge meaningless numbers. Also, setting the jumper to 66mhz unlocked full bios fsb settings. I can now set fsb from 66 up to 150mhz(it fails to boot on 133,I tried, but runs fine on 100mhz in bios with a jumper on 66)

Wow, that's a useful first step!.

flynth wrote on 2022-11-26, 11:54:

Edit2: Something works! 😀
8><CUT
Wow, that's a useful first step!.

Congrats ! 😀
Just curious.... what do the markings on the frequency generator chip say (I can't read it from the photos)

PARKE wrote on 2022-11-26, 13:15:

Congrats ! :-) Just curious.... what do the markings on the frequency generator chip say (I can't read it from the photos) FREq […]
Show full quote

flynth wrote on 2022-11-26, 11:54:

Edit2: Something works! 😀
8><CUT
Wow, that's a useful first step!.

Congrats ! 😀
Just curious.... what do the markings on the frequency generator chip say (I can't read it from the photos)
FREqgen.JPG

Thanks 😀

The chip appears to be
W144H
8955 1907FD

Now, it appears all those PGA370 P3 Cpus work. 866 is the fastest at 650Mhz (I've yet to try OCing the rest of the MB, at 133 it failed to post, but not like before where it was 00. It ended at what looked like P0 so much better.)

So it was a dodgy slotket and a broken Celeron in the end.

The multiplier setting doesn't matter at all on those locked cpus.

I also found a DOS tool that shows me the real cpu frequency and it confirms that the fsb is indeed 100mhz despite the jumper being on 66mhz (100mhz is set in bios).

So, thank you everyone that helped me to get here 😀

https://pdf1.alldatasheet.com/datasheet-pdf/v … RESS/W144H.html no readout function
you can try 105 112 124 fsb

>dodgy slotket and a broken Celeron

my first and third guesses 😜

>The blue thing in the ah2 pin is a piece of plastic wire isolation. Solder was sucked out from the pin hole and the isolation added to prevent the pin contacting the walls. The mod was tested with a multimeter and is reversible this way (rather than breaking off the pin). One can reverse it by pulling out the blue plastic and soldering the pin again.

why would you want to reverse it? it was a useless pin before coppermines, and in coppermines it was used to lock out older boards/slotkets. Afair it has no other function.

rasz_pl wrote on 2022-11-26, 13:58:

https://pdf1.alldatasheet.com/datasheet-pdf/v … RESS/W144H.html no readout function
you can try 105 112 124 fsb

The manual for this motherboard mentions the following:
oooooooooooooooooooooooooooooooo
irst enter/choose "BIOS" System. Second, choose " CPU Host clock" which is under the "Chipset Feature Setup ". Third , you can change the Frequency with the follaring numbers such as 66, 75, 83, 100, 112 and
133 for replacement.
oooooooooooooooooooooooooooooooo
This suggests that only frequency settings can be chosen from the section for which input address FS3 = 0.
The 133 mhz setting in this section has a PCI divider of 3 resulting in a PCI fequency of 44.4 mhz. It is not unlikely that this settng results in instability of the system. I ran into the same problem with an ASUS P2B motherboard.

PARKE wrote on 2022-11-26, 16:15:

The manual for this motherboard mentions the following: oooooooooooooooooooooooooooooooo irst enter/choose "BIOS" System. Second […]
Show full quote

rasz_pl wrote on 2022-11-26, 13:58:

https://pdf1.alldatasheet.com/datasheet-pdf/v … RESS/W144H.html no readout function
you can try 105 112 124 fsb

The manual for this motherboard mentions the following:
oooooooooooooooooooooooooooooooo
irst enter/choose "BIOS" System. Second, choose " CPU Host clock" which is under the "Chipset Feature Setup ". Third , you can change the Frequency with the follaring numbers such as 66, 75, 83, 100, 112 and
133 for replacement.
oooooooooooooooooooooooooooooooo
This suggests that only frequency settings can be chosen from the section for which input address FS3 = 0.
The 133 mhz setting in this section has a PCI divider of 3 resulting in a PCI fequency of 44.4 mhz. It is not unlikely that this settng results in instability of the system. I ran into the same problem with an ASUS P2B motherboard.

The BIOS shows the pci clock of 33mhz for the following fsb settings. 66/100/133 as if it was dividing it by 4 at 133. Perhaps it is just a label and really it is 44.4mhz which would explain failure to boot. I wonder if there is a way to verify that without measuring it directly.

Also, about the manual for this motherboard. Unfortunately it is useless (even worse than useless, it is misleading), because there is one manual for at least 4 versions of the motherboard with chipset ranging from via apollo Pro133, via apollo pro plus, zx, another one I don't remember etc. Additionally the BIOS that is supposed to be for this very board works only in 2 oldest versions. 2 newer versions fail to post at all (00 on post card again).

However, having said all that about the manual I'm still pretty happy with the motherboard.

Interestingly I experienced horrible GPU instability with p3. I immediately thought about what rasz_pl said, but I went through the usual troubleshooting and I found the cause to be lack of pressure on the slot1 connection.

Specifically, the instability manifested itself by 3d mark99 exiting the benchmark early (often on first or second test). I run 3dmark99 running on a loop for a whole day before with no issues so I knew it is capable of stability on 100mhz fsb. But I even saw the same thing when I popped back by p2 cpu. Then after I reseated the cpu, making sure to "clip" the little plastic catches the problem went away on p2.

So then I've put the P3 back again and this time I was physically pushing the slotket into the motherboard throughout the whole test. It did finish fine. Unfortunately I can't manually keep pushing it for hours to do a proper stability test, but I think I now know what's going on.

I definitely need something that will hold that cpu slot in tension. Like the rubber band I had all those years ago 😀

Ideally I would 3d print a proper frame, but there are no ready made models online and rather than expend that much effort I might get the se result with some zip ties...

Results wise my mx440 scored 4000 points and there are 8700 cpu points. So it is appreciably faster.

flynth wrote on 2022-11-26, 17:41:

The BIOS shows the pci clock of 33mhz for the following fsb settings. 66/100/133 as if it was dividing it by 4 at 133. Perhaps it is just a label and really it is 44.4mhz which would explain failure to boot.

Yes, I wondered about that as well - which setting is the BIOS addressing in reality ? When you have a BIOS or a set of jumpers that offers both choiches you could count on it.
For your system I would try to score a Slot 1 800MHz fsb 100.
They are not cheap but there are not that many around so it is not a bad investment imo.

PARKE wrote on 2022-11-26, 17:58:

Yes, I wondered about that as well - which setting is the BIOS addressing in reality ? When you have a BIOS or a set of jumpers […]
Show full quote

flynth wrote on 2022-11-26, 17:41:

The BIOS shows the pci clock of 33mhz for the following fsb settings. 66/100/133 as if it was dividing it by 4 at 133. Perhaps it is just a label and really it is 44.4mhz which would explain failure to boot.

Yes, I wondered about that as well - which setting is the BIOS addressing in reality ? When you have a BIOS or a set of jumpers that offers both choiches you could count on it.
For your system I would try to score a Slot 1 800MHz fsb 100.
They are not cheap but there are not that many around so it is not a bad investment imo.

Yes, I'll keep looking for one.

A week ago I ordered one 933/133 P3. This one would give me 700mhz at 100mhz fsb and I got it really cheap, but it has been a week and it hasn't been posted so it might not happen. Unfortunately, I've found this quite common in this hobby. Sometimes I find something really cool for very cheap and then the seller can't be bothered to post it...

In the meantime my "zip tie" trick seems to be working. 3dmark99 has been going on loop for ~4.5h without a crash. I'll probably leave it overnight just to be sure, but that looks pretty nice. Also I swapped the old whiny fans with new ones so it is much quieter. Still it is not as quiet as it could be. Unfortunately finding single Fan cheap pwm controllers proved more challenging than I thought.

W144 clock gen supports CPU/4 just fine. Table in datasheet lists all fsb/pci combinations

rasz_pl wrote on 2022-11-26, 23:33:

W144 clock gen supports CPU/4 just fine. Table in datasheet lists all fsb/pci combinations

Good to know. Although via said this chipset is not supported at 133 they used to sell it as "133mhz capable". So I think it should work with 133mhz(even if not very well).

According to via datasheets both chipsets support same read sdram timings as 440bx (x-1-1-1-1-1-1-1), but while searching for more info with Google setting custom date range to before 2002 I found one article that claimed via apollo pro 133 uses x-1-6-6-6-6-6-6 at 133mhz. So perhaps dropping those timings might help.

flynth wrote on 2022-11-27, 09:35:

Good to know. Although via said this chipset is not supported at 133 they used to sell it as "133mhz capable"

I have a slot/socket370 combo motherboard (P6BAT-A+ with via chipset), supporting 133 mhz on paper (via apollo pro 133). Never succesfully managed to run it at 133 or anything above 110, the hdd controller and io becomes unstable, getting dma errors and transfer all the way (dma errors are everywhere even at 100 unless udma is disabled in bios and in the system). I recommend to clock it at 100 and use it like that.

Geri wrote on 2022-11-27, 10:42:

flynth wrote on 2022-11-27, 09:35:

Good to know. Although via said this chipset is not supported at 133 they used to sell it as "133mhz capable"

I have a slot/socket370 combo motherboard (P6BAT-A+ with via chipset), supporting 133 mhz on paper (via apollo pro 133). Never succesfully managed to run it at 133 or anything above 110, the hdd controller and io becomes unstable, getting dma errors and transfer all the way (dma errors are everywhere even at 100 unless udma is disabled in bios and in the system). I recommend to clock it at 100 and use it like that.

It must have been quite disappointing to get a 133mhz board and have to run it at 100. At least I knew my chipset is not supported at 133 so if it run it woukd be a bonus rather than an expectation. (this is via apollo pro plus, it can allegedly do 133,but only via chip sets from via apollo Pro133 were "supported" at 133).

In general I think a lot of Via chipset's bad reputation came from them pushing to release a 133mhz capable chipset as soon as possible. I read some article from 1998 that claimed via had the first generally available 133mhz fsb chipset at the time. I imagine they probably had apollo pro plus running internally at 133mhz in an ideal environment and they thought "let's repackage this as our 133mhz chipset a year before anyone else". Then they sold it cheap as pro133 to cut out the competition, but as a result lots of shoddy manufacturers made boards with bad designs, low quality substrates etc. Lots of those crappy boards from brands like Ezza and others gave them a bad name. Also, at the time(I suppose due to its availability) even quality brands like Asus released motherboards with via apollo Pro plus and Pro133. So they couldn't have perceived it as a shitty product or they wouldn't risk their brand on it.

Then the reality came. Slot1 as a cpu mounting method had to cope with people sticking various low quality slotkets in there (some with no, or bad quality frames). Also even original Slot1 celerons had frames that didn't always provide tension from the factory (see my rubber band mod). And so on. People had lots of bad experiences.

So why are we using those old slot1 systems now? I can only speak for myself and to me the appeal is not just that this was my first pc platform, but also that it is the last AT form factor. And that it is the only cpu slot/socket one can go as low as 133mhz cpu/66fsb and as high as 1.4ghz (with tualatin) at 100mhz fsb. Such a wide range of performances is pretty cool to have to me.

flynth wrote on 2022-11-27, 11:36:

In general I think a lot of Via chipset's bad reputation came from them pushing to release a 133mhz capable chipset as soon as possible.

VIA KT333 supported 333 mhz fsb.
And they repeated the mistake.. khm.. scam again.
Guess what VIA KT400 supported: up to 333 mhz only 😁

Geri wrote on 2022-11-27, 22:57:

VIA KT333 supported 333 mhz fsb. And they repeated the mistake.. khm.. scam again. Guess what VIA KT400 supported: up to 333 mh […]
Show full quote

flynth wrote on 2022-11-27, 11:36:

In general I think a lot of Via chipset's bad reputation came from them pushing to release a 133mhz capable chipset as soon as possible.

VIA KT333 supported 333 mhz fsb.
And they repeated the mistake.. khm.. scam again.
Guess what VIA KT400 supported: up to 333 mhz only 😁

Some companies have a certain "culture" that is pretty difficult to get rid of...

My 933/133mhz P3 cpu finally arrived and if it manages to run the 3dmark99 benchmark on loop for the whole night it will remain in this pc for the time being. It runs on 700mhz of course. This is good, because I replaced original cpu and gpu fans with 40mm "silentium" brand that are really quite silent (almost not there with case closed).

This time I got ~3800 gpu points and ~8600 cpu points. For the cpu it exceeds the doubling of frequency. On the gpu running on the same fsb extra 800 points is also not bad.

Of course not everything went great. I've always preferred pliers for removing and installing P3 socket 370 coolers (rather than screwdriver). I've done that hundreds of times before with no problem, but this time pliers slip and gouged a proper hole in my gigabyte slotket (previously borrowed, now it is mine). Thankfully having electronics as a hobby I managed to fix it, just barely. If the density of traces was double, or there were more elements there it would be much higher.

So I have few photos for your viewing pleasure :
The damage in macro photo:

Then first stage of repair with tiny (0.02mm) wires:

Then it was tested and finally UV hardening solder mask was added (I only have green) :

This was close.

you are not supposed to use such coolers on p3. the celeron 400-type coolers are enough, and those can be installed and removed by pressing on them with your finger.

My ECS board is also not stable at 133FSB, therefore I got a PIII 900E. PIII 1Ghz E is very rare. In my case the manufacturer never claimed it supported 133FSB. Run Sisoft Sandra 2002 (or other similar version) memory benchmark. You will have to use wpcredit to enable some chipset options to get decent performance out of it. 440BX at 100FSB scores about 700/750 in SSE benchmark version. With wpcredit you may get to 500, without it maybe 300.