dionb wrote on 2022-12-03, 15:16:

According to spec, i430NX does NOT support 128MB SIMMs - it supports max 256MB, but that's using a lot more than two SIMMs.

Hmm, something odd in the manual ( https://theretroweb.com/motherboard/manual/cm … 5e926995047.pdf ). Page 59 agrees with you and lists 32MB as the maximum per stick, but page 52 gives a 384MB maximum by using 6 @ 64MB. Also, RetroWeb's entry for the 82434NX ( https://theretroweb.com/chipsets/1357 ) gives 512MB as the maximum, using 128MB FPM. It also says the cache can be either async or pipelined burst. That seems to agree with the datasheet for the 82434NX ( https://datasheetspdf.com/pdf-file/45336/Inte … ation/82434NX/1 ). Datasheet actually says SIMMs of 16M x 36bit, which only makes 64MB, but I think that's for a single bank, so 2 banks (think that sometimes gets called double sided) would be 128MB. Looks like maybe the LX is only 32MB?

I might be misinterpreting something though, so please point out if I've gone wrong.

Sphere478 wrote on 2022-12-03, 18:12:

-Install the second socket.

Yeah, still think it'd be better to get everything else checked out first. But if you get to this bit then one thing that will probably need to happen is swapping the 82378 for an 82379 for multi-processor support. Plus other 430NX MP boards that maybe have BIOSes to experiment with: https://theretroweb.com/motherboards/?showIma … &chipsetId=1357

dionb wrote on 2022-12-03, 21:30:

I have to admire your chutzpah - you treat known limitations as challenges, which could be described as foolish or worse, but you just chip away at it and discover stuff like the K6-2+ cache unlock by ignoring other people telling you what's not possible. That one alone earns you a place in the annals 😀

Here again I'd say you're bordering on insane with what you want to do, but if anyone can do it, you can. That said, I'll eat my hat if you get two K6 CPUs w0rking on this board. But afterwards I will get myself a similar board and reproduce it 😉

I honestly don’t think it will work. But there are at least two people claiming to have seen it in the wild. Chances are that they are trolling. But trying it on a socket 5 has been something I have been meaning to try to confirm.

Some of the quad socket 5 servers seem more likely to be able to pull this off but even that is unconfirmed

The mere fact that this is so hard to test with the requirement of a voltage interposer and bf mods makes me pretty sure that not many people have actually tried it and seeds just enough doubt that I can’t be 100% sure.

Also, the smp pins are showing some kinda logic connection on a ohm meter so I have some hope that maybe amd had some kind of plan for smp k6 that never was realized perhaps because no smp ss7 chipset was ever made, and no motherboards as a result. Leaving the very rare s5 with interposer the only avenue to do it.

Theory is that k5 chips for s5 weren’t able to do it, therefore no one ended up trying it on the dual plane ss7 chips.

But,

I’m calling it now though. that it probably won’t work, but damnit if I’m not gonna try it with my extensive collection of k6 chips. 😀 for science!

I’m basically 90% sure that it isn’t going to work.

snufkin wrote on 2022-12-03, 21:52:

Hmm, something odd in the manual ( https://theretroweb.com/motherboard/manual/cm … 5e926995047.pdf ). Page 59 agrees with you a […]
Show full quote

dionb wrote on 2022-12-03, 15:16:

According to spec, i430NX does NOT support 128MB SIMMs - it supports max 256MB, but that's using a lot more than two SIMMs.

Hmm, something odd in the manual ( https://theretroweb.com/motherboard/manual/cm … 5e926995047.pdf ). Page 59 agrees with you and lists 32MB as the maximum per stick, but page 52 gives a 384MB maximum by using 6 @ 64MB. Also, RetroWeb's entry for the 82434NX ( https://theretroweb.com/chipsets/1357 ) gives 512MB as the maximum, using 128MB FPM. It also says the cache can be either async or pipelined burst. That seems to agree with the datasheet for the 82434NX ( https://datasheetspdf.com/pdf-file/45336/Inte … ation/82434NX/1 ). Datasheet actually says SIMMs of 16M x 36bit, which only makes 64MB, but I think that's for a single bank, so 2 banks (think that sometimes gets called double sided) would be 128MB. Looks like maybe the LX is only 32MB?

I might be misinterpreting something though, so please point out if I've gone wrong.

Sphere478 wrote on 2022-12-03, 18:12:

-Install the second socket.

Yeah, still think it'd be better to get everything else checked out first. But if you get to this bit then one thing that will probably need to happen is swapping the 82378 for an 82379 for multi-processor support. Plus other 430NX MP boards that maybe have BIOSes to experiment with: https://theretroweb.com/motherboards/?showIma … &chipsetId=1357

not in front of the board atm.

Is that the chip to the left that looks like a apic?

Shoot, so it’s not ready for mps yet. Okay, I’ll get one in that case then.

Is there a older or non intel apic that I can use, that will maximize chances of the dual amd longshot.

snufkin wrote on 2022-12-03, 21:52:

[...]

That seems to agree with the datasheet for the 82434NX ( https://datasheetspdf.com/pdf-file/45336/Inte … ation/82434NX/1 ). Datasheet actually says SIMMs of 16M x 36bit, which only makes 64MB, but I think that's for a single bank, so 2 banks (think that sometimes gets called double sided) would be 128MB. Looks like maybe the LX is only 32MB?

Interesting, that does look like 64Mb chips are supported by i430NX where they aren't in i430LX. If the 64Mb chips do work on NX I'd be very interested to see what they do on LX, as the chipsets are otherwise exceedingly similar. This sounds like i440LX officially only supporting 512MB SDRAM (yet 1GB of EDO) due to only supporting 64Mb chips, not 128Mb. Its successor i440BX supports 1GB of both. Guess what? Works on i440LX too. And i44oEX and i430TX too, albeit with 2 instead of 4 banks so 512MB max.

Sphere478 wrote on 2022-12-03, 22:19:

[...]

I honestly don’t think it will work. But there are at least two people claiming to have seen it in the wild. Chances are that they are trolling. But trying it on a socket 5 has been something I have been meaning to try to confirm.

Some of the quad socket 5 servers seem more likely to be able to pull this off but even that is unconfirmed

I strongly suspect you need a different kind of motherboard here. The difference is that Intel's built-in SMP supports 2-way max at this point. So a quad So5 system must use some 3rd party glue logic - which works around the chipsets supporting Intel SMP not supporting AMD SMP. i430NX is the first gen Intel SMP implementation only. Unless it's not...

The mere fact that this is so hard to test with the requirement of a voltage interposer and bf mods makes me pretty sure that not many people have actually tried it and seeds just enough doubt that I can’t be 100% sure.

Also, the smp pins are showing some kinda logic connection on a ohm meter so I have some hope that maybe amd had some kind of plan for smp k6 that never was realized perhaps because no smp ss7 chipset was ever made, and no motherboards as a result. Leaving the very rare s5 with interposer the only avenue to do it.

There's no such thing technically as "ss7", it's just a marketing term for So7 with higher FSB, AGP - oh and lower volrage and higher multiplier too, but apart from that BF2 pin, it's just So7. So a regular So7 SMP chipset with native dual-plane support (i.e. i430HX) would be a more logical place to start than So5 - except for the whole 'will only work with two separate buses' thing. Not sure if any officially go down to 2.2V, but having two separate planes to start with makes life a lot easier.

Theory is that k5 chips for s5 weren’t able to do it, therefore no one ended up trying it on the dual plane ss7 chips. […]
Show full quote

Theory is that k5 chips for s5 weren’t able to do it, therefore no one ended up trying it on the dual plane ss7 chips.

But,

I’m calling it now though. that it probably won’t work, but damnit if I’m not gonna try it with my extensive collection of k6 chips. 😀 for science!

I’m basically 90% sure that it isn’t going to work.

Your advantage: an equally extensive collection of split voltage adapters. They give you more options than most people who tried this before. Still, I think you need an obscure quad-CPU system to get two to work.

Yes the unicorn quad systems seem to have the best chance if it is going to happen.

I have tried basically every chip mfg except slow speed chip offerings on my tyan s1564d

430hx it seems is ruled out.

It’s such a long shot that we really that this will likely never amount to more than dreaming.

Anyway,

Does this board have onboard audio? Looks like I have empty pads for it

the bus has to be able to operate in something called tri state a perfect example of a chip that can do this but can’t smp is tillamook. At least this is what intel has to do. Perhaps intel broke amd’s version of this the hx on purpose? If even it is there. Probably not. I doubt it is even there. Again, I don’t think this is possible but I want to continue to check.

And a curious thing happens when you try to smp a tillamook. It posts. This is not the case when trying any combo of other brand cpu. Suggesting that either the bios or intel broke it on purpose or it simply isn’t there and and bus collisions are occurring.

Bet is that it simply isn’t there and bus collisions are occurring

I seem to have beaten the tillamook to death to the point that I am convinced we will never see a ppga tillamook in smp. The film chip ones though… prob not but further investigation needed.

Re: Interposer for mobile Pentium MMX Tillamook

Tillamook SMP and multiplier unlocking investigation thread.

I was however able to prove that the intel mmx overdrive can in fact do smp.

Re: Socket 5 / 7 Pentium Overdrives. Higher Multipliers? SMP? Let’s find out.. So there is a partial win 😀

dionb wrote on 2022-12-03, 23:09:

snufkin wrote on 2022-12-03, 21:52:

[...]

That seems to agree with the datasheet for the 82434NX ( https://datasheetspdf.com/pdf-file/45336/Inte … ation/82434NX/1 ). Datasheet actually says SIMMs of 16M x 36bit, which only makes 64MB, but I think that's for a single bank, so 2 banks (think that sometimes gets called double sided) would be 128MB. Looks like maybe the LX is only 32MB?

Interesting, that does look like 64Mb chips are supported by i430NX where they aren't in i430LX. If the 64Mb chips do work on NX I'd be very interested to see what they do on LX, as the chipsets are otherwise exceedingly similar. This sounds like i440LX officially only supporting 512MB SDRAM (yet 1GB of EDO) due to only supporting 64Mb chips, not 128Mb. Its successor i440BX supports 1GB of both. Guess what? Works on i440LX too. And i44oEX and i430TX too, albeit with 2 instead of 4 banks so 512MB max.

Datasheet says the LX has one less RAM address line, 11 (MA10:0] rather than 12 (MA11:0), so ends up as effectively 22 bits (4M addresses * 4 bytes = 16MB, at 2 ranks that's 32MB) rather than the NX 24 bits (16M addresses * 4 bytes = 64MB, 2 ranks gives 128MB).

Looking at the datasheet pinout then pin 109 is MA11 on the NX and is NC on the LX. Maybe it's the same silicon underneath but Intel didn't bond out to the package pin for the LX?

Sphere478 wrote on 2022-12-03, 22:36:

Is that the chip to the left that looks like a apic?

Shoot, so it’s not ready for mps yet. Okay, I’ll get one in that case then.

Is there a older or non intel apic that I can use, that will maximize chances of the dual amd longshot.

Yeah, between the /COAST and the PCI slots.

Slightly pointless aside, but I'm wondering about naming conventions. I've seen several ways schematics mark signals as active-low, or inverted, so like a NOT gate. So for 'not-a-COAST' should that be '/COAST', 'nCOAST', '!COAST', '#COAST' or '_COAST' (are there any others I've missed?). I usually avoid exclamation marks, but could be persuaded here due the consequences of getting it wrong.

My guess, and I haven't looked in to this much, is that you're stuck looking for the 82379. There's a datasheet on the two parts here: https://pdf1.alldatasheet.com/datasheet-pdf/v … EL/82379AB.html so you can see everything they handle, including PCI-ISA bridge, DMA, interrupts and timers. Looks like the '79 drops some some functions from DMA and ISA to free up 3 pins for the APIC. Handily the differences between the '78 and '79 are highlighted in grey. That datasheet might help for setting some the ROPT resistors on the motherboard, e.g. flicking through I noticed that IRQ8# (pin 172) needs to have an 8.2k pull-up when using the '79 so there's likely/hopefully a connection between pin 172 and one of the ROPT pads.

I consider all COAST and not-a-COAST sticks waste not, want not.

Lol.

I think maybe I need to track down a high res pic of a fully populated board.

Anyone run across one?

Sphere478 wrote on 2022-12-03, 23:32:

Does this board have onboard audio? Looks like I have empty pads for it

From a couple of blurry photos, looks like a 52 pin ESS chip. There was the ES1488, but I can't find much about it at all. Looks like any components starting with 'A' are part of the audio system, ending up at header AP1 which I think contains a Mic/Line In, Speaker Out and a PC Speaker output (I guess that's also on AP2). Maybe AU1 is an 8 pin amp?

Sonicx's suggestion of the Am79C974 LAN/SCSI chip ( https://www.amd.com/system/files/TechDocs/18681.pdf ) looks like a good shout, it's at least got the right number of pins. You could check the board connections against the datasheet page 12 pinout. Looks like it needs either a 20 or 40MHz oscillator. I'm not at all clear how the clocking for it works, the datasheet seems to say that the LAN part needs a separate crystal from the SCSI clock, but there's only the one oscillator socket on the board. For the LAN stuff the board also need the filter/magnetics FD1, maybe something like this https://www.yuandean.com/en/product/LAN-Filte … 14F-Series.html .

Might be some useful files here: https://www.elhvb.com/mobokive/archive/Zeos/M … tium/index.html , although it looks like your board is already on version 8 of the cmax (which I think is your board : 010-0050 ) versions. Also a useful file here: https://www.elhvb.com/mobokive/Archive/Zeos/M … os_files/id.txt with entries for the Coral Max and Coral Max 3, including part numbers for the original cache sticks (IDT7MPV6179/80/81/82).

[managed to double post whilst trying to edit to add:]

Sphere478 wrote on 2022-12-03, 22:36:

Is that the chip to the left that looks like a apic?

Realised I wasn't clear before, I'm talking about the big Intel chip. I think the smaller chip next to it is the PCI-IDE interface, maybe an RZ1000 from PCTech, which Wikipedia tells me is connected with Zeos.

re host

Okay, this is going very well 😀

512mb no problem. 4x 128mb

What’s next? 🤔

Pentium 200 at 100 is able to be kept cool by switching fingers every 5 seconds 🤣

Debating what I am going to do for the interposer. Leaning toward some sort of perminent version of my interposer to permanently convert it to socket 7 but maybe not.

Might be time to make a version of the interposer that has a spot for a murata or possibly a 9-16 bit card for isa that pulls from 5v

So one thing someone told me once is that this board wouldn’t work on a non intel cpu.

Here is an attempt at a k6-3+ 570.

The bf pins aren’t set up on this interposer so that may be why. 🤔 gonna throw a k5 in there and see what happens

Result:

Tried a tillamook. No dice.

Finally succeeded with a pentium mmx 233 on the interposer. (Still set at 2.0v 🤣)

Bios says ‘pentium 120’

I think I may be dealing with some sort of Prejudice bios 🤣. Or a hopelessly out of date one, but why would a k5 133 not work?

It was making post codes… but no bios screen. So I’m thinking bf pins aren’t problem.

What alternative bios should I try? Mr bios?

Tried a winchip.

Yeah, this bios is definitely doing this on purpose 🤣. Gonna have to hack it or ditch it.

I need my k6-3+ or I’ll start twitching and foaming at the mouth.

Edit:
Finally made it to cpu-z (I’m keyboard only)

Crap, where’s my L2..

I played with some bios settings:

Not the external cache setting though oddly,

And….

I see the name "Pantera" and my mind immediately went two places:

https://www.youtube.com/watch?v=i97OkCXwotE

🤣

well, so you think the second socket isn't gonna work without swapping that chip? where can I get it?

Yeah, the datasheet for the 82378/82379 lists one of the extra features of the '79 as having an APIC with "Support for Multi-Processor Systems". RetroWeb only shows 6 boards with the '79: https://theretroweb.com/motherboards/?showIma … &chipsetId=1357 and I think it'd take a stroke of luck to come across a broken one of those that you could salvage the '79 from. Usual IC source websites that always feel, perhaps unfairly, a bit dodgy don't have it either, except SeekIC who list it as price negotiable.

Nice work proving the memory capacity. I think entry level RAM around the time was 8 or 16MB. What does it do if you put those sticks you have in slots 5&6? For other possible BIOSes, might be worth checking https://theretroweb.com/bios/?chipsetId=273 . Other hardware on those boards might be different, so might lose IDE/Floppy/Serial.

On the POST codes, is the display E7 E6, with the missing digits in the photo just due to camera speed? A generic list of Phoenix codes says that could be RAM check. Might be worth trying going back to small amounts of RAM, and trying to work out which jumpers on the board control the cache options, so you can disable that for the moment.

Also, just for laughs, you could see what it does with a BIOS for the '79: https://theretroweb.com/bios/?chipsetId=1357 The ASUS has an update from '98.
Probably fail to boot when it tries to check interrupts.