teh_Foxx0rz wrote on 2022-04-30, 20:47:

Additionally (maybe relatedly?): This board's method of changing the FSB configuration is very different to the Socket 370 boards I have. The Socket 370 boards just need a jumper or DIP switch (or couple) switched, and it just simply runs the CPU slower accordingly. However, this board has the DIP switches, but also a jumper that needs to be on when it's 66 or 100, and needs to be off to make it anything above 124MHz, and this also doesn't simply run the CPU slower, it just simply doesn't boot if this doesn't match the CPU's listed spec.

Is this normal for Slot 1 boards?

Yes for some boards that is normal to have a jumper for FSB over-ride even if there are DIP switches. Some boards have a jumper or switch setting to force "jumper mode" versus BIOS mode mode to set the FSB.
The BIOS must recognize the CPU ID or else will not boot proper or will boot at slowest supported speed on most boards. Using a SLOT- 370 adapter is always at the users risk ;p

This board doesn't seem to be one of those with a way in the BIOS to set the FSB; it's all the DIP switches and the jumper. But I suppose it must still need to recognise the ID, and I guess that gives a limit to how late of CPUs it can support. I did update the BIOS in the hopes that that might help, but sadly not there.

Is there a chance it could work with an actual Slot 1 PIII 1000EB and not one on a slot/socket adapter?

have u checked if there is any modded bios made for that board like ones that would have came with one of those upgradeware or powerleap adapters that would let people put tualatin cpus in a board that normally wouldn't support such a thing. maybe one of those will have the code needed for the 1ghz chip. i was able to find one of those for a gigabyte GA-BX2000+ that lets me run a 1400 celeron tualatin in it. i think this board only ever officially supported upto an 850 p3.

i kinda think there is one because the guy on ebay who sells modded Tualatin cpus that can be put into a normal SLOKET adapter says your board will work with the 1400 tualatin but says it needs a new bios to do so witch he provides

teh_Foxx0rz wrote on 2022-05-01, 20:33:

This board doesn't seem to be one of those with a way in the BIOS to set the FSB; it's all the DIP switches and the jumper. But I suppose it must still need to recognise the ID, and I guess that gives a limit to how late of CPUs it can support. I did update the BIOS in the hopes that that might help, but sadly not there.

Is there a chance it could work with an actual Slot 1 PIII 1000EB and not one on a slot/socket adapter?

What board revision do you have - the later BIOS versions (with the widest CPU support) seem to be for revisions 4.x and above?

I presume the board is not stable at 133Mhz. VIA 693A was a poor man's choice back it the day, usually used only in budget Celeron builds. Get 694X or 694T.

AlexZ wrote on 2022-05-02, 22:24:

I presume the board is not stable at 133Mhz. VIA 693A was a poor man's choice back it the day, usually used only in budget Celeron builds. Get 694X or 694T.

All revisions of the board are compatible with at least the lower end 133MHz Slot 1 Coppermine CPUs, but the later revisions have broader support at the top end (if the OP can detail the board revision and which sloket is being used?). This Gigabyte doc includes revision 4.0 support for the GA-6VXE+ on BIOS F6 (F8 Beta & F9 improved this support later)

I also recall something about BIOS support is needed after ~800 MHz, maybe they have another core stepping.

PC Hoarder Patrol wrote on 2022-05-03, 07:23:

All revisions of the board are compatible with at least the lower end 133MHz Slot 1 Coppermine CPUs, but the later revisions have broader support at the top end (if the OP can detail the board revision and which sloket is being used?). This Gigabyte doc includes revision 4.0 support for the GA-6VXE+ on BIOS F6 (F8 Beta & F9 improved this support later)

That's the theory but given that the slotket could also be to blame in addition to an old board revision and VIA 693A having an extremely slow memory controller makes it inadequate for PIII build anyway. In my opinion not worth troubleshooting as performance will be a huge disappointment.

For PIII, 440BX or 694X / 694T should be used.

teh_Foxx0rz wrote on 2022-04-30, 20:47:

Hello all, […]
Show full quote

Hello all,

I've been building a Slot 1 PC that I've hoped could be versatile enough to alternate CPUs between a 233MHz Klamath Pentium II (downclocked to 133MHz), all the way up to a 1GHz Pentium III with a 133MHz FSB.

The board I've picked for now out of what came my way is a Gigabyte GA-6VXE+, which has a VIA VT82C693A (Apollo Pro 133), which of course does support 133MHz FSB (without overclocking the expansion slots).
However, as the title says, when I put a PIII 1000 EB in it (using a slotket adapter, though I don't think it makes a difference, it works fine on slower speeds), I simply get a loop of Windows 98SE saying it's "restored a good registry", and never getting past this point. I've also just tried a PIII 933MHz CPU, just in case the 1000EB was just slightly over some threshold (and they're cheap so it's no biggie), but sadly, as somewhat expected, still the same issue.

The manual does say it supports processors between 233-800MHz, but admittedly, I perhaps naively thought this was merely advisory, perhaps put as this simply because the fastest PIIIs at the time of its release were 800MHz rather than a hard limit in the scope of its design. Certainly, a 1GHz Slot 1 system seems to be a popular project, even on 100MHz FSB boards, so I don't feel like this was an implausible hope.

So, my question: am I simply out of luck, and need to look out for a different motherboard? Or is there a possible way I could get around this by any chance?

Additionally (maybe relatedly?): This board's method of changing the FSB configuration is very different to the Socket 370 boards I have. The Socket 370 boards just need a jumper or DIP switch (or couple) switched, and it just simply runs the CPU slower accordingly. However, this board has the DIP switches, but also a jumper that needs to be on when it's 66 or 100, and needs to be off to make it anything above 124MHz, and this also doesn't simply run the CPU slower, it just simply doesn't boot if this doesn't match the CPU's listed spec.

Is this normal for Slot 1 boards?

I realise the extra board the CPU is mounted on might make some electrical features of the CPU more distanced from the board, such as this. And I know slotket adapters tend to also have a FSB speed jumper on them (and this *does* downclock the CPU accordingly, if the motherboard is set the same so that it actually boots)

Some other, merely tangential info about my system, for flavour/context:

I was primarily building this particular PC to make the most of my Voodoo 3 3500 (I heard they scale well all the way up to a 1000MHz CPU); it was originally in a Socket 370 board with an Apollo Pro 133A chipset, which would be easy to put a 1000MHz PIII into, and downclock the FSB when needed for better compatibility with earlier '90s DOS games.

However, I was also wanting a PC for a Voodoo 1, but the case I have aside for that is only a Micro-ATX. Micro-ATX Slot 1 boards with AGP, two PCI slots (or three if one is a combo) and also an ISA slot seem to be rather hard to come by. I have a S370 board that fits that description that I was hoping to put a 533MHz PIII into so I could go up to 533MHz for some things, then all the way down to 266MHz for V1 stuff, but those seem to be scarce or expensive, so I looked to Slot 1 boards. But Slot 1 boards which fit all of that description seem to be even more hard to come by than PIII 533's. (I'm not sure I'm ready to dive into the AT form-factor yet either, haha.)

But then this GA-6VXE+ came up and looked ideal, supporting PIIIs as well as PIIs, and 133Mhz FSB, plus two ISA slots and other neat things (like a SB-Link header which I never imagine using but would be neat to have to play with if such a card came up), so I figured I would try just combining the two systems, and leveraging the Slot 1 platform's flexibility. I've got the Voodoo 1 working just fine alongside the Voodoo 3 3500 (and dipped into a bit of Dreams to Reality! Weird but intriguing game) and the Soundblaster 16 working just fine alongside an OPTi 82C929A chip ISA sound card (and Audigy 2 ZS without the DOS drivers installed) but sadly, maybe this is its compromise.

Of course, I'm fully aware that things are rarely completely ideal and straightforward with these old PCs, but we can always hope, you know?

Thanks for all the comments and help.

Personally I believe a Coppermine 800EB would be sufficient for your Voodoo 3 3500. I've run my Voodoo 3 3000 with a Celeron 800 even.
It's possible that your board (I have not looked into it, nor do I really know this particular board or this particular chipset well), seeing it officially going up to 800MHz, could indeed be some kind of bug, or the power delivery of the board is insufficient for your CPU.
Your CPU being on a slotket could also complicate things.

If you happen to have a 800MHz Coppermine (preferably one that works with 133MHz FSB and not the 100MHz FSB variant), it will be basically a drop-in replacement.
The 1000MHz Coppermine uses around 30W while the 800MHz is more a 20W CPU. It could actually be something with the power delivery if the CPU you're using requires 50% more power compared to what this board was build for.

If you end up deciding to get a replacement board, you could get one without ISA slots (i815 would be an easy pick for a 1GHz Coppermine build, it basically just works) and make your old board the Voodoo 1 board along with a (much) slower CPU.
Btw, there are even some 400MHz P2 CPUs with Deschutes core which can be underclocked as not all of these have not been fully locked. Iirc stepping was SL2S7. Thread is here if you want to do some reading on this subject 😋
But in essence you could (if you find one that isn't fully locked) set your board to run the 400MHz Deschutes at 66MHz FSB for a 366MHz CPU, but also underclock it by lowering the CPU multiplier for your Voodoo 1, if for whatever reason you need to underclock it further.

Would perhaps be nice if your 693A chipsetted board could be used stably with the Voodoo 1 since it does have some nice features, while your faster s370 system could field your Voodoo 3 3500, perhaps even at 1GHz (which imo is definitely more than enough for the V3).
My favorite chipset for a 1GHz Coppermine is by far the i815 chipset. It's virtually idiot-proof with the main disadvantage being the lack of ISA which imo a Windows system can live without, especially if you already have a dedicated rig for ISA sound.
The advantage of the 694 chipsets is especially that these support >512MB total system RAM, in case you ever need it.

Tetrium wrote on 2022-05-04, 12:13:

But in essence you could (if you find one that isn't fully locked) set your board to run the 400MHz Deschutes at 66MHz FSB for a 366MHz CPU, but also underclock it by lowering the CPU multiplier for your Voodoo 1, if for whatever reason you need to underclock it further.

By 366, you mean 266, right? 😉

H3nrik V! wrote on 2022-05-04, 12:43:

Tetrium wrote on 2022-05-04, 12:13:

But in essence you could (if you find one that isn't fully locked) set your board to run the 400MHz Deschutes at 66MHz FSB for a 366MHz CPU, but also underclock it by lowering the CPU multiplier for your Voodoo 1, if for whatever reason you need to underclock it further.

By 366, you mean 266, right? 😉

No, 366MHz
As seen here:

This CPU wouldn't work with a higher multiplier, or at least not as a 66MHz FSB 400MHz CPU. 5.5*multi was as high as it would go.

Tetrium wrote on 2022-05-04, 12:58:

No, 366MHz As seen here: DSC00272.jpg This CPU wouldn't work with a higher multiplier, or at least not as a 66MHz FSB 400MHz CPU […]
Show full quote

H3nrik V! wrote on 2022-05-04, 12:43:

Tetrium wrote on 2022-05-04, 12:13:

But in essence you could (if you find one that isn't fully locked) set your board to run the 400MHz Deschutes at 66MHz FSB for a 366MHz CPU, but also underclock it by lowering the CPU multiplier for your Voodoo 1, if for whatever reason you need to underclock it further.

By 366, you mean 266, right? 😉

No, 366MHz
As seen here:
DSC00272.jpg
This CPU wouldn't work with a higher multiplier, or at least not as a 66MHz FSB 400MHz CPU. 5.5*multi was as high as it would go.

Ah, sorry, missed the part about isn't fully locked 🤣 now it makes sense 😀

There are bios modder forums that can program updated microcodes into bios for newer cpu support (REVISED WITHOUT CUTTING OFF LEGACY). As for 98se registry corruption I hope your disk hasn't a weak cluster/sector on it.

H3nrik V! wrote on 2022-05-04, 15:46:

Tetrium wrote on 2022-05-04, 12:58:

No, 366MHz As seen here: DSC00272.jpg This CPU wouldn't work with a higher multiplier, or at least not as a 66MHz FSB 400MHz CPU […]
Show full quote

H3nrik V! wrote on 2022-05-04, 12:43:

By 366, you mean 266, right? 😉

No, 366MHz
As seen here:
DSC00272.jpg
This CPU wouldn't work with a higher multiplier, or at least not as a 66MHz FSB 400MHz CPU. 5.5*multi was as high as it would go.

Ah, sorry, missed the part about isn't fully locked 🤣 now it makes sense 😀

These aren't officially supported multipliers and it's not exactly perfectly consistent which multipliers will work with what CPU. An identical CPU may be fully locked or have unlocked multipliers like this 400MHz Deschutes of mine, but fewer of them.
There also may be ones with higher multipliers which the CPU itself cannot hope to even handle.

teh_Foxx0rz wrote on 2022-04-30, 20:47:

Additionally (maybe relatedly?): This board's method of changing the FSB configuration is very different to the Socket 370 boards I have. The Socket 370 boards just need a jumper or DIP switch (or couple) switched, and it just simply runs the CPU slower accordingly. However, this board has the DIP switches, but also a jumper that needs to be on when it's 66 or 100, and needs to be off to make it anything above 124MHz, and this also doesn't simply run the CPU slower, it just simply doesn't boot if this doesn't match the CPU's listed spec.

Normally there's a jumper somewhere that controls the PCI/etc clock dividers, but often it's purpose isn't clearly marked, so that extra jumper could have something to do with that. That doesn't explain why it would be the same for both 66 and 100FSB though. Maybe it's half of a 2-bit setting and the other half is buried in the "FSB" dip switches.

Are you heavily loading the RAM, and if so have you tried cutting it back? I recall these chipsets were flaky at PC133 when loaded with "too many" modules.
I'm looking in old notes for an old motherboard I had of this generation and apparently changing the BIOS setting "Delay DRAM Read Latch" to "No Delay" could improve stability.

Alright everyone -

Thank you for all your input. There's a lot interesting things to consider!

I found myself thinking about the mention that the BIOS would need to recognise the CPU ID. And since this is a Slot 1 board, I figured they might only consider Slot 1 CPUs. So, I decided I'd try the obvious, and fork out for an actual Slot 1 PIII 1GHz (which fortunately didn't break the bank even if it wasn't pennies heh, although I did have to replace a monstrously ugly Dell cooler on it!).

And, well...
That worked!! Thanks to those of you who mentioned about that!
So yeah, either it was the CPU ID not being recognised, or just simply a problem with the slotket adapter. And that problem is solved.
Thanks very much everyone!

But of course, there are still some other things to follow up on.

PC Hoarder Patrol wrote on 2022-05-02, 01:25:

What board revision do you have - the later BIOS versions (with the widest CPU support) seem to be for revisions 4.x and above?

It seems my revision is indeed 4.0! So it looks like I was lucky there!

On modded BIOSes - those would have been the next place I'd have looked if this didn't work. Might still be neat to look into though still, even if I don't use one now, heh!

And fortunately, now that it's working, it doesn't seem to be an issue with my hard drive. I didn't think it was anyway, since it distinctly worked at slower speeds even at 133MHz FSB (a 533MHz PIII), but it was always a potential thing still, yeah.

To those recommending Intel chipsets - I really wanted 133MHz FSB support and ISA slots! So a VIA chipset is really the only practical option. I might have picked an Apollo Pro 133A, and did have a board with that in this PC previously, but yeah, that was a socket 370 and I wanted the versatility of a Slot 1. And I didn't need AGP4x with a Voodoo 3. But it's working now, so that's all good heh.

Tetrium wrote on 2022-05-04, 12:13:

Would perhaps be nice if your 693A chipsetted board could be used stably with the Voodoo 1 since it does have some nice features, while your faster s370 system could field your Voodoo 3 3500, perhaps even at 1GHz (which imo is definitely more than enough for the V3).

My S370 systems are unfortunately not really options for my Voodoo 3, since one is actually a 1.4GHz Tualatin system I've put a Geforce FX 5700 in for (basic) DX9 support, and the other, the Micro-ATX system, has a case with absolutely inadequate ventilation for a Voodoo 3, haha. It has barely any ventilation and no place to mount case fans!

shamino wrote on 2022-05-04, 22:52:

Normally there's a jumper somewhere that controls the PCI/etc clock dividers, but often it's purpose isn't clearly marked, so th […]
Show full quote

teh_Foxx0rz wrote on 2022-04-30, 20:47:

Additionally (maybe relatedly?): This board's method of changing the FSB configuration is very different to the Socket 370 boards I have. The Socket 370 boards just need a jumper or DIP switch (or couple) switched, and it just simply runs the CPU slower accordingly. However, this board has the DIP switches, but also a jumper that needs to be on when it's 66 or 100, and needs to be off to make it anything above 124MHz, and this also doesn't simply run the CPU slower, it just simply doesn't boot if this doesn't match the CPU's listed spec.

Normally there's a jumper somewhere that controls the PCI/etc clock dividers, but often it's purpose isn't clearly marked, so that extra jumper could have something to do with that. That doesn't explain why it would be the same for both 66 and 100FSB though. Maybe it's half of a 2-bit setting and the other half is buried in the "FSB" dip switches.

Are you heavily loading the RAM, and if so have you tried cutting it back? I recall these chipsets were flaky at PC133 when loaded with "too many" modules.
I'm looking in old notes for an old motherboard I had of this generation and apparently changing the BIOS setting "Delay DRAM Read Latch" to "No Delay" could improve stability.

So, there are three jumper headers associated with RAM/PCI speed, which are the ones with the blue jumpers in this picture.
https://i.imgur.com/RdND3du.jpg

The two on the left; the top one (JP10) seems to be turbo, but I'm not sure what the bottom one (JP12) seems to be exactly, though I think it's PCI speed?
Then there's the blue jumper on the right, near the DIP switches (JP15) - this mentions PCI in its description, but it seems intrinsically linked with the DIP switch settings and FSB, to the extent where it's in the table to the lower left of the chipset heatsink; the first column of it, next to the numbers. And as you can see, it's the only thing which changes between 100MHz FSB and 133MHz FSB. Since 133Mhz is listed as officially supported and not listed as a risky FSB speed in the manual, I don't think it's anything to do with overclocking the PCI slots or such. Perhaps it does just set a multiplier. Though I'm not sure what any of the possibilities have to do with the description there of 33Mhz vs 44Mhz.

And yeah, just for the record, I'm only using two RAM modules of 256MB each, which is pretty standard and I wouldn't expect to cause any power problems.

In addition to being curious to know whatever's going on with this, I also wonder, is this kind of thing, and not being able to change the speed of the CPU just by changing the FSB configuration like you can with S370 boards I've used, normal for Slot 1 boards that support 133MHz FSB?
At least I can get around this very easily on a Slot 1 board just by buying a few more CPUs for the configurations I might want to switch them out easily, even if this isn't quite as economical and elegant as I was hoping!

The problem of VIA 693A is extremely poor memory performance. You need to fiddle with pcredit to get decent performance out of it. It is not possible to adjust the necessary settings in DOS. Games will suffer a lot. I had this chipset with PPGA Celeron back in the day. If you really want VIA then get at least 694X as it has a better memory controller. 440BX at 100Mhz FSB performs approximately as 133Mhz VIA694X memory wise. Run memory benchmark on Sisoft Sandra 2003 to see the difference.

teh_Foxx0rz wrote on 2022-05-05, 21:15:

[..]And since this is a Slot 1 board, I figured they might only consider Slot 1 CPUs. So, I decided I'd try the obvious, and for […]
Show full quote

[..]And since this is a Slot 1 board, I figured they might only consider Slot 1 CPUs. So, I decided I'd try the obvious, and fork out for an actual Slot 1 PIII 1GHz (which fortunately didn't break the bank even if it wasn't pennies heh, although I did have to replace a monstrously ugly Dell cooler on it!).

And, well...
That worked!! Thanks to those of you who mentioned about that!
So yeah, either it was the CPU ID not being recognised, or just simply a problem with the slotket adapter. And that problem is solved.

It's 100% because of the slotket. There's no magic microcode for socket 370 CPUs that would make them incompatible with Slot 1 boards. 😀
You probably had one of those very cheapo slotkets (maybe one that was never meant/tested to work with Coppermines in the first place).

bloodem wrote on 2022-05-06, 07:17:

teh_Foxx0rz wrote on 2022-05-05, 21:15:

[..]And since this is a Slot 1 board, I figured they might only consider Slot 1 CPUs. So, I decided I'd try the obvious, and for […]
Show full quote

[..]And since this is a Slot 1 board, I figured they might only consider Slot 1 CPUs. So, I decided I'd try the obvious, and fork out for an actual Slot 1 PIII 1GHz (which fortunately didn't break the bank even if it wasn't pennies heh, although I did have to replace a monstrously ugly Dell cooler on it!).

And, well...
That worked!! Thanks to those of you who mentioned about that!
So yeah, either it was the CPU ID not being recognised, or just simply a problem with the slotket adapter. And that problem is solved.

It's 100% because of the slotket. There's no magic microcode for socket 370 CPUs that would make them incompatible with Slot 1 boards. 😀
You probably had one of those very cheapo slotkets (maybe one that was never meant/tested to work with Coppermines in the first place).

There are even CPUID differences between different steppings of 1GHz P!!!s, might be pure luck that the slot1 part worked and the FCPGA didn't?

@OP: What was the sSpecs of the CPUs in question?