computergeek92 wrote:

K6-2 mainboards also used 83, 95 or 97 fsb bus speeds. (As well as the standard 66 and 100) Did the odd bus speeds cause hardware problems or incompatibilities due the mobo's clocks being on a synchronized bus?

I had Asus T2P4 motherboard. It officially supported FSB speeds up to 66 MHz, but with jumpers there were also undocumented 68, 75 and 83 MHz settings. I think I never got it to POST at 83 MHz, and it was quite unstable at 75 MHz. It was either the onboard IDE controller or the IDE drives, but helped a bit to limit PIO/UDMA modes to be a bit slower.

I have an Intel 430TX-based board which runs well with an 83 MHz FSB. To get the system stable, I had to disable the motherboard's L2 cache by pulling out the TAG RAM chip. This was not a problem because I am using an AMD K6-III+ with built-in L2 cache. I also found that with 83 MHz, EDO RAM (256MB) would run fine with fast RAM timings, however if using SDRAM, I had to reduce it to 128 MB. I did not notice any benchmark improvement with the SDRAM, so I kept 256 MB EDO.

I also have a Intel TX board, DFi P5BTX/L Rev C, it works without a problem with Cyrix MII CPU's at 75 & 83 MHz FSB. The ATX version of the DFI(586ITXD) with the same chipset however is unable to cope with 83MHz and occasionally flaky at 75 MHz. Both have the latest (1998) BIOS.

I also have a PCChips M590 with SiS 5591 chipset, that is fine with all speeds upto 95 MHz.

BSA Starfire wrote:

I also have a Intel TX board, DFi P5BTX/L Rev C, it works without a problem with Cyrix MII CPU's at 75 & 83 MHz FSB. The ATX version of the DFI(586ITXD) with the same chipset however is unable to cope with 83MHz and occasionally flaky at 75 MHz. Both have the latest (1998) BIOS.

I, too, originally noticed this flaky nature, which was was worse at 83 Mhz, occasional at 75 MHz, and non-existent at 66 MHz. I found switching to 72-pin RAM over SDRAM solved the problem entirely. Did you try 72-pin EDO RAM?

That's interesting, I must admit I never have, I just assumed that SDRAM would cope better than 72 pin. It's odd that the AT version of the board is fine with 75/83 MHz FSB and the ATX is so picky, currently have a intel Pentium 200 MMX in the ATX board & it's as good as gold, was fine with a Cyrix 166MX too, in fact now I think about it, that 586ITXD wouldn't play with a AMD K5 166 reliably either(66MHz x 2.5) maybe it's just really fussy about what CPU's it likes?

As long as the chipset and motherboard are designed properly, it shouldn't matter.
The PCI buses should run at 33 MHz regardless of the FSB speed. So if the FSB speed is not a multiple of 33 MHz, then the PCI bus needs to be run asynchronously from the FSB (some simple buffer to decouple them).
Even Intel does this, eg with the 50/60 MHz FSB on early Pentiums.

As far as I recall all those old Socket 7 chipsets just overclocked or underclocked the buses. And of course it often caused problems when PCI was 38 MHz or faster. Like with VLB too for that matter.

Decoupling didn't happen until maybe nForce?

swaaye wrote:

As far as I recall all those old Socket 7 chipsets just overclocked or underclocked the buses.

Pretty sure they don't. PCI has to run at 33 MHz period (not any higher, but not any lower either). Why would Intel not implement their own standard properly? Or why would they even design PCI with a fixed frequency when they want to implement it with various frequencies?

Scali wrote:

swaaye wrote:

As far as I recall all those old Socket 7 chipsets just overclocked or underclocked the buses.

Pretty sure they don't. PCI has to run at 33 MHz period. Why would Intel not implement their own standard properly?

Intel ran their CPUs at 60 or 66 MHz FSB. The chipsets were designed for that. It was Cyrix and AMD that had the 75/83/95 MHz FSBs specified in some cases.

swaaye wrote:

Intel ran their CPUs at 60 or 66. None of this 75 or 83 MHz stuff. The chipsets had the proper dividers for 60/66.

There is a Pentium 75 that runs at 1.5x 50 MHz.
Most socket 7 boards will have 50/60/66 FSB options.
See here: https://en.wikipedia.org/wiki/List_of_Intel_chipsets

Scali wrote:

swaaye wrote:

As far as I recall all those old Socket 7 chipsets just overclocked or underclocked the buses.

Pretty sure they don't. PCI has to run at 33 MHz period. Why would Intel not implement their own standard properly?

Because all of the Intel CPU's on skt7 ran at 66 MHz maximum FSB & they didn't want you to buy AMD or Cyrix CPU's that ran at 75MHz or 83 Mhz or faster. It was upto motherboard makers to "fix" the intel chipset's to run over 66 MHz with varying results board by board it seems.

BSA Starfire wrote:

Because all of the Intel CPU's on skt7 ran at 66 MHz FSB

Nope, see above, they supported 50/60/66 MHz FSB.

Yes. Ok. But Cyrix and AMD wanted you to try to run these chipsets at 75, 83 or 95MHz in some cases and the chipsets were not designed for that. They don't have the bus dividers for it.

sorry, i just edited my post to say maximum 66 MHz.

swaaye wrote:

Yes. Ok. But Cyrix and AMD wanted you to try to run these chipsets at 75, 83 or 95MHz in some cases and the chipsets were not designed for that. They don't have the bus dividers for it.

Cyrix and AMD didn't make CPUs with FSBs of more than 66 MHz until years after Intel launched these chipsets. This is the 'super socket 7' platform, not to be confused with the regular 'socket 7' from Intel (just because they kept the socket doesn't mean all CPUs are compatible. People often mention the 'advantage' that AMD keeps sockets around for longer, but ignore the fact that you can't always use new CPUs on old motherboards with the same socket. Things like FSB, voltage, TDP may change over time. Intel is more strict. Generally if the CPU fits in the motherboard, it is actually going to work).
Third-party chipsets from vendors such as SiS and VIA were designed for these new FSB speeds.

Cyrix 6x86 was 1995-1996. 😁

Same thing happened with the SIS 5591, it was certified to 83 MHz, but motherboard makers shipped it out as the first 100Mhz super 7 chipset, best it could ever do was 90-ish at a push.

Scali wrote:

swaaye wrote:

Yes. Ok. But Cyrix and AMD wanted you to try to run these chipsets at 75, 83 or 95MHz in some cases and the chipsets were not designed for that. They don't have the bus dividers for it.

Cyrix and AMD didn't make CPUs with FSBs of more than 66 MHz until years after Intel launched these chipsets. This is the 'super socket 7' platform, not to be confused with the regular 'socket 7' from Intel (just because they kept the socket doesn't mean all CPUs are compatible. People often mention the 'advantage' that AMD keeps sockets around for longer, but ignore the fact that you can't always use new CPUs on old motherboards with the same socket. Things like FSB, voltage, TDP may change over time. Intel is more strict. Generally if the CPU fits in the motherboard, it is actually going to work).
Third-party chipsets from vendors such as SiS and VIA were designed for these new FSB speeds.

The Cyrix 6x86 200 was a 75 MHz FSB part and that was before the Intel TX chipset.

swaaye wrote:

Cyrix 6x86 was 1995-1996. 😁

My point exactly, Pentium is 1993-1995. Intel had moved on to P6 by then.