I really appreciate all these suggestions, I have some thinking and research to do.

I won’t be going with any of the FX options because of their reliance on SSE, which we’ve seen can severely peg back a card if not supported. CPU itself isn’t the strongest either. There is a cheap Radeon 7000 PCI and a reasonably priced MX 4000 near me, both are SSE compatible, but how much they rely on it remains to be seen. Need to find somebody who has tried one with a Pentium II era chip.

I recently picked up a new old stock Cyrix PR 166 on Ebay as I have a ton of Socket 7 boards and I'm really bored building intel/amd socket 7 machines.

I had the genius idea to install OS/2 on my first Cyrix build... Is OS/2 Warp 4.0 a good match for this cpu + 32 mb of memory? This CPU predated Warp 4.0 by a year, so I'm curious if it would be underpowered.

It would be a perfectly fine CPU for Warp 4, especially with 32MB RAM.

So having checked out some benchmarks, I think the issue with my FX card performing so poorly with the 6x86MX was always to be expected really. I was surprised that 3DMark 2000 was five times faster on my P3, but the Ultimate 686 benchmark demonstrates the same kind of disparity in 3DM99.

SSE or not, my CPU is quite simply a bigger bottleneck than I expected. Phil ran an FX 5200 with a k6-III, which also lacked SSE but still scored close to a P3-500.

So I might pair I the 6x86mx a Matrox or something.

That said, it shouldn’t matter too much, I built it to be a mid 90s beast, and that it is, it flies through anything pre 1998; Quake 2 in OpenGL, WC3, Fade To Black, even some Team Apache.

appiah4 wrote on 2024-04-20, 08:54:

It would be a perfectly fine CPU for Warp 4, especially with 32MB RAM.

Thanks, I have it running well, but what a pain to install.

Dan386DX wrote on 2024-04-14, 00:24:

Sorry if already asked: […]
Show full quote

Sorry if already asked:

Is there any advantage to either of these CPUs? They're both tagged '6x86MX PR300' and both run at the same voltage but the multipliers and FSB are different:

- One is gold-topped and runs at 3 x 75MHz.
- The other is grey, and 3.5 x 66MHz.

The gold one looks nicer 🤣 - but I went with the other one because I don't want to be suckered out of my extra 6MHz.

In my benchmarks 3x75MHz was faster in 640x480 dos benchmarks (PC-Player, Quake, Chris's 3d benchmark) and 3,5x 66MHz was faster in 320x200 benchmarks.

Also 3x75MHz gain more points in 3D Mark 99.

elmatero wrote on 2024-04-22, 17:33:

Dan386DX wrote on 2024-04-14, 00:24:

Sorry if already asked: […]
Show full quote

Sorry if already asked:

Is there any advantage to either of these CPUs? They're both tagged '6x86MX PR300' and both run at the same voltage but the multipliers and FSB are different:

- One is gold-topped and runs at 3 x 75MHz.
- The other is grey, and 3.5 x 66MHz.

The gold one looks nicer 🤣 - but I went with the other one because I don't want to be suckered out of my extra 6MHz.

In my benchmarks 3x75MHz was faster in 640x480 dos benchmarks (PC-Player, Quake, Chris's 3d benchmark) and 3,5x 66MHz was faster in 320x200 benchmarks.

Also 3x75MHz gain more points in 3D Mark 99.

Interesting, thank you. I can manually jumper (well, dipswitch) to try that configuration.

I can actually get it stable at 3.5 x 83 but only if I bump it up to 3.2v which is a touch higher than I’d like. At 290MHz it’s a significantly better chip, but I’m not sure what a safe voltage range is for these CPUs.

Dan386DX wrote on 2024-04-22, 19:23:

elmatero wrote on 2024-04-22, 17:33:

Dan386DX wrote on 2024-04-14, 00:24:

Sorry if already asked: […]
Show full quote

Sorry if already asked:

Is there any advantage to either of these CPUs? They're both tagged '6x86MX PR300' and both run at the same voltage but the multipliers and FSB are different:

- One is gold-topped and runs at 3 x 75MHz.
- The other is grey, and 3.5 x 66MHz.

The gold one looks nicer 🤣 - but I went with the other one because I don't want to be suckered out of my extra 6MHz.

In my benchmarks 3x75MHz was faster in 640x480 dos benchmarks (PC-Player, Quake, Chris's 3d benchmark) and 3,5x 66MHz was faster in 320x200 benchmarks.

Also 3x75MHz gain more points in 3D Mark 99.

Interesting, thank you. I can manually jumper (well, dipswitch) to try that configuration.

I can actually get it stable at 3.5 x 83 but only if I bump it up to 3.2v which is a touch higher than I’d like. At 290MHz it’s a significantly better chip, but I’m not sure what a safe voltage range is for these CPUs.

The microprocessor databook always tells you what you need to know:

https://datasheets.chipdb.org/IBM/x86/6x86MX/mx_full.pdf

As far as your specific situation:

Vcc2 is the volage used for the processor core. 3.2v would put you beyond the Recommended Operating Condition but still within the Absolute Maximum Rating, so I guess it's a judgement call if you feel safe operating it that way.

Some of these older chips, you can actually back down the voltage after "burning in" at a higher volt apparently necessary to reach a certain speed. i.e. wont boot that high at first until voltage raised, but after you raise voltage and run it a bit, it will boot at a slightly lower voltage now. That stopped being a thing from about coppermine era.

Paralel wrote on 2024-04-23, 14:35:

Vcc2 is the volage used for the processor core. 3.2v would put you beyond the Recommended Operating Condition but still within the Absolute Maximum Rating, so I guess it's a judgement call if you feel safe operating it that way.

Just to point out what this means, so you can do the judgement appropriately: Exceeding the absolute maximum even for a fraction of a second may irreversibly degrade or damage the chip. There is no guaranteed safety margin in those values. If it says "absolute maximum 3.3V", and during power-up before the regulator stabilizes, you get a 10ms pulse of 3.4V before it settles to 3.2V (hypothetical example), the chip is permitted to break quite soon. Staying below the absolute maximum will prevent permanent damage, but it does not necessarily mean that the chip operates correctly! If you want a guarantee for the chip to behave as specified in the data sheet, you need to stay within the recommended operating conditions.

This is true not only for the voltage, but obviously also for the clock frequency. As soon as you overclock your chip, you are violating some "recommended operating conditions" anyways, and you obviously shouldn't use anything operating outside of "recommended operating conditions" for mission-critical systems. Higher voltage will mean more energy dissipation in the chip, and more energy dissipation will mean that the chip runs hotter. A hotter chip is more susceptible to electromigration, which is one of the contributors to chip aging. So even staying below the absolute maximums will not yield "indefinite lifetime", and approaching the absolute maximum may shorten the usable life of the chip.

Paralel wrote on 2024-04-23, 14:35:

The microprocessor databook always tells you what you need to know: […]
Show full quote

Dan386DX wrote on 2024-04-22, 19:23:

elmatero wrote on 2024-04-22, 17:33:

In my benchmarks 3x75MHz was faster in 640x480 dos benchmarks (PC-Player, Quake, Chris's 3d benchmark) and 3,5x 66MHz was faster in 320x200 benchmarks.

Also 3x75MHz gain more points in 3D Mark 99.

Interesting, thank you. I can manually jumper (well, dipswitch) to try that configuration.

I can actually get it stable at 3.5 x 83 but only if I bump it up to 3.2v which is a touch higher than I’d like. At 290MHz it’s a significantly better chip, but I’m not sure what a safe voltage range is for these CPUs.

The microprocessor databook always tells you what you need to know:

https://datasheets.chipdb.org/IBM/x86/6x86MX/mx_full.pdf

As far as your specific situation:

Vcc2 is the volage used for the processor core. 3.2v would put you beyond the Recommended Operating Condition but still within the Absolute Maximum Rating, so I guess it's a judgement call if you feel safe operating it that way.

Appreciate the reply, yeah I saw that a few nights ago; my only concern was that I wasn't sure these values applied to the (later?) gold top chips or the black top ones like mine. I saw a forum post somewhere that seems to point out that there are a few subtle differences between them, the most obvious being the difference in suggested multi and FSB.

For now I'll stick to running it at stock volts and either default 233 or 266 as it's basically doing everything I need it to; but I won't sweat it too much when I go for 3.2v and 290MHz for the odd benchmark - especially when the new GPU turns up.

Dan386DX wrote on 2024-04-23, 20:28:

Paralel wrote on 2024-04-23, 14:35:

The microprocessor databook always tells you what you need to know: […]
Show full quote

Dan386DX wrote on 2024-04-22, 19:23:

Interesting, thank you. I can manually jumper (well, dipswitch) to try that configuration.

I can actually get it stable at 3.5 x 83 but only if I bump it up to 3.2v which is a touch higher than I’d like. At 290MHz it’s a significantly better chip, but I’m not sure what a safe voltage range is for these CPUs.

The microprocessor databook always tells you what you need to know:

https://datasheets.chipdb.org/IBM/x86/6x86MX/mx_full.pdf

As far as your specific situation:

Vcc2 is the volage used for the processor core. 3.2v would put you beyond the Recommended Operating Condition but still within the Absolute Maximum Rating, so I guess it's a judgement call if you feel safe operating it that way.

Appreciate the reply, yeah I saw that a few nights ago; my only concern was that I wasn't sure these values applied to the (later?) gold top chips or the black top ones like mine. I saw a forum post somewhere that seems to point out that there are a few subtle differences between them, the most obvious being the difference in suggested multi and FSB.

For now I'll stick to running it at stock volts and either default 233 or 266 as it's basically doing everything I need it to; but I won't sweat it too much when I go for 3.2v and 290MHz for the odd benchmark - especially when the new GPU turns up.

This databook was the 6th edition, July 29, 1998, covering all the released chips up to the 263 MHz model. I don't know when your chip was released, before or after that date.

Fx5200 pci with my pr-300 runs all benchmarks almost the same as G450 pci, but compatibility with older games is terrible for Fx series. For example Colin 1 and Motorhead had issues in menu. For me this no go for win98 retro experience. G450 or some early Geforce is the best option.

So I’m sorry to keep going on about my 6x86 build but I’m just looking for some confirmation really.

It’s now paired with a TNT2 M64, a 32GB variant on PCI.

Even at the best overclock, 290MHz, the 3DMark99 CPU score is about half that of the same bench with the same card but with a Pentium II 233.

6x86MX PR300 (290MHz OC) + TNT2 M64.

1267 - 3D marks
1266 - CPU marks

P2-233MHz + TNT2 M64.

1985 - 3D marks
2371 - CPU marks

I’m assuming the difference is expected and normal (because of the slow FPU?) - or is this a board/external cache issue? (on my current SS7 board it’s 512kb l2 cache @66mhz)

EDIT: actually, I completely forgot that the ultimate 686 benchmark comparison tested both of these CPUs on 3DM99; and their results tally almost exactly, 2394 for the PII and 1420 for the Cyrix MII which is very similar to my chip.

The 3DMark CPU scores can have a huge variance between machines with the exact same processor, so I do not personally rely on them for comparison purposes.

Dan386DX wrote on 2024-04-30, 18:13:

So I’m sorry to keep going on about my 6x86 build but I’m just looking for some confirmation really.

No need to be sorry. I am interested to see your results and progress.

Dan386DX wrote on 2024-04-30, 18:13:

So I’m sorry to keep going on about my 6x86 build but I’m just looking for some confirmation really. […]
Show full quote

So I’m sorry to keep going on about my 6x86 build but I’m just looking for some confirmation really.

It’s now paired with a TNT2 M64, a 32GB variant on PCI.

Even at the best overclock, 290MHz, the 3DMark99 CPU score is about half that of the same bench with the same card but with a Pentium II 233.

6x86MX PR300 (290MHz OC) + TNT2 M64.

1267 - 3D marks
1266 - CPU marks

P2-233MHz + TNT2 M64.

1985 - 3D marks
2371 - CPU marks

I’m assuming the difference is expected and normal (because of the slow FPU?) - or is this a board/external cache issue? (on my current SS7 board it’s 512kb l2 cache @66mhz)

EDIT: actually, I completely forgot that the ultimate 686 benchmark comparison tested both of these CPUs on 3DM99; and their results tally almost exactly, 2394 for the PII and 1420 for the Cyrix MII which is very similar to my chip.

It's a Cyrix CPU, what exactly are your expectations here? FPU performance is notoriously not great.

Garrett W wrote on 2024-04-30, 19:25:

It's a Cyrix CPU, what exactly are your expectations here? FPU performance is notoriously not great.

It was an experiment really, I'd seen underwhelming benchmarks in the past but wanted to build a 6x86 myself since the mid 90s; I've now seen for myself just how much of an edge Intel had during that period. I'm fond of the build though, it's great for DOS gaming up until about 1998.

You need a good setup and you're near Intel or better ...

A 32GB TNT2 M64 eh? 😅