The CPU doesn’t really care about the FSB that’s more of a motherboard overclocking thing if you adjust the multiplier to a core speed that is appropriate for what the CPU can handle you will do fine

If your CPU is 2.2 V cyrix, then you have the 4X multiplier you would just have to enable it

But if you are running 100 MHz front side bus 400 MHz probably is not going to be stable

The best you can hope for on a 2.2 V cyrix is probably 350 MHz

Re: Cyrix 400gp multiplier settings

Sphere478 wrote on 2025-10-22, 04:01:

The CPU doesn’t really care about the FSB that’s more of a motherboard overclocking thing if you adjust the multiplier to a core […]
Show full quote

The CPU doesn’t really care about the FSB that’s more of a motherboard overclocking thing if you adjust the multiplier to a core speed that is appropriate for what the CPU can handle you will do fine

If your CPU is 2.2 V cyrix, then you have the 4X multiplier you would just have to enable it

But if you are running 100 MHz front side bus 400 MHz probably is not going to be stable

The best you can hope for on a 2.2 V cyrix is probably 350 MHz

Re: Cyrix 400gp multiplier settings

Thanks a lot
Unfortunately I don't have a 2.2V MII CPU, but only the 2.9V ones, in particular the 333 is the fastest I have.
Thanks to Jan who provided me with a correct BIOS, CPU 333 can also use the 4X 66 setting, instead of the original 3.5X 75, so 266 MHz instead of 263.
Unfortunately today 2.2V MIIs are expensive, in the past instead when they were cheap, you could buy an assorted batch, in which there were some 2.2V copies, today I don't think it's possible.
I was asking about the FSB above 100, to know if it would create problems, basically if you don't exceed the factory frequency, everything should work without problems, at least hopefully.
Jan's correct BIOS, I needed it for my ACORP A-5VIA3P (VIA VPX) PC, after turning it on I had problems with the drive and stability, the PC crashes shortly after loading MS-DOS 5.00.
I performed some tests at lower frequencies, and at 233 MHz it's fine, with multi 4X I tried at 240 MHz, and it works but some programs have problems, and maybe the PC crash takes longer.
Maybe it could be a problem of linear regulators, which heat up too much, or it could be a problem of VCORE, slightly lower, there is no 2.9V but only 2.8V.

Your best luck finding a 2.2V Cyrix MII is with one of the silver tops, like this: https://www.chipdb.org/img-cyrix-miiv-333gp-greytop-591.htm There was an abundance of these CPUs about 8 years back on eBay. They sold for $20-25.

As for FSB, I run my MII at 3.5 x 95 MHz because 3.5 x 100 wasn't quite stable on the motherboard I wanted to use. Plus, 3.5 x 95 was faster than 3 x 100. Enable some of the Cyrix register settings, pair it with a 3dfx Voodoo, and you'll have an acceptable pre-millennium gaming system.

I haven't come across an in-depth analysis to determine whether a Cyrix 6x86MX or MII rated for 66 MHz FSB is fully stable at 100+ MHz FSB. I find it peculiar that Cyrix would rate their MII-333 chips at 3x83 MHz rather than 2.5x100 MHz. It suggests to me that there was some aspect of their qualification tests which didn't like 100 MHz.

feipoa wrote on 2025-10-22, 09:07:

Your best luck finding a 2.2V Cyrix MII is with one of the silver tops, like this: https://www.chipdb.org/img-cyrix-miiv-333gp-greytop-591.htm There was an abundance of these CPUs about 8 years back on eBay. They sold for $20-25.

As for FSB, I run my MII at 3.5 x 95 MHz because 3.5 x 100 wasn't quite stable on the motherboard I wanted to use. Plus, 3.5 x 95 was faster than 3 x 100. Enable some of the Cyrix register settings, pair it with a 3dfx Voodoo, and you'll have an acceptable pre-millennium gaming system.

I haven't come across an in-depth analysis to determine whether a Cyrix 6x86MX or MII rated for 66 MHz FSB is fully stable at 100+ MHz FSB. I find it peculiar that Cyrix would rate their MII-333 chips at 3x83 MHz rather than 2.5x100 MHz. It suggests to me that there was some aspect of their qualification tests which didn't like 100 MHz.

Thank you very much, yes I wanted to buy one of the silver ones, from memory I don't remember if it was a 2.2V, but they were in high demand, I also have one of the black ones, a MII 300, I don't think it's different from the gold ones, in the collection I have almost everything Cyrix, except for some 6X86 or 6X86L from IBM.
Interesting this thing on the MII with FSB 95, faster than the 100 one, some ASUS motherboards for example, they also have an FSB 97, I don't know if it would be worth trying, of course at most they are about 8 MHz more, but who knows if it's okay.
In fact, it's strange, but on the Internet I saw that faster versions had to come out, including MII 466 500 533, I think they reach 350 MHz, unfortunately they didn't arrive on the market, maybe VIA chose the Cyrix III, and left the S.7 preferring the S.370, but I think that without an integrated L2 cache, it was even slower than Celeron equivalents.
I don't have the Cyrix III, while I have a pair of VIA C6, I tried a 600 MHz one, it totally disappointed me, it goes roughly like a 200 MHz (or a little more), the IDT derivation of this CPU, makes me think that the IDT Winchip 200, also goes slow.

I haven't messed with my Cyrix CPUs in a while (and was playing more with 486 sockets most recently a few years back), but when I was heavily tinkering with Super Socket 7 CPUs, I realized multiple of my 250 nm (mix of IBM and NS process) 2.9V rated parts would overclock like crazy, typically 233 or 250 MHz rated parts to 3x100 MHz without stability issues when running at stock voltage and windows 98 SE (plus multiple hours of playing Unreal Gold in software render mode, plus in 3DFX mode on). Asus P5A-B with Voodoo 3 2000.

I also discovered they would undervolt well with that motherboard but not necessarily with all (I had some issues on an MVP3 based board, I think Shuttle brand one, but forget right now, and didn't get around to testing on the FIC 503+ I have, or 503A). I routinely ran the 6x86 366 (2.5x 100 rated) at 2.5x 100 on 2.2V without issue as well. (same board has also had good luck running 2.2V K6-III 450 parts at 500 MHz at 2.0V without issue ... that's K6-III 250 nm, Sharptooth, not III+ 180 nm) Caution note on 250 nm AMD parts though: they instantly die at 3.3 or 3.5V, probably at around 3V too. I mean instantly die and might short out the motherboard. (I did that by accident with the jumper settings one time, no POST, and made the breaker buzz in the unusual AT PSU I was using at the time, one that has tripped from shorts before and saved me from destroying a 486 motherboard and CPU at least once due to socket mis-insertion, plus another when a tantalum capacitor failed short ... but 250 nm K6 wise, either it fried the chip itself, or managed to short circuit the surface mounted caps on the chip shorted, not sure; but AMD's 250 nm parts are nothing like the IBM and NS parts in late model 2.9V rated 6x86s)

Note: 350 and 250 nm 6x86 parts were rated at 2.9V and SOME 250 nm parts were also rated 2.2V while other 2.2V parts were 180 nm, so you need to do more digging using the identification guide on CPUShack.

Those 2.9V 350 nm chips I tested also overvolted for prolonged periods at 3.0 to 3.2V (the same voltages I was using for Pentium MMX 233 chips around the same time, though those are 350 nm) and didn't have signs of failure or damage. At the time I hadn't actually realized they were 250 nm and not 350 nm parts, as this was just when I was getting into CPU testing with retro tech stuff around 2010. And that MII 366 was the first Cyrix chip I bought. I vaguely recall trying it all the way up to 3.5V, but I'm not 100% sure on that and as I recally it didn't help pushing the speed any higher than 300 MHz beyond the POST screen.

I also haven't gotten that board to run reliably above 115 MHz FSB, and 115 is hit and miss (though it's hard to tell when it's CPU vs MB limits or a combination). 110 MHz tends to work reliably for CPUs that tolerate it, and I recall getting 330 MHz that way out of that Cyrix MII and maybe another 2.9V rated one, but didn't make it to windows before hanging. (plus it was probably at a higher voltage than I was comfortable with sticking to, and it ran at 300 MHz at 2.9V without issue, so I stuck with that ... I think it also worked at 2.8V at 300 MHz)

I've had no problem running 350 nm cyrix parts at 3.5V for longer periods, though, but mostly did that with 6x86L-200 (150 MHz rated), maxed out at 3x68 MHz (208 MHz) but stable ... also tried running Unreal Gold that way. I believe I've tried 5 different 6x86Ls that way and all did 208 MHz and ran benchmarks + games in Win98SE. I know I also walked back the voltage to see where the minimum overvolt was, but forget the nuanced results. I recall at least some working at 3.3V.
AMD's 350 nm parts also seem to tolerate 3.5V fine (which makes sense given the K-6 233 was factory overvolted to 3.2V and there was at least some rare 3.3V marked examples), plus I believe the same process was used for the AMD 5x86 (or 486 x 5) CPUs which have been pushed to well over 4 volts, though also known to fail at 5V, though not instantly like my 250 nm K6 accident. (I've also had Cyrix DX4-100 486s and PR-75 100 Mhz 5x86s running at 5V without any signs of destruction, but those are on larger fabs than that, .68 micron I think for both cases, and while the absolute rating voltage is below 5V in the MSDS for those chips, I never ran into tendency for failure, but also only ran the 5x86 at 2x66 MHz 5V for a few hours, mostly trying it in DOS running the floppy disk version of X-Wing for extended periods, but also without any of the register enhancements enabled; it ran at 2x60 MHz at 3.5V so I spent more time at that, while I'd already been running multiple of those DX4 100 CPUs ... specifically ST marked ones, not Cyrix marked, for months in an older 5V only 486 board, running at 4x40 120 MHz, playing games on it or running demos almost daily with no issues ... plus one instance where the heatsink had come lose and I hadn't realized it, so it was running really hot, but didn't self-destruct: it was in a tower and the plastic heatsink clamp or clip came half-off, so there was roughly 1 mm of gap between the CPU and heat sink)

All that said, and back on topic with the 2.2V rated Cyrix parts. I haven't risked overvolting the few 180 nm examples I have (several PR 366 266 MHz rated parts, and a single PR 400 3x83 MHz rated part) and also never got around to seriously stress testing the clock speed limits of those. I stumbled on some good ebay deals for those well after I'd given up on hunting them down, but then they popped up from some overseas sellers at reasonable prices. (I believe the same seller that appeared to have a ton of new old stock coming out of China or Hong Kong ... at some point around 2020 they had ST DX4 100 CPUs going for $7 plus offered combined shipping, so I'd picked up a handful of those cheap and was less afraid of burning out ... initially at least; the single Cyrix 5x86 I overvolted significantly was sold for scrap in a batch of surface mounted QFP to PGA adapted 5x86s with bent, mangled, or lifted pins and traces, but about half of them could be un-bent and un-shorted to make function, others probably work but I haven't gotten around to trying de-soldering and re-soldering with an air pencil ... need more practice with that)

I also can't recommend risking the same overvolting I've done, as it's obviously not fool proof. I can just say what my own experiences have been. (plus I would definitely not go above 2.6V for 250 nm AMD parts, from both modern and 90s/2000s era anecdotes, that's the max overvolt that actually works long-term and 2.8V has been known to work for a while and eventually kill the chip ... 2.6V is usually what's required to make K6-2 550 parts run at 600 MHz ... and I haven't tried pushing that rate 2.2V rated 250 nm K6-III 450 to 2.6V to see if it will go to 600, or even 550. I was happy using it at 5x100 at 2.0 or 2.2V for extended periods ... plus someone had lifted the lid on that one at some point, and mounting it was kind of funky ... I didn't try a direct die mount for fear of shorting the chip on the heatsink, so I applied thermal paste to the lid and placed it back on)

feipoa wrote on 2025-10-22, 09:07:

Your best luck finding a 2.2V Cyrix MII is with one of silver top.
I haven't come across an in-depth analysis to determine whether a Cyrix 6x86MX or MII rated for 66 MHz FSB is fully stable at 100+ MHz FSB. I find it peculiar that Cyrix would rate their MII-333 chips at 3x83 MHz rather than 2.5x100 MHz. It suggests to me that there was some aspect of their qualification tests which didn't like 100 MHz.

Want an even stranger one?
The Cyrix pr350 was spec’d at 90mhz x 3x, which meant they were only compatible with an obscure PCCHIPS motherboard. (And date code to the cusp of and early pc10 era)

The majority of pr333 “Cyrix “ chips were actually silver tops made by an ibm fab. (Most silver tops of specific grades were made in ibm fabs regardless of how they were marked even after the so called transition) as far as I know IBM never sold pc100 - pr366 chips only the pr333-83mhz chips that were both on sale at the same time.

Why IBM/Cyrix kept selling 75/83mhz FSB chips was twofold, these chips released in limited numbers before pc100 was fully standard. These chips could allow sleazy beige box OEMs to clear out older cheap motherboard designs like the PCCHIPS m571 that kept finding its way into low end cheap machines through 1999

The true 250mhz Cyrix chips (national ) were mostly sold as pr366 100 mhz x 2.5x in both gold and silver tops.

Later they were sold as 6x86v-pr366, obviously in the via era.

Do you have a photo of a silver top PR-366?
The only silver tops I've ever seen were made by IBM at PR300 (3.5x66) and PR333 (4x66).

Anonymous Coward wrote on 2026-04-18, 01:25:

Do you have a photo of a silver top PR-366?
The only silver tops I've ever seen were made by IBM at PR300 (3.5x66) and PR333 (4x66).

I’ve only ever seen the silver top PR333 83x3 in person.
Never saw a 4x66
I’ve been told there were silver top pr366+ and owned a silver top pr433v (VIA). I also had a black top but don’t remember the specifics.

Based on the 366+ that were silver, it seems they are all likely VIA, I don’t think anyone other than IBM OR VIA Made silver tops but I’ve seen the claim the end of production nationals were silver tops. (Never seen physical evidence though)

The 366 I’ve seen in persOn were gold top assumably National sourced and 2.9 volts.

Via is noted for making e low volt 366 versions, so I assume that’s where the confusion came from.

Almost never see mobile versions... there were a handful of cheap laptops with them. I assume laptops with 5X86 used the QFP

I have two gold-top 2.9V 83MHz PR333s. They came from Compaqs with the Desert Storm motherboard. Only one works, and I don't have a better motherboard to mess around with it. I'd like to see what it can do without permanently-enabled SiS 5598 integrated graphics stealing memory bandwidth.

I was looking at some other Cyrix chips and ran into a 3x83MHz chip marked 350. That must have been when National Semiconductor changed Winstone versions.

I seem to remember one of the hardware sites back in the day, possibly SysOpt.com before it had it's millennium refresh, claiming that PR was at first based on pentium classic integer scores, then changed to covington celeron.

lti wrote on 2026-04-19, 04:00:

I have two gold-top 2.9V 83MHz PR333s. They came from Compaqs with the Desert Storm motherboard. Only one works, and I don't have a better motherboard to mess around with it. I'd like to see what it can do without permanently-enabled SiS 5598 integrated graphics stealing memory bandwidth.

I was looking at some other Cyrix chips and ran into a 3x83MHz chip marked 350. That must have been when National Semiconductor changed Winstone versions.

PR350’s of any flavor are rare.
What region was it from?

You should grab all the serials and see if you can use the decoder to see what model, year process node it is.

Would also be worth submitting scans of it to cpushak or the collectors forums as your particular flavor is noted but no pics

The PR350 was just pictures on the Internet. I don't know where they came from, and they don't show the bottom. The 333 is the fastest I've seen in person.