I think any time you increase the voltage, it's a given you are decreasing it's lifespan. Going from 2.8 to 3.2 is actually a pretty hefty jump too. I usually read accounts of people bumping voltages up 0.1V to get a stable overclock. 0.4V isn't anything I would do, unless I were aiming for some sort of benchmark record and then backing off for daily use.

Aside from that, if you don't have to touch the voltage to go from 66x3 to 100x2, it should be fine? I'm running a 233 MMX at 2.5 x 100 instead of 3.5 x 66 to no ill effects myself. Didn't have to touch the voltage at all, and I've spent long enough with it to solidly believe it's totally stable.

Namrok wrote on 2022-11-28, 18:36:

I think any time you increase the voltage, it's a given you are decreasing it's lifespan. Going from 2.8 to 3.2 is actually a pretty hefty jump too. I usually read accounts of people bumping voltages up 0.1V to get a stable overclock. 0.4V isn't anything I would do, unless I were aiming for some sort of benchmark record and then backing off for daily use.

Aside from that, if you don't have to touch the voltage to go from 66x3 to 100x2, it should be fine? I'm running a 233 MMX at 2.5 x 100 instead of 3.5 x 66 to no ill effects myself. Didn't have to touch the voltage at all, and I've spent long enough with it to solidly believe it's totally stable.

Thanks for the quick response
So, for setting the 100mhz FSB I don't need to touch the voltage settings, I'll probably keep it that way, then.
The motherboard allows me to jump from 2.8v, which is the recommended setting for this CPU, to 2.9v and then the next available option is 3.2v. Tried with 2.9v but it's still not stable at 250mhz... maybe 3.0v would be enough but there is no such setting, sadly.

As old as this equipment is now I really don't see the reason to push those components to the brink. Especially when something faster is easily available and pretty cheap as well. Why do you need more than 2 x 100?

MMX work without any problems even on older boards with 3.3v only, just put a decent heatsink on it.
I have 233 MMX running on Asus P55TP4XE from 2001, no issues.

You can run your MMX at 2.5x100, no worries.

_tk wrote on 2022-11-28, 20:28:

As old as this equipment is now I really don't see the reason to push those components to the brink. Especially when something faster is easily available and pretty cheap as well. Why do you need more than 2 x 100?

No specific reason, just wanted to tweak this old PC and get a few more fps on Quake.

I agree that when you jack up the voltage you shorten the life of the chip, same if they run too hot. Thing is how many hours do you expect to use that machine and are you using a more modern heatsink that keeps the CPU cooler then the OEM ones?

Old chips made on large processes tend to last longer at stock voltages then something made today at bleeding edge processes putting out a hell of a lot of heat. Electron migration due to heat is what will kill newer CPUs over time.

Short answer is yes. I don't feel the need to over clock on older hardware when it is pretty cheap to get some of this stuff in the first place. Just upgrade 🤣. Sad truth is eventually all of this hardware will fail over time and I don't want to speed up the process on my collection. I'm sure things that are cheap now will become valuable in time to come.

gaffa2002 wrote on 2022-11-28, 21:20:

_tk wrote on 2022-11-28, 20:28:

As old as this equipment is now I really don't see the reason to push those components to the brink. Especially when something faster is easily available and pretty cheap as well. Why do you need more than 2 x 100?

No specific reason, just wanted to tweak this old PC and get a few more fps on Quake.

I get it and I remember trying to tweak my MMX processors for the same reason in the late 90's. But now, I don't see the need to overclock ~25 year old hardware. These parts likely won't last forever and why rush them to their demise? Especially when Klamath and Deschutes Pentium II processors are cheap and run Quake like a champ.

I did a Win95 C build with a PII 333 and LX chipset motherboard specifically for that era of games (and it technically still qualifies as a Win95 era build since all those parts were released before Win98).

For a PMMX, which is the model you have 350nm? I guess they were still 3.3V rated for IO. That could explain the ability to over volt so high. But that it could run at 100mhz ? I don't know about that. If it could then why didn't they just go for 100mhz if they bothered to put IO capable for that? They had certain design requirements, one would be how much power could be delivered by the system itself, and that they probably wouldn't have tested at that speed then for that as that wouldn't have been in the specification. Niether was to bump up internal frequency at the 100mhz level with the same multipliers. I guess that it helps that the chip is dual voltage. I guess it's a weird fluke based on socket 7's large configurability at the time, and why AMD went that route as an easy selling point, which doesn't help performance much.

I suspect it's a weird fluke based on the design, and really, it wasn't intended to run at high fsb speeds of the super socket 7 era, and you could shorten the life of the IOs. I mean c'mon, Intel probably does not overbudget their IOs... see here:
https://newsbeezer.com/czechrepubliceng/intel … -after-a-while/

I know more than I'd want to know about Intel's practices that their models really can't catch problems as they develop over time. I wouldn't overclock anything for a few fps, even if it could tolerate the voltage at 350nm.

During that era, IIRC, the 66MHz FSB wasn't really a limiting factor for software of the day. Quake (and TNT cards) liked CPU speed more than FSB speeds.

If anything I'd run it at 3.5 x 66 if you must overclock it. No more, however.

My CPU model is FV80503200 SY060. From the comments its clear that I should not mess with the voltages and keep it at 2.8v, so I'm keeping it at 2.8v. Btw I'm using a cheap P3 cooler.
Still not sure about the 100mhz FSB, but I might keep it at 2x100 instead of 3x66. 3.5x66 is not an option since this CPU does not support it (it sees it as 1.5).
Strange thing is that I saw a nice performance boost moving from 66 to 100 fsb, something around 10fps in quake. Didn't expect such a difference.
Edit: Is it possible to run the CPU at 66x3, but keep the motherboard bus at 100? Maybe the performance difference I'm seeing has more to do with the motherboard bus running faster?

I have a few P200 mmx and a P233 mmx and all will run all day at 100x2.5 2.8v. Maybe Yours has already been damaged somehow if it needs 3.2v for 250mhz.

Imperious wrote on 2022-11-29, 00:15:

I have a few P200 mmx and a P233 mmx and all will run all day at 100x2.5 2.8v. Maybe Yours has already been damaged somehow if it needs 3.2v for 250mhz.

This is also a possibility, I used to mess a lot with this motherboard once I had a k6-2 450 on it. I suspect I might have ran the MMX on it with the k6-2 voltage (3.3v) by mistake for a little while but changed it to 2.8 once I realized it.
The plastic piece in the motherboard which used to keep the heatsink in place on top of the CPU is also broken, so I had to glue the heatsink to it using thermal glue.

I ran a P166 MMX at 3.5x66 for quite a while and no issues whatsoever. This thread does a breakdown about the Pentium MMX's that were/were not locked:

Re: Pentium MMX fixed multiplier?

Then again, this was all on the Intel HX and TX chipsets. Not sure what the Ali chipset is up to but other than SiS, I avoided anything that wasn't Intel during that era (mainly bad drivers and odd compatibility/performance with certain programs).

Funny to see that most people who say don't do it only say it based on nothing than theory (which isn't wrong in itself but certainly does not apply to specific case). But I guess it is always that way: there are people who talk and there are people who do 😀

Paadam wrote on 2022-11-29, 06:38:

Funny to see that most people who say don't do it only say it based on nothing than theory (which isn't wrong in itself but certainly does not apply to specific case). But I guess it is always that way: there are people who talk and there are people who do 😀

As was mentioned, overclocking a Pentium MMX was pretty common back in the day. We all know that it works and what can be done. We've done it, period.

However, it's now 25+ years later and thus those same electronics are 25+ years old. Will it still work? Likely, assuming all parts are still within spec. Is it a good idea? Debatable. I'm not going to run any of my MMX processors at an overclock anymore because, as said, PII/PIII parts from that era are still relatively cheap and faster than an overclocked MMX will ever be. And I've owned my MMX processors since new and frying them doing a silly overclock is not something I want to do.

Otherwise, do you have any experience or knowledge to add to this?

Paadam wrote on 2022-11-29, 06:38:

Funny to see that most people who say don't do it only say it based on nothing than theory (which isn't wrong in itself but certainly does not apply to specific case). But I guess it is always that way: there are people who talk and there are people who do 😀

Yep. These were some of the most over-engineered chips in history, limited to 3.5x 66MHz by marketing (fear of out-competing more expensive P2 chips) rather than technical limitations. In cases like this 'stock speeds' are a political statement, nothing more. Yes, theoretical technical life will be shortened by running the chip at higher speeds/voltages than lower ones, but a good many CPUs sold at closer to their limits will be worse affected by running at said stock speed than many of these P55C with goodly overclocks. Based on this line of reasoning, you should run any K5, 6x86 or high-end Athlon CPUs at a significant underclock. In many cases the motherboards they are run on will be pushed further than the CPUs themselves. Moreover they are very common CPUs so even if something does go inadvertently wrong, the loss to posterity is minimal. I strongly suspect there will still be thousand of working P55C CPUs - including ones run above stock speeds - left after the last motherboard they could run on has long died.

I cant say I even saw Intel CPU killed by electro migration caused by OC, something that worked fine overclocked until it died.
Overvoltage? Yes definitely, a couple. Funnily enough cant remember a single Intel killed by heat, but then again I never build or sold a single Pentium 4 😀. Rest was user error - mechanical screwdriver damage, ripped/bend pins, crushed smd components.

AMD sure. Fried by heat, overvoltage, crushed cores, CPUs that with time overclocked worse and worse. AMD was always flying closer to the sun.

It will decrease the lifespan so it will die within 2 hours - 300 years, give or take a century. Be careful!