Cyrix chips didn't have a lot of margin in them, clock-speed wise. IBM had a reputation for leaving a little more margin in theirs. It was easier to find an Intel or AMD chip with overclocking margin in it, so that was why Cyrix chips weren't popular with overclockers.

With newer cooling methods, Cyrix chips might fare better now than they did in their heyday. Cyrix chips did run hot and I found them less tolerant of low-quality power supplies, so using a rock-solid power supply would probably also help.

RetroSpector78 wrote:

What would be a safe / responsible way to continue ?

Not to overclock it at all. What's the point anyway? If you want more speed, install a faster CPU instead of running that one out of specs and risking to damage it.

The 2.2 volt PR433 I have overclocks easily.

Too bad they didn’t release it 3 years earlier

Dave Farquhar wrote:

Cyrix chips didn't have a lot of margin in them, clock-speed wise. IBM had a reputation for leaving a little more margin in theirs.

I was surprised to get it up to 187 and consider that a win.
Would like to check the temps somehow. Will see if I can add a probe somewhere.
Dunno if there are certain measures I can take to prevent the chip from getting killed (or is such a thing not possible and is overclocking (even in very small increments) always a risk (even if you ensure that the cpu doesn’t get too hot ?

Any thoughts on the 66 vs 75 mhz FSB ? The 66 x 2.5 runs at the stock voltage (and as such cooler).
What would be the impact of a slower FSB here vs the higher clock speed when compared to stock ?

derSammler wrote:

RetroSpector78 wrote:

What would be a safe / responsible way to continue ?

Not to overclock it at all. What's the point anyway? If you want more speed, install a faster CPU instead of running that one out of specs and risking to damage it.

It’s just a bit of fun. As this is a vintage forum and overclocking has some historical value I find it interesting to experiment with these things today. IM

RetroSpector78 wrote:

Dunno if there are certain measures I can take to prevent the chip from getting killed (or is such a thing not possible and is overclocking (even in very small increments) always a risk (even if you ensure that the cpu doesn’t get too hot ?

There are always risks involved with overclocking, including stability. You won't kill the chip as long as you keep the heat at an acceptable range. The main thing you can do is control the temperature.

RetroSpector78 wrote:

Any thoughts on the 66 vs 75 mhz FSB ? The 66 x 2.5 runs at the stock voltage (and as such cooler).
What would be the impact of a slower FSB here vs the higher clock speed when compared to stock ?

The higher FSB speeds up access to memory and it definitely helped performance in the Socket 7 days. A Pentium running at 75 MHz x 2 could keep pace with a Pentium 166 at 66 x 2.5. A Pentium at 83 MHz x 2 could keep pace with a Pentium 200 at 66 x 3. So, you can figure approximately a 10%-15% performance hit by using the slower FSB while keeping the clock rate as close as possible. The 75 MHz bus definitely helped, even if it did push the PCI bus out of spec.

Cyrix Socket 7 CPUs are interesting to mess around for "academic" purposes. But it's not something you would want to use overclocked in a daily retro PC for games.

The Serpent Rider wrote:

Cyrix Socket 7 CPUs are interesting to mess around for "academic" purposes. But it's not something you would want to use overclocked in a daily retro PC for games.

It is kinda an academic exercise. I also want to do it in a controller way where I don't leave the thing unattended for a long time, and control what kind of apps / games I run it. Ideally I would also like to check the temps in realtime somehow. I have a temperature probe on my multimeter (a wire), but not really sure if that is the way to go. Putting in on the center of the CPU will cause a substantial gap between the CPU and the heatsink/fan. Placing it off-center is better but might give inaccurate readings.

So not really sure what the way forward is. Looking for an application (running on msdos / win98) that is able to read the cpu temps. I'm guessing the CPU will store its internal temp somewhere in a register that can be read ... hopefully someone create an app for that (cyrix specific), as apps like CPU-Z and HWMonitor don't show anything.

Dave Farquhar wrote:

There are always risks involved with overclocking, including stability. You won't kill the chip as long as you keep the heat at an acceptable range. The main thing you can do is control the temperature.

First step would be to find a way accurately measure the temperature. Using my temperature probe from my multimeter, putting in on the center of the CPU results in a substantial gap between the CPU and the heatsink/fan. Placing it off-center is better but might give inaccurate readings. So looking for an app / program that is able to read the temperature from the cpu directly. (if such a thing exists)

Dave Farquhar wrote:

The higher FSB speeds up access to memory and it definitely helped performance in the Socket 7 days. A Pentium running at 75 MHz x 2 could keep pace with a Pentium 166 at 66 x 2.5. A Pentium at 83 MHz x 2 could keep pace with a Pentium 200 at 66 x 3. So, you can figure approximately a 10%-15% performance hit by using the slower FSB while keeping the clock rate as close as possible. The 75 MHz bus definitely helped, even if it did push the PCI bus out of spec.

Are there any specific benchmarks that cover this type of FSB performance ? Or will this performance gain simply show up in regular benchmarks and will most app / games benefit from it.

I don't remember, whether electromigration was an issue back in the days with 350 or 600 nm processes? But this will drastically speed up when increasing the voltage ..

null

I would think that if you can put a socket A cooler or a big socket 370 cooler on there it should be enough to offer peace of mind regarding CPU temperature. In terms of wattage the Cyrix chips would never come close to a 60W+ Athlon, even with a small bump in the voltage.

But maybe one of those infrared thermometers pointed at the heatsink would at least provide some measurements to compare?

rmay635703 wrote:

The 2.2 volt PR433 I have overclocks easily.

Too bad they didn’t release it 3 years earlier

Where did you find an MII-433? Those are quite rare!

RetroSpector78 wrote:

Are there any specific benchmarks that cover this type of FSB performance ? Or will this performance gain simply show up in regular benchmarks and will most app / games benefit from it.

It showed up in regular benchmarks and most apps/games benefited from it, as I recall.

I'm really surprised to see such a thread, because...
back in the era, those Cyrix chips were often called "factory overclocked" 😁

Anders- wrote:

Sure, you just need a thin temperature probe (something like the "cpu" probes bundled with the digital doc 5) - remove a small amount of heatsink material to get the probe right on top of the center of the cpu. Depending on how much heat said chip produce it might make sense to put the probe somewhat off center, in order not to diminish cooling performance (on a hot socket A system I'd settle for keeping the probe next to the core, sacrificing correct temp reading rather than heat transfer).

So you mean using a tool like a Dremel to cut a path in the heatsink material for the temperature probe wire to pass, to ensure that there is sufficient surface contact with the cpu ?
Cause that is indeed the current problem ... with the temperature probe in between the CPU and heatsink I'm reading 70 degrees celcius (= max operating temperature) when the cpu is running stock, but I can imagine that the heat transfer to the heatsink is pretty terrible. If I press my finger on the little visible part of the CPU it also gets quite hot. However, with the heatsink installed without the probe, I don't "feel" any heat at all. Heatsink remains very cool , and also the visible part of the CPU can be touched.

Anders- wrote:

This is the best way to get an accurate temp reading for old systems not featuring cpu integrated temperature probes.

Are you aware of any Cyrix/IBM specific tools that could read the temperature from one of the CPU registers ? I'm guessing that there is some kind of sensor in the CPU and this value should be available somewhere.

Grzyb wrote:

I'm really surprised to see such a thread, because...
back in the era, those Cyrix chips were often called "factory overclocked" 😁

Consider it an academic study 😀 Just wanted to play around with it a bit ... I was able to get it up to 187,5Mhz (a 25% increase from the stock 150Mhz), and there is a big improvement in the benchmarks.
It also survived the benchmarks (dos and windows) and the CPU isn't getting extremely hot as far as I can tell.... but trying to figure out the best way to measure the temperature now.

bakemono wrote:

I would think that if you can put a socket A cooler or a big socket 370 cooler on there it should be enough to offer peace of mind regarding CPU temperature. In terms of wattage the Cyrix chips would never come close to a 60W+ Athlon, even with a small bump in the voltage.

What do you consider a "small bump" ? The stock CPU runs at 2.9v and to get it stable at 187.5Mhz I need to bump the voltage to 3.2v. At this point the only thing you can do (I think) is to keep a close eye on the temperature ? Is it safe to say that the bump in voltage in itself is not enough to kill the CPU ?

bakemono wrote:

But maybe one of those infrared thermometers pointed at the heatsink would at least provide some measurements to compare?

The heatsink itself doens't get hot at all .... I can comfortably hold it with my fingers .... Doesn't come close to feeling hot.

Is it safe to say that the bump in voltage in itself is not enough to kill the CPU ?

3.2v is not enough to kill that CPU. I would start worrying after 3.5v.