CPUs take power in spikes. This is why capacitors are very important -- to satisfy the demands of power hungry chips quickly. When you overclock, the demands on the capacitors increase, but if the little resistor inside each capacitor (known as ESR) is too large, the capacitor cannot fully satisfy the needs of the CPU, and the result is a freeze or crash, which you may interpret as "I can't seem to overclock any higher." This is what Drake and Kay experienced on the LC575 motherboard until they swapped out their tantalum capacitors with OS-CON. We know that is true on the LC575 motherboard, but I don't have enough data from other people to know how it applies to other motherboards. In theory though, lower ESR capacitors should help. But in practice, well, I need the owners of those other computers to chime in during their overclocking experience to know for sure. Kay has a lot of experience, which is why I directed my previous post at him.
When overclocking on a modern Intel or AMD PC, there are numerous factors that make up the formula.
1. The quality of the individual CPU. Different CPUs individually allow different settings, and often require a specific voltage individual to that CPU in order to overclock. Heat also plays a role here.
2. Motherboard quality and ability. Some PC motherboards are better at overclocking than others, mostly due to design of how it is setup but also what components are used (including capacitors) and cooling of those components.
3. RAM quality. If memory is not fast enough or high enough in quality, it limits the upper limits of stability.
4. Power delivery. All of the above is dependent upon a quality reliable and stable power supply.
5. Temperature. Lower ambient room temperatures allow the components to reach higher temperature without affect. Cooling of components such as the CPU greatly affect the ability to overclock.
On a modern Intel or AMD PC, all of the above are part of a successful overclocking recipe, and having one part being deficient results in the entire system being unstable.
I could see the same applying to vintage hardware. One component in upper limit testing is what is the mask of the 68040 CPU being used, as it’s lithography could affect performance. Newer mask chips (legitimate ones), should theoretically overclock easier and higher than older mask chips.
I have tested a variety of 68040 chips and analysed their individual temperatures at 33mhz and found a variation of temperature from over 80’c to as low as 48’c. That alone would affect the limits of overclocking beyond everything else. If used on a board with OSCON caps, a 90’c CPU at 33mhz would not overclock very well.
To obtain the fastest possible 68040 clocks, one would need the most ideal design Mac logic board, with the best performing on-board RAM chips, a late-mask 68040 CPU, a good cooler and some thermal paste, good on-board capacitors, and very stable power delivery. Cooling of that system would also be crucial to ensure stability.
On the topic of the CPUs, Motorola binned them based upon order requirements. If Apple needed 25mhz CPUs, Motorola would simply test a batch at 25mhz, and if they passed internal testing, would be marked 25mhz. Those that didn’t, would be retested at 20mhz and masked off at that speed.
If Apple ordered 33mhz chips, the same process would happen. It’s very possible a 40mhz chip has a 33mhz or even 25mhz printing because that is the certified limit. However, I don’t think Motorola ever thought of these running higher than 40mhz so they never would have a process for certifying them above 40mhz.
So what does that mean? A 68040 CPU that is marked 33mhz may very well outperform one marked 40mhz. I would say it more depends upon the mask revision than anything. A “true” L88M 68040 at 33mhz should theoretically outperform an older 40mhz mask (take your pick on which one). I say “true” because over the past decade, but even more so over the past several years, recyclers overseas have pulled 68040 and 68LC040 CPUs off old machines, ground down the surface and reprinted them as 68040 L88M 40mhz. Which would make them “fake”.
My initial findings for 68040 CPUs that I purchased on eBay to test can be found here:
https://tinkerdifferent.com/threads...fake-68040-cpus-especially-the-l88m-mask.820/