testing and verifying overclocking

[Twospruces]

I am working on validating some overclocking modes for the TRS-80 Model 100.

To date, I have just been letting the system run at predefined clock rates, with pretty large jumps. I can tell when it really dies, but it is harder to say what is a reliable overclock setting, because it is statistical.

In order to get a better handle on this, I've been thinking I should find a way to generate or synthesize a finely tunable clock, so I can very accurately change the frequency / bus cycle time.

Then if I had a good way to measure an "error", I could plot Error Rate vs Cycle time.
Looking at such a thin in log-linear, I could then extrapolate to a suitable low error rate.


Question:

What would be a good way to measure a bus error? Any thoughts?

Any other ideas on how to deduce "safe overclocking"?


thanks!

 

[Ruud]

Every system has it own reaction to overclocking: in some systems the hardware will fail, in others the CPU will fail. The only thing what you can do is to write a program that test as many things of your system as possible in a huge loop. The moment errors occur or your system stops, you know you have reached the maximum frequency.

Personal question: why do you want to overclock that Tandy? I see it a bit of trying to tune an old T-Ford to make it run 35 MPH instead of 30. (actually I have no idea what its max speed is) A reason could be curiosity but in this case I don't consider it as a valid reason: you can blow up the engine/system by doing this and then you end up empty handed. And that answers your second question: IMHO there is no way to find out what is still safe.

Just popped up: if a firm finds out that a CPU that has been branded to run at X MHz can safely run at the higher frequency Y MHz, it will will surely rebrand them to Y MHz to have better chances in the market. So IMHO per definition the by the manufacturer given frequency is the max frequency.

 

[Timo W.]

Just popped up: if a firm finds out that a CPU that has been branded to run at X MHz can safely run at the higher frequency Y MHz, it will will surely rebrand them to Y MHz to have better chances in the market. So IMHO per definition the by the manufacturer given frequency is the max frequency.

You would think that, but that's not how it was done. Faster CPUs were often branded with a lower speed if market demand for lower-speed CPUs was higher than the output of such CPUs. Motorola was most known for doing this, but it was common practice.

 

[Twospruces]

My interest is personal. Just because. I have been able to increase from 2.5 MHz to 6MHz so far and continuing to explore.

I have developed a process..
1) a good frequency synthesizer capable of fine tune clock period
2) a test program that reports time-to-failure

I will plot average time-to-failure vs clock period and extrapolate to a low error level.

 

[Gary C]

Certainly there was a rake of branded DX4-80's that ran very happily as DX4-100's