Maximum stable P2 clock - did someone try this?
pik33
Posts: 2,352
I am experimenting with the module player (SimpleSound) adding some MHz from run to run. The maximum I got was 376000000, which works, while 377000000 doesn't
The P2 eval board is now at thermal equilibrium playing modules for several hours, the ambient temperature is about 23C, the P2 is warm, but not hot, when touched.
Have someone else experimented with P2 max clock frequency?
Comments
Yes I know others like evanh and Tubular have played with the extremes in the early P2 days and there are threads that covered it somewhere (new forum search sucks), but you are in the ballpark of where things usually start to fail. I try to remain under 350MHz myself when overclocking with my video stuff and it typically worked well to that value.
On PCs, overclockers increase the CPU core voltage. I wonder if that'd work here...
Maybe yes, but this doesn't give much more MHz, and instead it gives more heat. Then they use liquid nitrogen
No, I will not overclock further. If 377 seems to be top, let consider 360 is safe value. I want to get as high HDMI resolution as I can: 960x600 works now, 1024x576 should be proper 9:16
If you ever have strange problems, slow down and see if that makes them go away. The 1.8V core will fail all at once if the speed gets too high, but I've noticed the 3.3V I/O pin ADC's getting very noise at over 320MHz.
About HDMI, you can signal some monitors at 30, and even 24, frames per second, cutting the required speed in half, allowing for higher resolution.
The symptoms I observed were:
The simplest audio demo, which produces 2 tones on 2 channels works until 400 MHz, then audio frequencies slow down when declared frequency goes higher (I tried 410 MHz) - this means PLL reached its limit while the core still worked.
I didn't try ADC yet.
Our early test results confirm about the same limit, Pik33, details here
https://forums.parallax.com/discussion/comment/1448299/#Comment_1448299
The freeze spray let us get past an initialisation, after that hurdle things kept running. These were 'glob top' chips so can have slight marginal differences to the main silicon
Like Chip said, there was a 'wall of failures' with many things failing all at once. It would be worth increasing the 1v8 rail to 2.0v and see whether that allows a little bit more.
1024x576 @ 50 Hz can be achieved with 330 MHz on a P2.
I found something similar when working on my SD Driver (there’s an old thread).
I found catastrophic failure which was around 380MHz. Serial stopped working and I though the base instructions just started failing on masse.
Most of my experiences of crashes were on the revA silicon. My extreme clocking tests on revB were only that, a smartpin producing a 1/1000 clock to monitor on the scope. Mainly for testing temperature responses. No real programs running, so not much to see even if it did occasionally crash. More effort was put into voltage and power testing.
PS: My high power testing didn't seem to crash when thermally speed limiting, but then they were only ever running for a minute or so at a time. And, again, the data produced was never verified nor displayed in any way.