Raspberry Pi 4 Can Now Overclock to 2.147 GHz. Here's How.

Now try a higher voltage with 7-Zip and report the results. If you could also do emulator testing and report on the stability with the new overclock. At least basics like SNES with Star Fox and Top Gear (both are very resource intensive for Pi Zero/W and prone to change performance with overclocking) [Yes different model, maybe with PS1?].
 
Nice article.

But I've found that the most demanding load is a linpack test. When testing at 1900MHz, I had unstable system with over_voltage=3, but not with over_voltage=2. Thermal images confirmed, that after sufficiently cooling the main BCM, USB and ETH chips, the PMIC (power management - switching voltage changer) becomes the hottest. Hence these reboots - are overload of the PMIC. See https://www.coolipi.com/Overclocking.html

What a pity Alvram haven't used our heatsinks for the test that were sent to Gareth. Would be a nice read :)
 
Now try a higher voltage with 7-Zip and report the results. If you could also do emulator testing and report on the stability with the new overclock. At least basics like SNES with Star Fox and Top Gear (both are very resource intensive for Pi Zero/W and prone to change performance with overclocking) [Yes different model, maybe with PS1?].

No, try with all over_voltages, even lower. As I've written, if you hit a power limit, decreasing over_voltage may help.

But this case may be the other situation, you're not limited by power delivery and need to increase the voltage to get the main chip to whatever the highest speed it can run at.
 
Now try a higher voltage with 7-Zip and report the results. If you could also do emulator testing and report on the stability with the new overclock. At least basics like SNES with Star Fox and Top Gear (both are very resource intensive for Pi Zero/W and prone to change performance with overclocking) [Yes different model, maybe with PS1?].


So I tried this with up to an over_voltage of 16 and it didn't help.
 
Nice article.

But I've found that the most demanding load is a linpack test. When testing at 1900MHz, I had unstable system with over_voltage=3, but not with over_voltage=2. Thermal images confirmed, that after sufficiently cooling the main BCM, USB and ETH chips, the PMIC (power management - switching voltage changer) becomes the hottest. Hence these reboots - are overload of the PMIC. See https://www.coolipi.com/Overclocking.html

What a pity Alvram haven't used our heatsinks for the test that were sent to Gareth. Would be a nice read :)

Gareth is a freelancer for us and lives in another country so we definitely don't send cases back and forth. However, I don't think this would have helped, because even with the Pimoroni Fan Shim, I never saw the core temp go above 74 degrees so, as long as it stays below 80, we won't throttle.
 
  • Like
Reactions: NightHawkRMX
But we send cases forth and don't need them back :) So if you feel you'd want to try it - we can send more. Might be interesting.

One thing is lack of throttling, stability another. At higher temperatures, the stability of all the components is lower.

CooliPi 4B heatsink with a Noctua 60mm fan running at 3.3V can cool the BCM chip to a temperature about 8-9 degree Celsius hotter than ambient. With 4 core linpack load. So I think with these relatively low temperatures (on air cooling), any added instabilities due to high temperatures should not manifest themselves.

What interesting benchmarks and tests do you guys suppose?

I'll try to overclock the CPU to 2147MHz and let you know what happened. I'll try it with the fan on and off.
I have a phoronix test suite installed, I'll try it. But it's not a power hog, some critical tests finish faster than the heatsink's thermal equilibrium time (about 30 minutes in passive mode).