What is the latest world record of an overclocked AVR?
I just managed to run my ATmega32A-PU at 50MHz (!) for a few minutes. And during those minutes I managed to get it to boot properly into a test program (which itself is non-trivial, using one timer and the UART, all interrupt-driven) and load an updated test program from ISP (!!!)
I used a 4-pin crystal oscillator to clock the chip. The chip is fused to use external clock directly.
ATmega328P is a chip clocked at 20MHz maximum but ATmega32A can be clocked at 16MHz maximum according to the datasheet, so clocking that chip at 50MHz is definitely more overclock than ATmega328P at the same clock speed.
I do not know if anyone has kept a running log, but 50MHz does sound speedy. I did note that writing to eeprom at high-clock may present an opportunity for damage.
technix:
What is the latest world record of an overclocked AVR?
I just managed to run my ATmega32A-PU at 50MHz (!) for a few minutes. And during those minutes I managed to get it to boot properly into a test program (which itself is non-trivial, using one timer and the UART, all interrupt-driven) and load an updated test program from ISP (!!!)
I used a 4-pin crystal oscillator to clock the chip. The chip is fused to use external clock directly.
ATmega328P is a chip clocked at 20MHz maximum but ATmega32A can be clocked at 16MHz maximum according to the datasheet, so clocking that chip at 50MHz is definitely more overclock than ATmega328P at the same clock speed.
It's interesting idea. What happened during overclock period? So hot, strange behavior or anything else.
More than likely because the micro no longer works till it has cooled down enough.
Robert10:
It's interesting idea. What happened during overclock period? So hot, strange behavior or anything else.
One of the problems is that you give the internal electronics not enough time to reach the logic levels. As a result instructions can read wrong from flash, data read from SRAM can be read wrong and so on, writing to SRAM will write the wrong data and so on. It will read / write a zero when it was supposed to be a one and vice versa.
I have Arduino with ATmega1284P overclocked at 24MHz and it works reliably, no heating, no errors. Max. frequency for 1284P according the datasheet is 20MHz. I've read that the 16MHz Atmegas usually run up to 30MHz without the problems.
@technix
Why don't you test for a longer time? Provide more info about.
... it's all very interesting and I'm sure it is "fun" to cook a 386P ... pushing it toward thermal breakdown. But, I routinely run my ESP8266 at 160Mz and it is fat & happy the whole time; the 80MHz and 160MHz settings are available through the ArduinoIDE with the ESP8266 core installed. The 12E and 12F modules are under $2 U.S.D. and are ARM-design 32-bit processor. If you do not require the RF section of the module, you can just disable it in software and the 80MHz current drops to around 40mA which is on par with a 72MHz STM32F103.
Point being, if there is a point here, is that you can overclock an AVR 8-bit until it turns cherry red and it still will not out perform a 32-bit ARM design. To be fair to the workhorse 8-bit AVR, the 32-bit chips do not sink/source nearly as much current as the 8-bit older AVR, so in many situations the 8-bitters makes for a simpler hardware design.
@mrburnette
In my case, it is just curiousness to find the limits of ATmega's reliability. I understand that the 8-bits are not competitor for the ARM chip. However, we should bear on mind that ATmega's chip technology is old and probably will not improve in near future, otherwise the chip frequency could be pretty high even now.
