anyone who has experience OCing CPUs will know that at some point you need to up the voltage to keep going... well, in the case of these Z9M chips, you can't do that so as you OC the MHz up and up, it will hit a ceiling and crash. trip the VRMs enough times and the darn thing will go into a "i don't want to play anymore, i quit" mode where it won't hash at all and you will think you have bricked your miner. if and when this happens, just power down the miner, exhaust all the power from the miner, disconnect the psu, exhaust the power from the psu, and fire it back up (or just disconnect everything cables and all and let it sit off for 15-20 minutes).
if you want to go above what Bitmain has set, you will need to have some electrical engineering background or knows somebody who does and have some equipment, like a USB-PMBus dongle tool at the minimum to regulate the voltage as well as needing to solder some connectors to certain points on the PCB itself. this is no easy task as you need to be able to read electrical schematics let alone know what the hell you're doing so you don't fry your boards on the first power-up after you mod it. FPGA-guys are already doing this as they really need to regulate voltage when they push their chips while mining. of course this isn't the only way as the direct interface to the PCB may allow modification.
also, heat is going to be an issue if you can actually accomplish what i mentioned above. just like OCing CPUs, the higher it goes, the better fan heatsink or waterblock you better have. i recommend liquid immersion but that's just me. if you disregard this, your chip will fry really quick.
either way, take what i said with a grain of salt cause i am still a noob in regards to this forum and have zero trust so what do i know, eh? just an observation, nothing personal, hehe...
i'm pretty sure at least 99% of you Z9M owners will not be attempting this and the ones who have done this will keep it to themselves and not share out. why give more hashing power to the rest to increase difficulty, right? welcome to the world we live in. happy mining!
I'm thinking you may be correct. It's also possible they may have skimped on MLCCs (power supply decoupling capacitors) in the newer batch, thus creating more ripple voltage at the chips, and lower overclocking potential.
The MLCC market is EXTREMELY tight right now, and prices are going up, and availability is way down.
So it may be as simple as replacing a few capacitors per hashing chip. I began to suspect this because I own an electronics manufacturing company, and am an electrical engineer. But I have yet to verify (or attempt to verify) my suspicion.