I still think it's pretty likely we'll see a work around, but if that work around is a custom made firmware patch (like it was for the driver update they deployed IIRC) then that will probably mean the chips you purchased have a less rosy long term cost as future driver updates won't be applicable to your altered firmware. If it requires any physical tinkering with the chip then it definitely increases the effective cost by forcing a labour/card cost and opening up the possibility of expected chip defect as you break some proportion of the hardware you purchase in the process of fixing it.

That all said - yea some people are totally going to hack that or my name isn't 46 DC EA D3 17 FE 45 D8 09 23 EB 97 E4 95 64 10 D4 CD B2 C2.

100% this in the short term, but in the long term workarounds will become harder to find. Have a look at iOS jailbreak history.

I just hope they don’t kill functionality which matters for other meaningful purposes, like deep learning.

There’s not money to be earned by jailbreaking iOS that increases in value the harder it is to jailbreak.

Yes, there is currently a ~$1m public bounty, probably more if you know where to look privately. You’d have to check your ethics at the door in both cases, but they do exist.

Nvidia did not say it will be a hardware limitation. In fact, they will almost certainly implement it in software. For example a "Lite Hash Rate" card will probably have a bit permanently set to 1 in the firmware/eeprom. The driver will read the bit, and arbitrarily enforce the restriction on such LHR card. This solves potential legal issues of retroactively crippling cards already out in the market, since only new cards sold from now on as "LHR models" will have the bit set to 1.

But miners will find another software hack (just like they did for the RTX 3060 earlier) to bypass the restriction.

IIRC wasn't that software hack based on NVIDIA accidentally releasing a driver version that supported the 3060 but didn't enforce the LHR?