Monday, March 1st 2021
Apple M1 Processor Manages to Mine Ethereum
Ethereum mining has been a crazy ride over the years. In recent times, it has become very popular due to a huge surge in Ethereum prices, following those of the main coin currently present on the market - Bitcoin. However, Ethereum miners use a customized PC stocked with many graphics cards to mine the Ethereum coin. Any other alternative is not viable and graphics cards have a high hash rate of the KECCAK-256 hashing algorithm. But have you ever wondered could you mine Ethereum on your shiny new Apple M1-equipped Mac? Our guess is no, however, there are still some people making experiments with the new Apple M1 processor and testing its capabilities.
Software engineer Yifan Gu, working for Zensors, has found a way to use Apple's M1 GPU to mine Ethereum. Mr. Gu has ported Ethminer utility to Apple's macOS for Apple Silicon and has managed to get GPU mining the coins. While technically it was possible, the results were rather poor. The integrated GPU has managed to get only 2 MH/s of mining power, which is rather low compared to alternatives (desktop GPUs). Being possible doesn't mean it is a good idea. The software will consume all of the GPU power and it will limit your work with the GPU, so it isn't exactly a profitable solution.
Sources:
Yifan Gu Personal Blog, ETHminer M1 Port (GitHub)
Software engineer Yifan Gu, working for Zensors, has found a way to use Apple's M1 GPU to mine Ethereum. Mr. Gu has ported Ethminer utility to Apple's macOS for Apple Silicon and has managed to get GPU mining the coins. While technically it was possible, the results were rather poor. The integrated GPU has managed to get only 2 MH/s of mining power, which is rather low compared to alternatives (desktop GPUs). Being possible doesn't mean it is a good idea. The software will consume all of the GPU power and it will limit your work with the GPU, so it isn't exactly a profitable solution.
17 Comments on Apple M1 Processor Manages to Mine Ethereum
but 20 or 40 A15 chip on a card pcie ? this is revolution !!!!
a keybaord is 130 key 1 key = chip size
when 17,4 billion transistor 3070 does 60Mh ~ RTX 5050 on 5nm.
Mining with M1
Now there is a disturbing ad. The ad-blocker is going to come back on if that doesn't go away on it's own.
No? Let us still shove this Apple ad down your throat!" Amazing stuff...
In a typical PC.. to do CUDA calcs, you'd do something like this..
You start with a data structure in system RAM, often an array of data (or it's not worth using the GPU).
1. You create an identical data structure in the GPU card's RAM.
2. Copy array data from System to GPU's RAM
3. Do your work on data in the GPU's RAM.
4. Copy your results from GPU RAM back to System RAM.
5. Destroy the data structure on the GPU (clean up).
On an M1.. System RAM is GPU RAM. More importantly.. they share the same memory space so instead of copying the data to be worked on, you just need to pass a reference to the original data structure.
You get to ignore steps 1, 2, 4 & 5. Instead, you just go straight to having the GPU do your work on the data in System RAM.
I'd be interested to know how optimized the M1 port of the mining software was.
And just for my own curiosity (why I ended up here), I'd love to know how a properly optimized mining client would run on the M1 Ultra chip that Apple announced today. The M1 in an Air ships with 7 or 8 GPU cores. The new M1 Ultra chip ships with 64 GPU cores and way more memory bandwidth.
The new Apple Studio computers with the big M1 Ultra Chips start at $4k so I doubt we'll see mining farms made out of them either. I'm just curious how the on-die GPU on the new chip would hold up against a large fast dedicated RTX GPU.. for the same ultimately pointless reason I've watched videos of a Tesla Plaid racing muscle cars.. even though I own neither. :)
.. just occurred to me to compare the published theoretical performance. Apple's claimed GFlops perf puts their new chip in the ballpark of an RTX 3070.