It's about like this (please correct me if I'm wrong here)
Lifetime left tells you about the wear of the cells. So lets assume you have a disk with 1000 P/E cycles NAND. That means, every cell of the NAND can be programmed and erased
at least 1000 times. So, lifetime left is the percentage of available P/E cycles across all cells. So if all cells on average have 600P/E cycles left, this means that lifetime left indicator is at 60%. No matter how many TB have been written to the disk.
In TLC NAND, 1 cell stores 3 bits, QLC 4 bits.
I don't know what Apple uses for the M1, but lets use QLC as 4 is easier to calculate with for this example.
Now assume, you're writing 100GB (100,000,00 Bytes) of data. This means you're writing to
at least 25,000,000 QLC cells.
This assumes the disk is brand new and has no data on it. For now it looks linear.
But:
If the disk gets more used over time, the controller uses something called Wear leveling. Certain cells are written to less often than others. So they get worn out faster. To prevent that, the controller will re-arange data while you write to wear out the disk more evenly. This causes additional writes. This is called write amplification.
View attachment 1814556
Source:
https://en.wikipedia.org/wiki/Write_amplification
So if you write 100GB to half full and well-used disk, wear leveling causes additional writes. So, lets assume the average write amplification for this 100GB write is 2. This means, this write would cause 50,000,000 cells to be written to. So it "wears out" twice as many cells as when the disk was new and fresh.
Also important to note: Writing can be done to pages, which are mutliple cells. Erasing can only be done to blocks, which are multiple pages. So this means, if you write data, a whole block needs to be read into the controller's RAM and modified in-memory. Then the whole block has to be erased and the whole modified block has to be re-written. This causes more writes than you actually wanted. Typically it's around 32 tp 128 pages per block. See:
https://en.wikipedia.org/wiki/Flash_memory#NAND_memories
If this is a erased block, which means all pages of that block are empty, then the page can be written without any of that above. But the fuller your SSD gets and the more you use it and modify files, the more this will happen. And causes more NAND writes than actual host writes.
The TBW is data written by the host (aka: your computer.). It's the data that was sent to the device. Not the data that was written to the NAND. Sending 100GB to the SSD can cause it to write (e.g.) 200GB in total as it has to re-arange data for wear leveling purpose or because you write smaller parts than can be erased.
If your disk is always 80% full, and you constantly write to the pther 20% free space, your Lifetime left will go down much quicker with less TBW than if you have 50% free and only write 20% of the disk. This way it has more free cells and doesn't need to shift data round that much.
So. TBW and Lifetime left do not correlate each other linearly. It's way more complex. And this is the reason why most SSDs have such "low" TBW ratings, even though the NANDs can do more. The manufacturer calculates with a worse case of write amplification as the TBW value is often used for warranty purpose.
As the controllers get better and better and more stuff like compression is used, write amplification can be reduced more and more. And therefore, SSDs last longer with more TBW, even though "lower quality" (means: rated for less P/E cycles) is used.
Write Amplification is not gone and will probably never go away. We don't know about the specs of apple's SSD controllers. They made their own since the T2 macs. And Apple isn't know for boasting tech specs for such things.
