Re:Apple Silicon Macs can't boot from external drive if inte...
by ece on Wednesday November 03, 2021 @02:17AM
(#29091588)
Some PC motherboards and laptops also have dual BIOSes/BIOS recovery[1] to prevent bricking. You can build a redundant system capable of surviving failures. Apple just doesn't want to spend the cost/volume of a couple 32Mbit roms, or however big iBoot is. Yes, simple is good, but it doesn't need to be this simple at the cost of longevity and upgradeability of the device.
Look at how much TLC NAND has improved in the last ~5 years, and ask yourself if QLC will improve as much. It probably will.
[1] https://www.dell.com/support/kbdoc/en-us/000132453/how-to-re...
Apple-branded SSD module failures are pretty common. Bricked BIOS/EFI images are not.
I've been supporting Mac users for two decades and the only time I've used a hardware SPI flasher was when I needed to pull my firmware image, modify the EFI bootloader to include NVME drivers, and flash it back. It wasn't hard. I attached a clip to the motherboard that connected the chip to a Raspberry Pi.
Seriously, my first mac was an LC. I've never seen a Mac brick itself doing a firmware update.
I had to do that because my Apple SSD died about 5 years into owning the system. That system uses an m2 PCIe port but it's a weird pinout. Apple doesn't sell parts, the local apple authorized shop wanted $250 just for diagnostics, the only thing on ebay were used drives of the same and, and only one company makes an overpriced replacement aftermarket drive for Apple's weird pre-NVME pcie storage connector and they have compatibility issues, underperform compared to much cheaper standard NVME drives, etc.
Thankfully, you can plug in a standard NVME drive via a $10 adapter from ebay.
I was back in business for a fraction of what a used 512GB Apple drive would have cost me, with twice the storage and even more speed over the already fast Apple module. But I had to pull NVME drivers from a later revision of the system and flash them to my ROM.
What pisses me off the most in the whole thing is that Samsung and Apple designed the drive such that when it runs out of write cycles, it completely fails; it doesn't even show up on the PCIe bus.
I'm typing this now on that very system, which is now almost exactly 8 years old. Do you think there will be 8 year old T2 macs? Not if they see daily use.
Repeat after me: user storage for a computer should never be permanently attached.
> This kind of failure (high voltage parts close to low voltage ones, close to liquid damage prone areas, without any underfilling) is documented by Louis Rossmann.
> It is, however, not always feasible or possible to entirely prevent.
It is 100% feasible/possible to entirely prevent.
Don't permanently attach storage on non-mobile devices.
Use conformal coating, something Apple could definitely afford to do with their profit margin on these machines. Even if they just coated the high voltage stuff. In fact, there's no reason T2 mac owners can't buy a can of conformal spray off Digikey or Mouser and do it themselves.
Don't put high voltage components near low voltage ones; especially ones that have user data and are non-replaceable. There's a reason Apple used to have a board for the DC input.
Every T2 mac has a finite lifetime before it becomes completely useless scrap thanks to SSD wear. And it's even worse for the M1 people because reportedly they do a great job of wearing out their SSDs.
This isn't about boot flash reliability or tiny increments in security-for-user-privacy. It's about slowly walling everyone in. Every new update of MacOS, every new hardware update - has put another brick in the wall.