M1 Mac SSD Wear Issue Tested and Explored [video](youtube.com)
youtube.com
M1 Mac SSD Wear Issue Tested and Explored [video]
https://www.youtube.com/watch?v=e8aCgnb_RQs&ab_channel=BarTechTV
24 comments
Isn’t this the YouTube way now however? Most videos are around 10 minutes or longer because that’s what YT thinks drives engagement, and therefore adverts? I’m sure there was something about videos over 10 minutes getting preferential placements in search results etc..
They’re like the 20 minute meeting that could have been a one line email. No thanks, I’ll take skimmable text any day.
Well if all you wanted to know was the truth the video title has it right there (chopped off from HN title) - M1 Mac SSD Wear Issue Tested and Explored. Yes, it's a problem.
I have YT Premium so without ads I found the video to be informative albeit a little slow paced. He goes through debunking various theories including the smart data being incorrect or misinterpreted.
I have YT Premium so without ads I found the video to be informative albeit a little slow paced. He goes through debunking various theories including the smart data being incorrect or misinterpreted.
What's the truth in the title? It says he tested and explored it. Eh.
Maybe it's a problem on 8GB units (and to that Richie video, my assessment about 8 vs 16 is that 8 shouldn't exist. That just isn't enough memory). On my 16GB device I've written 5TB despite installing all the OS betas, and pounding on XCode 13 betas all day. That puts it at 1% wear. I'm not too concerned.
Maybe it's a problem on 8GB units (and to that Richie video, my assessment about 8 vs 16 is that 8 shouldn't exist. That just isn't enough memory). On my 16GB device I've written 5TB despite installing all the OS betas, and pounding on XCode 13 betas all day. That puts it at 1% wear. I'm not too concerned.
What is the reason for this? Is there something about M1 that requires lots of data to be written/read from disk? If no, can we not expect that Apple will fix this with a software update?
I believe it's fixed in macOS 11.4... looks like it was a bug that was affecting the data reporting tools.
https://twitter.com/marcan42/status/1396374313591140357?s=21
https://twitter.com/marcan42/status/1396374313591140357?s=21
Guess you didn't see his second video:
https://www.youtube.com/watch?v=ZnHgdqU-1ws
The second video is crap too. It's all based on that the diagnostic tools shows the correct values of written data...
Have you tried the SponsorBlock browser add-on?
and obnoxiously cringe. He is the only techtuber that interview Apple employees and only one side at a time can be seen. LOL.
Large scale chia plotter here. I've used over 20 different brands of SSDs; you could say I "stress-test" NVMes for a living.
The "TBW warrantied" claims are not so meaningful. On every single NVMe, I have gotten at least 2-4x the rated TBW warrantied without encountering ANY issues, whatsoever.
The NVMe spec classifies the rating as being able to retain data, powered off, for 1 year. If your macbook is connected to power more than once in a year, you wouldn't expect to lose data. So real ratings are higher than claimed in the video.
I'd expect the Mac's NVMes to work for 4-8 years of heavy usage, before failure.
The "TBW warrantied" claims are not so meaningful. On every single NVMe, I have gotten at least 2-4x the rated TBW warrantied without encountering ANY issues, whatsoever.
The NVMe spec classifies the rating as being able to retain data, powered off, for 1 year. If your macbook is connected to power more than once in a year, you wouldn't expect to lose data. So real ratings are higher than claimed in the video.
I'd expect the Mac's NVMes to work for 4-8 years of heavy usage, before failure.
Which SSD would you say is your favorite?
The short answer is: every NVMe is different, some may be better at burst speeds, some may be better at sustained writes, some for endurance, and of course there's the question of value.
If you make me pick:
• I have nearly zero money to spend: MX500 2.5" or BX500.
• Value: Kingston A2000
• Balanced: WD SN750
• Endurance / professional scratch disk use: Samsung 970 Pro (NOT 980 Pro! Tlc instead of MLC), or Gigabyte Aorus, or Corsair MP600
These are consumer NVMes for desktop pc/workstation usage. If you're building servers, you probably want to consider server-grade hardware.
If you make me pick:
• I have nearly zero money to spend: MX500 2.5" or BX500.
• Value: Kingston A2000
• Balanced: WD SN750
• Endurance / professional scratch disk use: Samsung 970 Pro (NOT 980 Pro! Tlc instead of MLC), or Gigabyte Aorus, or Corsair MP600
These are consumer NVMes for desktop pc/workstation usage. If you're building servers, you probably want to consider server-grade hardware.
Thanks a lot! I’m planning on upgrading from my 2010 PC to the new Ryzen APU and am looking for some good storage. I appreciate your thoughtful response!
Chia plotting seems to very different workload compared to swap. TBW rating client workload is defined by JEDEC, I believe it's harder workload than Chia plotting (larger block size).
1. This video is old (April); Apple released a "fix" in macOS 11.4 in June. The author made a followup video showing that excessive writing still happens: https://www.youtube.com/watch?v=ZnHgdqU-1ws
2. This video guesses what NAND Apple is using instead of opening the case and looking. They also make assumptions about the endurance given that Apple doesn't specify it.
3. Total speculation: Heavy swapping combined with 16 KB pages is very bad?
2. This video guesses what NAND Apple is using instead of opening the case and looking. They also make assumptions about the endurance given that Apple doesn't specify it.
3. Total speculation: Heavy swapping combined with 16 KB pages is very bad?
>showing that excessive writing still happens
It only shows that excessive writing is still reported. Whether it still happens that's another matter.
It only shows that excessive writing is still reported. Whether it still happens that's another matter.
It's a reporting issue according to this article and has been fixed in MacOS 11.4
https://appleinsider.com/articles/21/06/04/apple-resolves-m1...
It's still not fixed.
The problem has only been reduced in macOS 11.4, but not solved. It's still really bad.
According to what? The reported written data counts? Who says those are accurate for the new architecture?
The main reason would be that it's only certain usage scenarios where M1 macOS disk writes fly off the handle, while others match the Intel ditto. What would be the apparent reason to you that both the OS' and S.M.A.R.T's readouts would match eachother yet still be selectively inaccurate on M1?
Different write patterns combined with different SSD controller cache logic?
Ain't no one got time for that.