Hacker News Reader

Comment by brianpgordon | Hacker News Reader

> 2018 MacBook Pro Has Fastest SSD Ever

This title is misleading. I have a faster SSD in my year-old home desktop. They say in the article that they mean the fastest SSD they've ever seen in a production laptop.

View on HN

No you don’t, that speed is for an OS level copy not the advertised max seq write speed.

I have the 960 pro in raid and I don’t get more than 2000-2200 MB/s when copying a files of that size on an actual file system.

Based on the copy speed vs advertised speed I would put the max seq write speed of that drive at least in the 3000 MB/s range this is at least as fast as the 970 evo in its largest capacity if not faster since Apple tends to use Sammy SSDs I’m betting that this is a newer version of the PM981 which might some day be released as the 980 evo or even pro.

My XPS 15 shipped with a PM981 (1TB) and while it is fast it isn't as fast as the speeds in the article. However I think this is more down to Apple "cheating" at the file system level with delayed write and snapshots or whatever other tricks they are using.

Also the PM981 is the same tech as used in the 970 EVO. Samsung have also had the SM line previous which was the same tech as used in the PRO line but there doesn't seem to be an SM981 so maybe they are no longer doing OEM MLC?

It’s not the PM981 but it quite possibly be a newer version of it, the WD Black reaches speeds of 2800 MB/s in its 1TB version so it’s not unreasonable that a Sammy refresh of the 981/970 will be at least as fast as the newer WDs.

And while Apple can cheat at least on the mid 2015 15” MBP with a 1TB NVME I’m not getting any speeds that aren’t in line with my drive this is on a Mojave preview with APFS.

I will have to test on my 2018 MBP when it arrives in a couple of weeks. Will be interesting to see side-by-side against the PM981.

You can't infer anything from the copy speed. The article mentions that it is using the new apple file system apfs. Which includes a clone feature that upon copying a file, a reference to that file is created. The file system then only actually stores changes made to individual copies. So I'm surprised it actually took 2 seconds.

The disk benchmark shouldn’t copy just write and it had managed to get a bit higher speeds, I have and older version of these drives which isn’t that much slower and I can get simmilar speeds in Windows.

Running Black Magic Disk Mark in my mid 2015 MBP with an NVME SSD with Mojave preview and APFS does produces valid results if this would be just a soft link copy I would get the same speeds even with a slower NVME.

Really? Faster than 2,519MB/s? I thought they were achieving this by ganging together 2 drives via their proprietary T2 chip.

They admit in the article that a lot of this performance could be an illusion caused by the APFS instant cloning feature, which essentially just creates a new directory listing pointing to the existing file when you copy a file to the same volume. A separate copy of the data is only made when you make a change to one of the files.

The speed seems to be accurate for a high capacity PM981 or newer especially in raid Apple is just using bonker fast SSDs.

It also appears that it doesn't copy the whole file upon changing one of the files. It only stores the changed portion. I assume this is block level. Pretty cool. But yes, SSD speed has very little to do with copy speed in this case.

I wonder if this could be used to infer information that users may not intend. For example, perhaps copying a file and fraudulently altering something in it at a specific time could be used as evidence in a trial.

Isn’t this from the custom SSD controller they’ve been iterating over for years? [0] A variation of it even made its way into the iPhone [1]. Of course there are likely more hardware factors, such as binning.

0: http://www.iphonehacks.com/2015/04/apple-custom-ssd-controll...

1: https://bgr.com/2015/09/29/iphone-6s-macbook-ssd-storage-con...

Reply on news.ycombinator.com