Entry level M2 Mac mini, 2023 MacBook Pro have slower SSD than predecessors
The base model of Apple's newest Mac mini and the 2023 MacBook Pro have has significantly slower SSD read and write speeds because of engineering choices, compared to that of the previous generation models.
A teardown by Brandon Geekbit has uncovered the reason on the Mac mini -- the 256GB Mac mini comes with a single 256GB storage chip. Last generation's M1 Mac mini came equipped with two 128GB flash chips in parallel, allowing for faster speeds.
{"@context":"https://schema.org/","@type":"VideoObject","name":"M2 Mac Mini Teardown / Disassembly (4K - Up Close)","description":"Teardown on the brand new Apple M2 Mac Mini. Base Model 8GB Ram 256GB SSD.","thumbnailUrl":"https://i.ytimg.com/vi/H7nBlq-FlXE/sddefault.jpg","uploadDate":"2023-01-24T17:09:40Z","duration":"PT8M17S","embedUrl":"
"}
As it turns out, the 512GB M2 Mac mini also features a single NAND chip, meaning that it would still have slower read/write speeds than a model with two 256GB chips. However, as there was no 512GB Mac mini with an M1 Pro chip, there's no comparison to be made.
The situation is similar on the 2023 MacBook Pro as well, as discovered by 9to5 Mac. The lower storage capacity models have fewer Flash chips at a higher density than the previous model. This cuts down speed versus what it would be, versus what you get from parallelization of multiple flash chips.
We've seen this scenario before, too. For example, the M2 MacBook Air had significantly slower read and write speeds over the 2020 MacBook Air.
Read on AppleInsider
A teardown by Brandon Geekbit has uncovered the reason on the Mac mini -- the 256GB Mac mini comes with a single 256GB storage chip. Last generation's M1 Mac mini came equipped with two 128GB flash chips in parallel, allowing for faster speeds.
"}As it turns out, the 512GB M2 Mac mini also features a single NAND chip, meaning that it would still have slower read/write speeds than a model with two 256GB chips. However, as there was no 512GB Mac mini with an M1 Pro chip, there's no comparison to be made.
The situation is similar on the 2023 MacBook Pro as well, as discovered by 9to5 Mac. The lower storage capacity models have fewer Flash chips at a higher density than the previous model. This cuts down speed versus what it would be, versus what you get from parallelization of multiple flash chips.
We've seen this scenario before, too. For example, the M2 MacBook Air had significantly slower read and write speeds over the 2020 MacBook Air.
Read on AppleInsider
Comments
Some actual numbers though, would be useful.
Here's the motherboard of the M2 Pro Mac Mini (single 512GB chip in green rectangle). It looks to be flat on the table - no second nand chip on the other side.
From Brian Stucki https://www.macstadium.com/blog/first-look-mac-mini-with-m2-pro
*At least that's what Matt says regarding the 512GB models of the M2 MacBook pro and Mac Mini M2 Pro.
*
But then 9-5 Mac says otherwise about the 14" M2 Pro MacBook Pro's 512 GB model performance...
https://9to5mac.com/2023/01/24/macbook-pro-ssd-performance-drop/
What's going on here? Are 14" M2 Pro MacBook Pro with 512GB ssds being made in both ways? Some with single 512GB chips and some with 2x 256GB nand chips?
Are some European market Macs being made differently from US models?
When it comes to the M2 regular computers, things are a bit different. The issue is having 1 chip rather than 2 (vs 2 rather than 4 on the pro models)
So, while the new M2 Pro computers with the base 512GB storage do have fewer NAND chips than their M1 Pro counterparts, they actually have two rather than four. Not one, as was mistakenly stated in the article. The one chip issue is true for the M2 regular models however, so it's easy to get confused.
It's not yet clear whether the 512GB M2 regular mini has only one NAND chip rather than two, but I think it's unlikely that it would have only one. I think the author was getting confused between the M2 regulars and M2 Pro models (the 512 M2 Pro is reduced, but not from 2 to 1, but rather from 4 to 2).
The YouTuber MaxTech, who used to work for this site will be benchmarking all the different configs once they have all arrived. But so far he has only posted about the 256GB M2 Regular model. The 9to5Mac article did benchmarks with the 512GB M2 Pro model, and many reviewers posted results with the 1TB or higher models. What I have not yet seen are any benchmarks of the 512GB M2 Regular model. But I expect it to have similar performance to the M2 512GB config in the MBA and 13" Pro, which uses two NAND chips, not one like the 256GB.
Finally, it should be noted that there can be a lot of variance in performance of SSDs, even among the same capacity based on manufacturing variances and also sometimes Apple has used different vendors for parts for the same SKUs. For example, I have seen the M1 Pro SSDs benchmark as high as 7000 read and 5400 write, but the 9to5Mac author was only getting 4000 and 5000.
- as someone posted, it will be interesting to see the performance of the M2 Mini with 512 GB
- Apple has cost, chip availability and/or other reasons for designing the various computer configurations. That said, they are marketing the silicon in a big way, and the SoC is fixed … therefore, for clarity, no matter the computer product line the M2 Pro should have a standard level of CPU and disk performance for a given number of C & G cores.
The Tom's Hardware test is using 2TB MacBook M2 versions that Apple sent them. Are they using faster chips in the 2TB models than the low spec (and Mac Mini) models?
Where did 9-5 Mac get those low BlackMagic 14"M2 Pro MB (3150 MB/s) test scores compared to Matt Talks Tech (6300 MB/s)? Supposedly they both used 512GB models.
If one buys the 1TB M2 regular or Pro Mac Mini will it have BlackMagic scores in the 6000 MB/s range or the 3200 MB/s range?
"The M2 Pro chip starts with a 10-core CPU and 16-core GPU, 16GB of unified memory and 512GB of storage. Apple sent us a version to test that costs an obscene $4,099, but it offers obscene power to match with a 12-core CPU, 38-core GPU, 64GB of memory and 2TB of storage.
Thanks to our sister site Laptop Mag, we also have the test results for the M2 Pro version of this machine, although this is a pricier configuration than the entry-level model. It offers a 12-core CPU, 19-core GPU, 32GB of RAM and 2TB of storage for $3,299."
https://www.tomsguide.com/reviews/macbook-pro-14-inch-2023#section-macbook-pro-14-inch-2023-specs
Some typical speeds from reviews on Thunderbolt 4 enclosures from Amazon using Samsung 980 pro:
Versus a Samsung 980 pro inside a fast computer:
https://www.youtube.com/watch?v=1y7NJXZ-5kU&t=216s
Although some of us upgrade more frequently (and push the performance envelope), many of us are no where near power hungry users. I do understand there are creative types who may upgrade more frequently. Many of us will not soon be using 8K displays or USB4. Many of us just want a few more ports than the Air models, and the base level 14" seems to fit the requirements for the casual user (spreadsheets, email, web browsing and viewing videos). If the 13" mbp didn't have that gimmicky touch bar and had a few more ports, I'd likely go that way instead. I initially considered a M2Pro mini, but prefer to have the screen, camera, microphones, speakers, keyboard, touch pad and such bundled together, plus the internal battery (handy during power failures).
I can imagine Apple not being too fussed about non-pro disk performance, since these SSDs are crazy fast regardless, and you won't see that speed in real life.
I always have iStat Pro on showing HD speeds and of course if I copy files where it takes a while, I take a peek
I have never seen speeds exceed 150MB/s in real life (hundred 50)... nowhere near the 5,000 MB/s this MacBook Pro 16 with M1 Pro is supposed to be capable of.
So I am guessing that this raw speed is highly theoretical and probably only applies to specialist software applications, ie, moving editing.
Otherwise, the file system overhead prevents these numbers from getting anywhere even close to 1,000.
Unless this was in Mbit/s... which I doubt.. MB is usually Megabyte (8x Megabit)...
Try it - copy a folder with 100k items and see how it performs.
I have no problems with SSD speed by the way - in day to day, that's never a bottleneck. Only when really copying very large folders does the progress dialog even pop up. Most of the time everything is instant.
Real life folder - 10,000 items, 4.4GB
Manually tested this, it took 37 seconds to copy - translates to 120MB/s
That's around 40x slower than the max speed
So I am thinking in real life, the SSD speed is not the bottleneck for almost all operations, except maybe speed tests and things that work with huge volumes of data. Video,, 3D, and so on.
For the rest of us - we'll never notice if the SSD is 1,000 MB/s or 5,000 MB/s because either way the Finder is limited to 120MB/s for some reason.
"Chip count" is not the way to extrapolate speed. Chips can be formed into banks with each bank on a single data bus (channel). So whether there's one, two or eight chips in a single bank, it will always be the same speed.
You’ll see that it’s not finder. The speed will be very close to the black magic numbers.
The Finder is not limited to 120MB/s. There is a lot of file system overhead in allocating space, creating nodes and error checking for thousands of items, especially if some data needs be moved around to prevent data fragmentation. Copying a single large file would be much closer to actually testing the speed of the SSD.
However...
I do need to point out that APFS does not actually copy data until it needs to. Simply copying a file (on the same physical volume) in the Finder does not mean that data has been duplicated and rewritten in some other area of storage. The actual copy doesn't occur until one of the "files" has been modified.
So for instance you duplicate a file 5 times in the Finder. All that's initially created are new nodes that all point to the same block of data in storage. So no data has been "copied" yet. Now let's say you open the fifth copy and make changes and then go to save it... Only then is that block of data copied and saved with any modifications. That fifth node now points to this new block of data and the first 4 "copies" all still point to the original block of storage.