Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

ryanflucas

macrumors regular
Original poster
Mar 28, 2006
148
17
Milwaukee, WI
It's safe to say that any Mac Mini model I buy will be faster then what I use now. I have a 21.5" iMac 2017 4K (16GB ram, 500GB SSD, Core i7 3.6GHZ, Radeon Pro 560 4GB, Mac OS Ventura). I'm considering replacement with a Mac Mini (M4 Pro chip with 12‑core CPU, 16‑core GPU, 16‑core Neural Engine, 24GB ram, 512GB SSD, 10 Gigabit Ethernet). However I keep reading about slow SSD speeds based on what memory chips are used. Does anyone know what capacity I would need to make sure I'm on the next fastest bus speed? Can that be 1TB?
 
Followup: I'll be using external SSD for most storage and backup to a hard wired Synology NAS. The internal drive would probably just be used for OS and Apps.
 
This video is a good discussion of NVMe drive performance and also shows how bad the 256 GB internal is.

This video compares 256 to 512 on the M4 mini, also 512 on the M4 Pro, and the Samsung 990 EVO Plus in an external Thunderbolt 4 enclosure.

An M4 mini Pro has Thunderbolt 5 so you can get better performance than the internal at 512 GB by going Thunderbolt 5. Besides initial speed, you also need to look at SLC cache if you plan on transferring huge files.

If you are doing fairly ordinary consumer-type stuff, you won't really notice the potential performance issues. Where performance suffers is if you need to copy large files.

My setup is an M1 Max Studio with 512 GB SSD and an external Samsung 990 Pro 4 TB in an OWC 1M2 enclosure and I get about 3,100 MBps with that setup. The internal is 5,263 / 6,528 MBps but the Studio has better performance than the base SSDs back in the M1 days.

Thunderbolt 5 enclosures cost quite a bit more than Thunderbolt 4 enclosures but costs usually come down with time and more competition.

I've not looked at the 1 TB internal performance.
 
For an m4pro Mini, I'd want at least a 1tb SSD...
24gb RAM is ok.
But if you can afford it, 48gb would be better...
 
However I keep reading about slow SSD speeds based on what memory chips are used. Does anyone know what capacity I would need to make sure I'm on the next fastest bus speed? Can that be 1TB?
In the past, the primary discussion of difference was the configuration, that is, the number of NAND modules. More modules allows for faster performance because they can work in parallel (e.g., more RAM channels, RAID vs single disk). Although, I’m not seeing any configuration differences in the (base) M4 lineup. Therefore, perhaps, it is more about the NAND manufacturer now.
¯\_(ツ)_/¯


Anyway…. Here are my own numbers:

AmorphousDiskMark

Measurement size: 1 GiB v 64 GiB
Single run each

M4 Pro - 2TB
• SEQ1M-QD8:
— Read: 6849 v 6606 MB/s
— Write: 8038 v 5855 MB/s
• SEQ1M-QD1:
— Read: 3378 v 3873 MB/s
— Write: 7512 v 6752 MB/s
• RND4K-QD64:
— Read: 1080 v 1032 MB/s
— Write: 194 v 210 MB/s
• RND4K-QD1:
— Read: 57 v 65 MB/s
— Write: 34 v 39 MB/s

OWC 1M2 (USB4) - WD_BLACK SN850X - 2TB
• SEQ1M-QD8:
— Read: 3846 v 3837 MB/s
— Write: 3710 v 3704 MB/s
• SEQ1M-QD1:
— Read: 3119 v 2644 MB/s
— Write: 3208 v 2961 MB/s
• RND4K-QD64:
— Read: 712 v 334 MB/s
— Write: 300 v 182 MB/s
• RND4K-QD1:
— Read: 63 v 51 MB/s
— Write: 44 v 37 MB/s

OWC 1M2 (USB4) - Samsung 990 Pro - 4TB
• SEQ1M-QD8:
— Read: 3853 v 3855 MB/s
— Write: 3614 v 3651 MB/s
• SEQ1M-QD1:
— Read: 3293 v 3304 MB/s
— Write: 3083 v 3087 MB/s
• RND4K-QD64:
— Read: 664 v 647 MB/s
— Write: 281 v 281 MB/s
• RND4K-QD1:
— Read: 70 v 70 MB/s
— Write: 52 v 52 MB/s

Now also available is the OWC Express 1M2 80G (USB4v2).
 
Does anyone know what capacity I would need to make sure I'm on the next fastest bus speed? Can that be 1TB?
The number of chips is what matters. The old M2 256GB model had one chip, while the 512GB model used two chips and was twice as fast. All larger models had an even number of chips and were as fast as the 512.

I thought this only applied to the M2

But what the YouTubers fail to understand is that effective speed depends on caching in RAM. speed is (hit ratio)*(ram speed) + (1-hit ratio)*(ssg speed)
Unless you are running a synthetic benchmark that intentionally keeps the hit ratio at zero.
 
Last edited:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.