Is this the general consensus??
M1 Max and M1 Pro chips should not exhibit such behaviour?
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
USB on M1 Macs isn't actually 10Gb/s? (Also definitely not USB4)
- Thread starter Anonymous Freak
- Start date
-
- Tags
- apple silicon mac mini usb 3.2
- Sort by reaction score
Is this the general consensus??
M1 Max and M1 Pro chips should not exhibit such behaviour?
I think my 14" Pro is little faster than my M1 Mac Mini but not by a whole lot. I think it depends more on the controller used by the ext SSD's. Both lag behind my 2018 Intel Mac Mini I used to have. See my post above you.
I remember reading that in the acasis website when I was about to buy it, but I cannot find it now. But also from looking around in posts/testimonials it does seem so.
ASM235CM chip (for USB-C to SATA enclosures) does not work at 10 Gb/s using the built-in USB-C ports of the MacBook Pro 14". System Information states the device is "Limited to 5 Gb/s."
It is only able to connect at 10 Gb/s when connected to the 10 Gb/s USB-C port of a Thunderbolt 3 dock.
It is only able to connect at 10 Gb/s when connected to the 10 Gb/s USB-C port of a Thunderbolt 3 dock.
Last edited:
non-Thunderbolt USB-C ports of Mac Studio are limited to < 8 Gbps because the ASMedia ASM3142 only has one PCIe gen 3 lane.
https://forums.macrumors.com/threads/issues-with-usb-ports-on-m1-max.2355896/post-31367729
What about the non-Thunderbolt USB-C ports of the M1 iMac? That also uses a ASMedia ASM3142 chip but it might have two lanes? Need ioreg dump to check.
The name "pcie-xhci" might be different on the M1 iMac. In that case, check the list of names with
https://forums.macrumors.com/threads/issues-with-usb-ports-on-m1-max.2355896/post-31367729
What about the non-Thunderbolt USB-C ports of the M1 iMac? That also uses a ASMedia ASM3142 chip but it might have two lanes? Need ioreg dump to check.
ioreg -filrtw0 -n pcie-xhci > ioreg_MacStudioAsm3142.txtThe name "pcie-xhci" might be different on the M1 iMac. In that case, check the list of names with
ioreg to find the correct one.Sorry in advance for not completely reading through all the pages of this post.
I also got a adata SX8200 pro 2TB new replacement with the slower controller and got about 1200 W/R speeds, similar to yours using blackmagic. Tried several TB3 enclosures and very minor differences. Put it in a Windows laptop, and got 1700 read, 1300 write using crystaldiskmark,. This is about half the speed expected. Did you ever resolve the issue? I ended up RMA this to Adata.
I have a hypothesis for your results based on extensive testing: Certain controllers perform worse with M1 macs. For example, the JM based controllers are much slower with the TB4 ports. The USBc front ports are based on ASM. Could xplains #1 result if the sandisk controller was an older model. The Studio front USB ports have one PCIe lane, which limits the bandwidth. The sandisk pro could have a more compatible controller which explains the faster speed with the TB4 port.
Never got a resolution to this. I ended up just putting that SX8200 into a laptop (which I actually already sold).
If I have a Late 2014 iMac 5K with a USB 3.0 Bus listed in System Information, what speed might I get from this here drive would you know?
Amazon linked page says 440MB/second, less than USB 3.0 bus speed, so 440 MB/sec
I doubt it. This has been a recurrent issue for ever with macs and usb (and other) controllers. It is absurd because we are talking about huge companies and products that are often marketed for macs, but well.
In general, it seems that thunderbolt works much more reliably than pure usb hubs/enclosures etc. Just do a favour to yourself and avoiding wasting money on usb3 enclosures/hubs etc, and get thunderbolt ones. Otherwise you will never know when things will just for no apparent reason stop working, or when you will start having whatever weird issue.
Take a look in the madness of usb3 hubs, for instance:
USB-C hubs and my slow descent into madness - Dennis Schubert
A short story of me trying out different USB-C hubs, taking them apart, and slowly descending into madness.
My experience is quite different. With my previous Macs with TB3 (MBP 2017 with Touchbar and a few more) I had multiple TB3 docks, from Belkin ("TB3 Express"), Elgato ("TB3") and another I can't recall. On all Macs all docks had identical behaviour: RJ45 Ethernet 50:50 chance of being permanently lost after unplugging the dock and plugging it in again until a reboot of the Mac (the Mac would claim no cable connected), 4k display turning off and on again for a moment and so on.
I ended up returning all docks over the course of a couple weeks (after it became clear that there's no support from the manufacturers), which at the time (2017) were relatively expensive so it was almost like a grand just in docks and TB3 cables (as the included cables were short and good quality Startech cables cost a pretty penny). At work I tried another TB3 dock which worked reasonably well until some MacOS update at which point the Mac no longer charged through it and the dock was rendered useless.
Ended up selling that 2017 Macbook as well once the keyboard broke and the butterfly issues came to light. After that I converted to more of a desktop setup where I didn't need a dock to begin with and skipped all Macbook generations until the 2020 Intel that I got with 4 ports so I wouldn't need a dock there either.
For that Macbook I bought a simple very cheap USB-C hub that has four USB-C ports with one of them reserved for the charger. Each port is limited to 5Gbps and 10W for charging, but the thing cost 25 bucks, works on my 2020 Intel MBP, on my 2020 M1 MBP, and on my 2021 M1 Max MBP flawlessly with all MacOS versions.
I'd rather buy a bunch more of these cheap hubs rather than a single expensive Thunderbolt dock that could give out at any moment with no real manufacturer customer support. I also disliked the big 100W external power bricks for these docks, I understand why they are required but they were quite visible and ugly with no good way to hide them as they were too big to fit in my cable management desk compartment.
My trusty hub says:
Code:
USB3.1 Hub :
Product ID: 0x0822
Vendor ID: 0x2109 (VIA Labs, Inc.)
Version: 90.13
Serial Number: 000000001
Speed: Up to 10 Gb/sI don't know what ****** bridge chip might be in there, all I know is that this hub never once dropped a device connection in the last 2 years and I connect a lot through it, Pixel phone, iPad, iPhone, Watch charger, USB-C-to-A adapters with lots more USB-A peripherals, Focusrite USB-C interfaces, USB microphones, and many more.
Granted though, I don't use any external SSDs where I need the transfer speeds since I store all data over the network.
Today I set up my new (to me) M1 Mac mini with Plugable Thunderbolt 4 / USB4 5-in-1 hub (USB4-HUB3A) and Samsung T7 Shield SSD. I tried various connections, USB-A and USB-C, direct to Mac and through hub.
M1 Mac mini USB-A -USB 3.1 Gen 2 (10 Gbps) (5 Gbps)
TB4/USB4 hub USB-A - USB 3.2 Gen 2 (10 Gbps)
For USB-C, I get the fastest speeds directly connected to the Mac, but speed through the hub's USB-C is reasonably close.
For USB-A, the hub is twice as fast as the Mac, and pretty close to USB-C speeds.
Mac USB-A - 375 MB/s write, 300 MB/s read
Mac USB-C - 815 MB/s write, 580 MB/s read
Hub USB-A - 695 MB/s write, 455 MB/s read
Hub USB-C - 710 MB/s write, 510 MB/s read
M1 Mac mini USB-A -
TB4/USB4 hub USB-A - USB 3.2 Gen 2 (10 Gbps)
For USB-C, I get the fastest speeds directly connected to the Mac, but speed through the hub's USB-C is reasonably close.
For USB-A, the hub is twice as fast as the Mac, and pretty close to USB-C speeds.
Mac USB-A - 375 MB/s write, 300 MB/s read
Mac USB-C - 815 MB/s write, 580 MB/s read
Hub USB-A - 695 MB/s write, 455 MB/s read
Hub USB-C - 710 MB/s write, 510 MB/s read
Last edited:
Which benchmark?
Tech specs says M1 Mac mini has 5 Gbps USB-A ports.
Check the USB hierarchy. It seems the hub's USB-A ports might from a child hub of the hub that controls the USB-C ports.
I think you'll get better results from a Thunderbolt 3 dock since it won't use USB tunnelling like Thunderbolt 4 docks use.
Need a test from an Intel Mac to see the max of the T7 to show how much is lost with the M1.
Black Magic.
You are correct, 5 Gbps for the Mac mini’s USB-A. I misread the specs page (screenshot below). That alone may explain it, since the USB-C speed is double the Mac mini’s USB-A speed. BTW, the Mac Studio offers no paper spec advantage here. It also lists 5 Gbps for its USB-A ports.
T7 Shield is USB 3.2 Gen 2 (10 Gbps).
I have no complaints about the hub’s USB-A speed though, as it’s reasonably close to the hub’s USB-C speed.
I’ll do some Intel Mac testing later.
Attachments
Last edited:
I would use AmorphousDiskMark.app since it usually gives higher numbers and therefore shows more closely what the bandwidth limit is.
USB 5 Gbps is actually 4 Gbps since it uses 8b/10b encoding.
USB 10 Gbps is actually 9.7 Gbps since it uses 128b/132b encoding so it's a little more than twice as fast which is what your read score shows.
Samsung T7 Shield 2 TB tests - This drive is 10 Gbps, and rated for 1050 MB/s read. APFS formatted.
M1 Mac mini + Plugable TB4/USB4 Hub - USB-C:
M1 Mac mini + Plugable TB4/USB4 Hub - USB-A:
M1 Mac mini direct USB-C (Plugable hub not connected):
M1 Mac mini direct USB-A (Plugable hub not connected):
M1 Mac mini direct USB-A (Plugable hub connected to Mac mini USB-C, but Samsung drive not attached to hub):
M1 Mac mini 1 TB (Ventura) internal SSD:
Both USB-C 10 Gbps and USB-A 5 Gbps on my 5 year-old Kaby Lake iMac are much faster than on my M1 Mac mini.
Core i5-7600 iMac (Monterey) direct USB-C:
Core i5-7600 iMac (Monterey) direct USB-A:
Last edited:
Were these drives encrypted? It makes a big difference in the sequential performance. For instance, here's a 2 TB Kingston XS2000 plugged into the TB3 port on my 2019 iMac, before encryption:
And after encryption:
Also, I believe there's something wrong with the measurement you made of the 1 TB internal SSD on your M1 Mini. 33.20 MB/s is much too slow for 4k random writes with queue depth = 1 (figure at lower R). Here are measurements made on my 2019iMac (Monterey 12.4). Without encryption, I got ~420 MB/s, and with encryption I got ~275 MB/s.
Last edited: