Thunderbolt 4 supporting M1 Macs don't support 10Gbps USB 3.1 Gen 2 leading to 2x worse USB speeds

[AlTech]

Summary

 

 9to5mac reports that testing on M1 Pro and M1 Max supporting Macs with Thunderbolt 4 reveals that they don't support USB 3.1 Gen 2 10Gbit/s as they are expected to do through Thunderbolt 4.

 

The testing has shown that the USB file transfer performance is half of what is expected from a USB 3.1 Gen 2 containing Thunderbolt 4 port in both theoretical and real world performance.

 

9to5mac has reached out to Apple for comment and so far Apple hasn't responded.

 

Quotes

Quote

Pro users of M1 Macs have reported disappointing transfer speeds with external SSDs, and tests appear to show that most M1 Mac Thunderbolt ports don’t support USB 3.1 Gen 2 – which means they don’t offer the maximum 10Gb/s transfer speeds that would be expected from Thunderbolt 4.

Quote

The piece details the actual speeds achieved, suggesting the following conclusions. Key among them is that the fastest storage devices perform at about half their expected speeds; the limitation appears to be present in all M1 Macs; using a Thunderbolt 4 cable to connect a USB-C device to the front USB-C ports on a Mac Studio Max results in speeds below 10% of that expected; even the latest Mac Studio models don’t support USB 3.1 Gen 2.

 

My thoughts

Honestly this is incredible. With Apple I feel like there's product defect after produxt defect. Do they not test devices before selling them ot what? Or, if this is an intentional feature then why not come clean and just say hey our USB ports suck but at least you have Thunderbolt 4.

 

Sources

 https://9to5mac.com/2022/04/18/m1-mac-thunderbolt-4-ports-speed-tests/

[darknessblade]

Let me guess Apple will soon blame it on the Users, saying they have to use the "Official" ovepriced apple cable to get to that speed. and 3rd party cables will not work at the rated speed, even if they are rated to be compatible with that speed

[Lightwreather]

So source article:

Quote

Currently, and as far as I’m aware since their release in November 2020, no Thunderbolt port in any M1 model appears to fully support 10 Gb/s SuperSpeed+ in USB 3.1 Gen 2, at least for SSDs.

Then direct contradiction:

Quote

The only ports in M1 Macs that currently appear to support USB 3.1 Gen 2 fully are those on the front of the Mac Studio Max.

And apparently the issue lies solely with USB 3.1 Gen2

Quote

10 Gb/s SuperSpeed+ in USB 3.2 devices does appear to be supported by the Thunderbolt ports of M1 Macs.

My feeling: This entire thing feels like a plug for the primary source's app:

Quote

Transfer rates were measured using my free app Stibium, version 1.0 (55), which wrote a total of 160 files of sizes from 2 MB to 2 GB in size to a folder on the SSD being tested, and read those same files back (Series Write and Series Read Tests as detailed in Stibium’s Help reference). The procedure detailed as the ‘Gold Standard’ test was followed.

Don't get me wrong, if this is an actual issue this should indeed be fixed and APple held accountable but I doubt that this is an actual issue.
That said if anyone has the requisite hardware and can test it with an actual file transfer, please do tell here. I'd love to be proven wrong

[Alex Atkin UK]

That's weird, they're supposed to be USB 4 which surely is backwards compatible with USB 3.1 Gen 2?

 

This just proves how badly the USB-IF have screwed up the specification if backwards compatibility is now optional.  Its not like this gives me much confidence.

 

But then calling them Thunderbolt 4 ports is wrong to begin with, they're USB 4 which includes Thunderbolt 4 support, not the other way around.

[leadeater]

49 minutes ago, Alex Atkin UK said:

But then calling them Thunderbolt 4 ports is wrong to begin with, they're USB 4 which includes Thunderbolt 4 support, not the other way around.

I miss when were able to just laugh at USB and then look over at how clean Thunderbolt was. USB-IF getting their hands on it seems to be slowly spreading it's chaos to TB.

[Alex Atkin UK]

18 minutes ago, leadeater said:

I miss when were able to just laugh at USB and then look over at how clean Thunderbolt was. USB-IF getting their hands on it seems to be slowly spreading it's chaos to TB.

The whole concept of standards seems to be going out of the window.  With HDMI, DisplayPort and Thunderbolt all having "optional" components that are not made clear to end users.

I just don't understand it, as people buying one thing then getting something else, surely customer dissatisfaction is going to be bad for manufacturers with higher returns?

[Origami Cactus]

While Apple hate seems to be a popular past time on this forum, this blunder seems to be on the USB standard corporation, not apple.

USB 4.0 doesn't need to support USB 3.1 Gen 2 (dual lane 5gbit to achieve 10gbit), but it needs to instead support USB 3.2 gen 1 (single 10gbit lane).

You can see that dual lane transfer seems to be optional, from the USB 3.2 gen 2 being "optional".

Spoiler

 

The original original article,, which you didn't link, but you linked some news site that copied from there, states that they used "basic UGREEN external SSD enclosure" for testing, which is USB 3.1 Gen 2, so to get 10gbit you need 2 lanes, otherwise falls back to 5gbit. 

 

So USB 3 1x2 (10GBit) runs at 5Gbit speed, but USB 3.2 Gen 1 would work fine at 10gbit.

Spoiler

 

 

I would say the article is pretty pointless, they should have also tested USB 3.2 enclosures, not only dual lane 3.1.

 

"Although USB4 is required to support dual-lane modes, it uses single-lane operations during initialization of a dual-lane link; single-lane link can also be used as a fallback mode in case of a lane bonding error."

 

And because the original article just used a cheap external SSD enclosure, instead of an actual external SSD like Samsung T5, this fallback mode sounds plausible, due to some error in Ugreen external SSD enclosure. And I still don't understand If USB 3.1 Gen 2 needs to be supported by USB 4 at all, seems like only 3.2 10gbit is "required"?  The wikipedia article doesn't make it any clearer, with a confusing stance on dual-lane. One sentance says that it is optional, other says it is required. So either way more testing is needed.

 

So we can't draw "apple bad" conclusions from that half test, if you can even call that, more that USB specifications are really trash.

 

[wanderingfool2]

2 hours ago, J-from-Nucleon said:

So source article:

Quote

Currently, and as far as I’m aware since their release in November 2020, no Thunderbolt port in any M1 model appears to fully support 10 Gb/s SuperSpeed+ in USB 3.1 Gen 2, at least for SSDs.

Then direct contradiction:

Quote

The only ports in M1 Macs that currently appear to support USB 3.1 Gen 2 fully are those on the front of the Mac Studio Max.

And apparently the issue lies solely with USB 3.1 Gen2

Makes you really appreciate how stupid the USB standard has become in terms of all the terms and differences between them.  Honestly, they really should have stuck with an incremental approach.

 

On a side note though, it's not contradictory.  The second statement is wrong, but not contradictory.  In their testing, whether it's flawed or not is up for debate, they found none of them worked with 3.1G2 at 10 Gb/s.  From what they say the only ones that appear to support it are Studio Max's front ports.  Which I mean isn't really a contradiction.  Don't get me wrong, the second statement is still wrong but it's not a contradiction.  Apple's own tech spec document on the Studio Max is that it supports 3.1G2 up to 10 Gb/s on the back ports. Which I think makes it much worse that they tested the front ports

 

Can we let that sink in for a second though, they tested the front ports which is rated for Thunderbolt 4, and didn't test the back ports which are advertised as 3.1G2 at 10 Gb/s according to this particular article.

 

With that said, I find it a bit troubling if they have 3.1G2 on the back ports but not the front ports (because the front ports are the ones more likely to have a high speed device plugged into it).  On another note though, the other article posted says they tested it on all USB ports

 

1 hour ago, Origami Cactus said:

While Apple hate seems to be a popular past time on this forum, this blunder seems to be on the USB standard corporation, not apple.

USB 4.0 doesn't need to support USB 3.1 Gen 2 (dual lane 5gbit to achieve 10gbit), but it needs to instead support USB 3.2 gen 1 (single 10gbit lane).

You can see that dual lane transfer seems to be optional, from the USB 3.2 gen 2 being "optional".

  Reveal hidden contents

 

The original original article,, which you didn't link, but you linked some news site that copied from there, states that they used "basic UGREEN external SSD enclosure" for testing, which is USB 3.1 Gen 2, so to get 10gbit you need 2 lanes, otherwise falls back to 5gbit. 

 

So USB 3 1x2 (10GBit) runs at 5Gbit speed, but USB 3.2 Gen 1 would work fine at 10gbit.

The thing is though, Apple does actually advertise the back ports as supporting 3.1G2 up to 10 Gb/s.  So no matter what occurs as long as the device they were testing with hits the 10Gb/s on other machines then the same should occur on a Studio Max, which it doesn't.

 

The article you posted tested it on an Intel Mac as well, where they said they achieved the rated speeds.  So effectively this is testing their 3.1G2 implementation, because in the article they already mentioned that 3.2 is working correctly (10Gb/s).

[Senzelian]

3 hours ago, Origami Cactus said:

And I still don't understand If USB 3.1 Gen 2 needs to be supported by USB 4 at all, seems like only 3.2 10gbit is "required"? 

3.2 Gen 2 is the same as 3.1 Gen 2. It got renamed with the introduction of 3.2 Gen2x2. 

 

Welcome to the bullsh*t that is USB naming.

[Origami Cactus]

5 minutes ago, Senzelian said:

3.2 Gen 2 is the same as 3.1 Gen 2. It got renamed with the introduction of 3.2 Gen2x2. 

 

Welcome to the bullsh*t that is USB naming.

Just gets worse the longer you look at it, especially with the dual/single channel BS also piling on top of the already steaming pile.

[Kronoton]

3 hours ago, wanderingfool2 said:

Can we let that sink in for a second though, they tested the front ports which is rated for Thunderbolt 4, and didn't test the back ports which are advertised as 3.1G2 at 10 Gb/s according to this particular article.

 

You got something confused here.

 

On a Studio with M1Max the front ports are not Thunderbolt only USB3.something. With the Ultra all 6 ports are Thunderbolt.

 

So it seems (to me) that what ever Apple does with those USB-C that don't have a dedicated TB controller also includes that 2 lane 10GBit mode while it is not available at those ports that do have their own TB controller.

 

Might be interesting test that theory on an iMac as AFAIR only 2 out the 4 ports (on the "bigger" configs) are TB and the other 2 just USB3.

[Paul Thexton]

<looks at USB info to try and make head or tails of it>

<gives up, makes a cup of tea, and remains happy that my USB-C connector drives work as I expect them to with no obvious (unbenchmarked) speed problems that I notice while using them>

[soldier_ph]

Apple be like: You're Mac isn't Premium enough to be blessed by the full Thunderbolt 4 spec. Please buy a Higher End Mac to get the full Exclusive Premium Experience.

[hishnash]

On my M1 Max MBP system information reports having a USB4 controler and a USB 3.1 controler but no 3.2 controllers. 

However when I attach a USB 3.2 Gen.2 device it does report 10 Gb/s max throughput and I do get close to that in speeds. Defiantly more than 5 Gb/s (this is a Samsun T7 ssd so it is limited by the ssd). I wander if the USB controller/driver is very particular about the handshake and if its not just in the way apple expected it falls back to a slower protocole.   I did notice when I attempted to attach my laptop to my large Philips screen (that I use as a tv) it did not establish the USB 3.2 Gen 2x2 connection.. 

People who have TB/USB-4 docs have reported that they do not have any speed issues since in those cases the handshake is handled by the controller on the dock..  its very apple to implement the USB spec to the letter and maybe provide absolutely no tolerance so if there is even the siltiest bit of noice or timing shift on the signal it will fallback to 3.1 (this SSD I have falls back to 3.1 if I swap the cable from the very short 30cm to a longer 1m cable even through the 1m cable is rated for 20GB/s)...
 

[GhostRoadieBL]

so long thread short:

- the thunderbolt ports run at 10Gb/s with the correct hardware and when the handshake happens

- USB naming conventions are absolute garbage

- if you want USB 3.2 gen2 (or USB 3.1 gen2 or TB or USB 3 10G, USB4 etc) you have to use the correct hardware within the spec

 

I don't think I learned anything from this news story other than already knowing bad press is rampant in the misinformation era

[hishnash]

10 minutes ago, GhostRoadieBL said:

- if you want USB 3.2 gen2 (or USB 3.1 gen2 or TB or USB 3 10G, USB4 etc) you have to use the correct hardware within the spec

this one line just emphasises how broken USB is.. why did they not just call it USB-C 5Gb, USB-C 10Gb USB-C 20Gb... etc

Im currently looking for a 2m+ USB-C to C 20 Gb/s cable and really don't want to buy the wrong thing (as this length is costly) but I just cant trust product descriptions to be correct.. I have a feeling that a TB3 cable will not work with the device im trying to connect to. 

[GhostRoadieBL]

Just now, hishnash said:

this one line just emphasises how broken USB is.. why did they not just call it USB-C 5Gb, USB-C 10Gb USB-C 20Gb... etc

Im currently looking for a 2m+ USB-C to C 20 Gb/s cable and really don't want to buy the wrong thing (as this length is costly) but I just cant trust product descriptions to be correct.. I have a feeling that a TB3 cable will not work with the device im trying to connect to. 

I understand the idea when USB 3, 3.1, 3.2 and dividing into lanes per spec became a mess when 3.1 originally had 1 lane and was 5Gbit but 3.2 had 1 lane and double the bandwidth that made sense. I think when they switched both standards to gen2 with 2 lanes each it became too messy to fix and at the time (I think it's still the case) "thunderbolt" was an intel property and needed to be licensed with fees and all the patent garbage that comes from, horribly toxic for the industry, giants bring with them.

 

There was nothing wrong with USB3 being 5gig and then USB4 being 10gig, then let intel play in their walled off sandbox with 40gig which USB5 can jump to and be praised for skipping 20 and 30gig to a new speed standard which would completely undercut intel's patent making it worthless to anyone but their company.

[wanderingfool2]

On 4/21/2022 at 12:42 PM, Kronoton said:

You got something confused here.

 

On a Studio with M1Max the front ports are not Thunderbolt only USB3.something. With the Ultra all 6 ports are Thunderbolt.

 

So it seems (to me) that what ever Apple does with those USB-C that don't have a dedicated TB controller also includes that 2 lane 10GBit mode while it is not available at those ports that do have their own TB controller.

 

Might be interesting test that theory on an iMac as AFAIR only 2 out the 4 ports (on the "bigger" configs) are TB and the other 2 just USB3.

I read the ultra stats, which mentioned TB instead of USB

 

[tikker]

10 hours ago, hishnash said:

this one line just emphasises how broken USB is.. why did they not just call it USB-C 5Gb, USB-C 10Gb USB-C 20Gb... etc

They did:

Quote

https://www.usb.org/sites/default/files/usb_3_2_language_product_and_packaging_guidelines_final.pdf
The USB 3.2 specification absorbed all prior 3.x specifications. USB 3.2 identifies three
transfer rates, USB 3.2 Gen 1 at 5Gbps, USB 3.2 Gen 2 at 10Gbps and USB 3.2 Gen 2x2 at
20Gbps. It is important that vendors clearly communicate the performance signaling that a
product delivers in the product’s packaging, advertising content, and any other marketing
materials.
• USB 3.2 Gen 1
o Product capability: product signals at 5Gbps
o Marketing name: SuperSpeed USB
• USB 3.2 Gen 2
o Product capability: product signals at 10Gbps
o Marketing name: SuperSpeed USB 10Gbps
• USB 3.2 Gen 2x2
o Product capability: product signals at 20Gbps
o Marketing name: SuperSpeed USB 20Gbps

The X.Y Gen Z(xW) notation is internal naming for the specification to distinguish the technicalities of the various transfer modes. All a consumer should have to worry about is the speed they can achieve. The connector has less to do with the USB version directly and has its own specification.

 

I think the article's source falls short in that it provides no information as to which specific enclosures and cables were used for the tests. Given that they achive speeds in excess of 450 MB/s (as noted in the USB 3.2 spec), it seems the link is operating at speeds higher than 5 Gbps.

Quote

For links that operate at Gen 1x1 speed (5 Gbps and 8b/10b line encoding) the raw
throughput is 500 MB/sec. Accounting for flow control, packet framing and protocol
overheads reduces the effective bandwidth down to 450 MB/sec or less to be delivered to an
application. The effective bandwidth is doubled in Gen 1x2 operation.

and the fact that they do achieve nearly full speed in other circumstances

Quote

However, a Sabrent+Samsung NVMe conforming to USB 3.2 was found to connect at 10 Gb/s, and returned a read rate of 911 MB/s and write of 973 MB/s.

indicates to me that this is more subtle than a blanket "they don't support it" (because USB 3.2 Gen 2x1 is what used to be the USB 3.1 Gen 2 transfer mode).

[hishnash]

13 hours ago, tikker said:

indicates to me that this is more subtle than a blanket "they don't support it" (because USB 3.2 Gen 2x1 is what used to be the USB 3.1 Gen 2 transfer mode).

All of these cheap USB implementation will have been developed a few years ago and tested on the avoidable laptops/desktops on the market at the time, they will have make 1000s of iterations to save money (cable, chipset, power circuit etc) many of these bring the devices well outside spec but they have been designed to still work for existing host controllers at the time of design. 

Then when you bring a new host chipset to the market how the designers of this integrated the spec and how they weighted the different signal quality parameters to determine if the connection is good will de different so existing cost cutting designs that are intentionally running very close to the edge will inevitably fail.  

What this really shows is that there are way to few companies developing USB host controllers/controller firmware so the usb-client controller industry has been able to be so bad.. just look at the cable testing done by Linus most of what you buy is utter garbage.  The nature of USB is that if your a vendor of usb client chipsets etc it is a price war to produce the cheapest product that 'works' on enough of the current host devices. 

There is a reason TB client devices (like docks etc) all cost a LOT more as there is not open market to product the controller chipsets etc. 
 

[RedRound2]

Is it only with USB 3.1 gen 2 speeds? If port itself supports full 40 Gb/s speeds, I dont see why it shouldn't support 10 Gb/s speeds and the issue is more likely a software or handshake problem.

[AlTech]

19 minutes ago, RedRound2 said:

Is it only with USB 3.1 gen 2 speeds? If port itself supports full 40 Gb/s speeds, I dont see why it shouldn't support 10 Gb/s speeds and the issue is more likely a software or handshake problem.

Seems to be. TB4 works fine on those devices.

[hishnash]

21 hours ago, RedRound2 said:

Is it only with USB 3.1 gen 2 speeds? If port itself supports full 40 Gb/s speeds, I dont see why it shouldn't support 10 Gb/s speeds and the issue is more likely a software or handshake problem.

I expect its an issue with the quality of the controller/connection to the controller on the device your trying to use. I have SSDs that work perfectly at 10Gb/s and others that don't even through they both are labeled with 3.1 gen 2.  Just look at Linus' video on cable quality there are LOT of cables labeled as supported a given spec that don't quite hit that target! I expect this is also true of usb controllers within devices you plug in.

Its not that apple have set a more strick set of rules but that the thresholds they are using are different. Current devices on the market will be using USB controllers and cables that have been optimised to be as cheep as possible and still pas the quality checks needed by AMD's and Intels's USB host controllers (this does not mean they are within spec at all).  The USB controler/cable market is all about being as cheap as possible to hit a given perf spec and this is not about complying with the on paper spec list but rather does it work with the give host devices on the market! 

When something is labeled as 3.1 gen2 that does not mean the USB group have tested it compiles with the 3.1 gen2 spec it just means that the person selling it expect most host controllers to be tricked into accepting it at 3.1 gen 2 even if the quality of the connection is in fact a long way from perfect. 

 

<removed by staff>

[WkdPaul]

* thread cleaned *

 

Please stay respectful.