Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Doesn't USB4 (which Apple does not write correctly, despite the horrible ongoing disjointed naming saga of USB-IF) require, what the heck do they call it, Gen 3.2 2x2 speeds at 20Gbps? Oh, is that the optional "tunneled" mode? What is "native" USB4 20Gbps? Is it really just 20Gbps TB3? This stuff is beyond ridiculous. I've been deep in the IT and tech world for over a quarter century, and I'm confused by this BS.
 
I think the bigger issue (at least for me!) is that it only has two of them! The Intel MacBook Pro’s had four.

The low end 13” Intel MBP only has 2 ports too. Perhaps they’ll release that long rumored 14” MBP with 4 ports when the 16” comes out
 
Doesn't USB4 (which Apple does not write correctly, despite the horrible ongoing disjointed naming saga of USB-IF) require, what the heck do they call it, Gen 3.2 2x2 speeds at 20Gbps? Oh, is that the optional "tunneled" mode? What is "native" USB4 20Gbps? Is it really just 20Gbps TB3? This stuff is beyond ridiculous. I've been deep in the IT and tech world for over a quarter century, and I'm confused by this BS.
I don't think USB4 requires a USB 3.2 gen 2 x 2 mode (20 Gbps dual lane USB, 10 Gbps per lane). Have to check the USB4 spec (which you can download).

Thunderbolt 3 has a 20 Gbps mode which is actually 20.625 Gbps (dual 10.3125 Gbps) for tunnelled DisplayPort, PCIe
Thunderbolt 3 has a 40 Gbps mode which is actually 41.25 Gbps (dual 20.625 Gbps) for tunnelled DisplayPort, PCIe
USB4 has a 20 Gbps mode (slightly slower than Thunderbolt) (20 Gbps using dual 10 Gbps) for tunnelled DisplayPort, PCIe, USB 3.x
USB4 has a 40 Gbps mode (slightly slower than Thunderbolt) (40 Gbps using dual 20 Gbps) for tunnelled DisplayPort, PCIe, USB 3.x
I suppose if you have a Thunderbolt 3 device between a USB4 host and a USB4 device, it will allow tunnelled USB 3.x to pass (the Thunderbolt 3 controller just doesn't have the USB Adapter to output USB 3.x from those packets).

Whether the ports are Thunderbolt 3 or Thunderbolt 4 doesn't matter. What does matter, is what can you connect to it?

How many displays per Thunderbolt port? How many displays total? It is unclear from the specs whether the answer is one or two. DisplayPort 1.4? DSC?

Does it support Thunderbolt 4 hubs like the OWC Thunderbolt hub? The hub has 3 downstream ports. USB4 hubs can also have multiple downstream ports.
 
  • Like
Reactions: frisbfreek
The reduced number of ports is a bigger bummer for me than the absence of TB4. The Mini was always its own hub, with only 2 Thunderbolt ports I’d need an external one.

Furthermore they should flat out be ashamed to still have 8GB RAM / 256GB storage as the baseline. Jesus, those were the default configs when I bought a 2014 MBP 15” six years ago; either the then-outdated and useless 8/256 or the reasonable 16/512. Computers with 8GB RAM / 256GB SSD are from a time when entry-level iPhones/iPads came with 16 GB storage. That number has quadrupled since. For the Pro phones, 128 GB is now the floor.

Yet when they introduce their new Apple Silicon computers to the world, they do it with configs that belong in 2010. Even on the MacBook Pro. Shocking.

So... pass. Get back to me when you have non-insulting default configs to offer. Bye.
 
Last edited:


The new M1-equipped MacBook Air, MacBook Pro, and Mac mini are each equipped with two USB-C ports that support USB 4 and Thunderbolt, but as it turns out, Apple is continuing to use Thunderbolt 3 rather than upgrading the new models to Thunderbolt 4.

macbook-pro-thunderbolt-3-ports.jpg

Intel in July shared details on Thunderbolt 4, which is coming out in new PCs with Tiger Lake processors. Thunderbolt 4 offers the same 40Gb/s maximum speeds available through Thunderbolt 3, but it does bring some notable improvements, such as support for docks with four downstream Thunderbolt ports. From Intel:
  • Double the minimum video and data requirements of Thunderbolt 3.
  • Video: Support for two 4K displays or one 8K display.


The M1 Mac "fail" Thunderbolt 4 compliance here. The two laptops can only manage output just one video stream. It appears that the M1 can only do two total. ( which is actually a retreat backward from Intel iGPU limit of 3 (and which Intel improved on with the current Xe-LP graphics to a limit of 4 ... so surprise , surprise their TB4 Gen11 CPU(SoC) powered laptops 'pass'. ). In this respect the M1 looks like it is basically just an iPad Pro SoC with a different badge slapped on it.


Technically I suppose the Mini could qualify since take away the build in display but it is probably better to be consistent with the Laptops not passing. Otherwise Apple would illuminate the display count backsliding they are doing here even more.

Thunderbolt 4 requirement makes for a more uniform experience. See a TBv4 port. If plug in a monitor it will work. No guessing on dependencies about what you have plugged in on other ports. Keeping track of monitors in use . Just plug ... works.

Very similar with data bandwidth consistency. Is this TB controller only feed x2 or x4 . Left side port has different data rate than right side port. If TB4 then they are same.

Intel putting their TBv4 controllers in the CPU makes it "fall out of bed in the morning" easy to hit compliance. Other folks are just going to do "USB 4 implementation with optional TBv3 included ".
 
Rushed? Not really. When Apple first moved from PowerPC to Intel, the form factors were identical, just the guts were changed. That we're seeing the same with the first models moving from Intel to ASi is not surprising to me.

And on the 68K -> PPC or PPC -> Intel moves did Mac loose I/O ports like the Mini did this time ?

The laptops fit.... but the Mini smacks of being pushed out the door with the only SoC option possible.
 
It looks more like the M1 is a slightly modified iPad A14X with the addition of two TB3 controllers and an extra bit in the memory controller to support a maximum of 16GB of memory. May likely be because when the official decision to go Apple Silicon was made, the A14X for the iPad may have already been taped out and finalised.

I expect the M2 to be different next year. It may have SMT-cores (disabled for the A15X variants), 128-256 GB memory addressing, and 4 TB4 ports. This is but the most logical choices for the bigger MBPs and Desktop Macs if Apple needs to to justify Apple Silicon’s legitimacy. The M2 may be the arch that Apple Silicon have designed for the desktop from the ground up.

Having said that, by just looking at the M1’s features, it’s very likely to expect the 2021 iPad Pro to feature a TB3-capable port and 8GB of memory (since Samsung, Kioxia and Hynix have already began shipping the next-gen, higher capacity 16Gbit mobile memory modules late last year).
 
I don't think USB4 requires a USB 3.2 gen 2 x 2 mode (20 Gbps dual lane USB, 10 Gbps per lane). Have to check the USB4 spec (which you can download).

Thunderbolt 3 has a 20 Gbps mode which is actually 20.625 Gbps (dual 10.3125 Gbps) for tunnelled DisplayPort, PCIe
Thunderbolt 3 has a 40 Gbps mode which is actually 41.25 Gbps (dual 20.625 Gbps) for tunnelled DisplayPort, PCIe
USB4 has a 20 Gbps mode (slightly slower than Thunderbolt) (20 Gbps using dual 10 Gbps) for tunnelled DisplayPort, PCIe, USB 3.x
USB4 has a 40 Gbps mode (slightly slower than Thunderbolt) (40 Gbps using dual 20 Gbps) for tunnelled DisplayPort, PCIe, USB 3.x
I suppose if you have a Thunderbolt 3 device between a USB4 host and a USB4 device, it will allow tunnelled USB 3.x to pass (the Thunderbolt 3 controller just doesn't have the USB Adapter to output USB 3.x from those packets).

Whether the ports are Thunderbolt 3 or Thunderbolt 4 doesn't matter. What does matter, is what can you connect to it?

How many displays per Thunderbolt port? How many displays total? It is unclear from the specs whether the answer is one or two. DisplayPort 1.4? DSC?

Does it support Thunderbolt 4 hubs like the OWC Thunderbolt hub? The hub has 3 downstream ports. USB4 hubs can also have multiple downstream ports.

So I downloaded the USB4 spec. I had to dig around a bit, but I found:
"A USB4 host supports 20G USB4 operation (Gen2x2) and optionally 40G USB4 operation (Gen3x2)."

So it seems like Apple is giving false advertising on USB4 "support". Either that or the M1 actually does support Gen2x2 and the specs are wrong... but I'm guessing it's the former.
 
Can anyone justify a customer-focused reason why Apple got rid of 2 of the Thunderbolt ports on the new macbook pros? I was just about to trade mine in until I realized.
 
And on the 68K -> PPC or PPC -> Intel moves did Mac loose I/O ports like the Mini did this time ?

The laptops fit.... but the Mini smacks of being pushed out the door with the only SoC option possible.

Since the M1 is the only SoC option available, that is not surprising.

I have not been able to watch the keynote, but from what little I have seen, the M1 is an A14X with additional hardware specific to the Mac / USB4 / TB3. It's clear it only has a single USB/TB controller so not surprising it is limited to two ports at the moment. I also think the 16GB limit might be either the memory controller or just the available space they had on the die.

The Intel MacBook Air only has two TB ports and the old non-Touchbar MacBook Pro 13 also only has two ports. So the M1 is meeting that and it's $100 cheaper than the base Intel model so you could argue that is to reflect losing those two ports.

As for the M1 Mac Mini, I see this as the model for software developers (who don't want to use an M-series laptop) and for folks who use their Mac Mini as a home media server or such. So the RAM and I/O restrictions are not crippling and you pay less.

The next model of ASi SoC (M1X / M2) will likely offer more cores (CPU and GPU), a higher unified memory limit (32GB+) and will have additional controllers for more USB/TB ports. That is the SoC I expect to go into the 14" and 16" MacBook Pros and they will have four ports. And I could see another Mac Mini with it and more ports sold either as a specific configuration or a new model (Mac Mini Pro).
 
Did anyone else notice how empty the Mac mini case shown in the video was ? Why use the exact same form factor, unless there are plans to add more hardware in the future ?

Couple of reasons:
1) Current mini form factor is exactly 1U tall, and two side by side fit perfectly in a standard network cabinet rack.
2) There's still several million dollars cost involved in designing a new chassis. May seem like peanuts in the scope of Apple's revenue, but they're still a public company with margins to maintain.

If it ain't broke....
 
Since the M1 is the only SoC option available, that is not surprising.

I have not been able to watch the keynote, but from what little I have seen, the M1 is an A14X with additional hardware specific to the Mac / USB4 / TB3. It's clear it only has a single USB/TB controller so not surprising it is limited to two ports at the moment. I also think the 16GB limit might be either the memory controller or just the available space they had on the die.

Since the A14X hasn't been shown the position that this is "extra Mac only" hardware really isn't grounded.

The current iPad Pro comes with a USB-C socket. It connects to a keyboard that also provides a separate USB-C data line feed from the socket on iPad Pro. So two.

If Apple drops a "A14X" into a new iPad Pro then it will be USB something. It seriously would not be hard for that to be USB 4. The A12Z in the current iPad Pro is over two years old at this point ( basically it is a A12X).

If Apple doesn't want Thunderbolt drivers mixed in with iPad OS then they can nominally just ignore the Thunderbolt part .... and it would still technically qualify as officially USB 4. If Apple did their own custom TB controller so they could lock it down hard enough for iPad OS usage then IPad Pro's with Thunderbolt would be a decent feature differentiator from the mainstream iPad ( which is getting FaceID , etc ... that contributed differentiator. )

If the USB 3.2 2x2 was some annoying power draw ... again it is their own TB+USB 3.x controller. So they can just flip off the the higher power mode on iPad Pro. At that point they'd loose the USB 4 badge but still could do it with same controller. (but again does Apple really want to walk the iPad Pro back to the iPad Air. ) .

They could also just provision the iPad Pro with just less memory. 6GB instead of 8GB. Don't need a new memory controller for that .


An article over at Anandtech did some back of the envelope on the size of the M1 ....

"...If Apple was able to keep the transistor density between the two chips similar, we should expect a die size of around 120mm². This would be considerably smaller than past generation of Intel chips inside of Apple's MacBooks.... "

The A12X size ??

" ...This is approximately 45% more than the 6.9 billion of the A12. In terms of silicon the die, it will likely be around 120 mm2, based on the A12’s 83 mm2 and the same 7 nm process. The “X” variant is, once again, sporting a good sized piece of silicon. ..."

the 16GB limit probably more to do with limiting to just two DRAM packages and the max memory density can get out of those (at price points Apple wants to pay). The memory controller is probably tunes to that, but that "two package" limit is probably the constraint. Two chip limit... same as iPad Pro.

From 11" teardown (step 8)

WgkFM4qTwDa42kDf.medium


Yellow box above. Two DRAM packages.

Walks like a duck , quacks like a duck .... probably a duck.

If the 14X dramatically shrinks in size then the iPad Pro is on a different development track. Time will tell.
But the highly established track record of the Apple Slicon solutions is for them to be reused in multiple products.
The 14X 'mode' would probably have a different power threshold curve, but a whole different chip optimization? Probably not. If there was a silicon gap then they might shrink the L2 caches back to the A14 size levels, but I doubt that is really "buying" a whole lot that isn't offset by economies of scale of doing more uniform larger volume batches. The winked out only 7 GPU cores M1s in the entry MBA . Pretty good chance that is where the A14X is aligned with. ( Apple could just wink out 1-2 GPU and/or 1-2 Performance CPU cores ) and slap a different label on it.


The Intel MacBook Air only has two TB ports and the old non-Touchbar MacBook Pro 13 also only has two ports. So the M1 is meeting that and it's $100 cheaper than the base Intel model so you could argue that is to reflect losing those two ports.

Once even up the M1 MBA to 8 GPU cores and match up the SSD level the price to MBP 13". The price the same.
The "touch bar" haters can buy the MBA. The folks don't mind the "touch bar" can get the MBP 13".
Those two are almost the same system. ( Better thermal curve probably on MBP 13" and 100 nits of brightness on MBP 13' but that is probably lost on most users. ).

The "butterfly keyboard" dustup eclipsed the backlash on the "touch bar" , but the two laptops are pretty close to the same model just with "user selected" option out/into the "touch bar".

The Mini also has major problems compared with WinPC desktops as the price range goes up. Ryzen 7 5800X paired with a modest discrete GPU and the Mini isn't going to look like some kind of "Performance King" anymore of a wide variety of workloads. ( yeah it is smaller , but faster starts to slip away. ) . Apple is a bit luckly that AMD bumped the Ryzen 7 5000 prices up a bit.




As for the M1 Mac Mini, I see this as the model for software developers (who don't want to use an M-series laptop) and for folks who use their Mac Mini as a home media server or such. So the RAM and I/O restrictions are not crippling and you pay less.

An M1 Mac Mini as a media server is just gross overkill. I'm sure there is some fringe that will do it . There are probably more units allocated to cloud services ( LAN and remote : virtual machine hosting , on demand "test and build" systems, etc. )

A mini is going to help more revised software get out there faster as some major development build chains use them. ( that is why the DTK form factor made sense also. )

The RAM constraint will hit folks with substantive concurrent virtual machines images running. Apple is throwing other minor roadblocks are virtual machine hosting but that cap is going to limit things.

The cheaper helps in that it will probably encourage more exploring by more groups of whether they want to make the move. Also going to be some folks engaged if the iMacs take longer than folks are expecting.



The next model of ASi SoC (M1X / M2) will likely offer more cores (CPU and GPU), a higher unified memory limit (32GB+) and will have additional controllers for more USB/TB ports. That is the SoC I expect to go into the 14" and 16" MacBook Pros and they will have four ports.

Adding another TB controller isn't going to help if still stuck with just two DisplayPort streams. The GPU subsystem will nee substantive improvements too. At least one more DP output stream.
 
Since the A14X hasn't been shown the position that this is "extra Mac only" hardware really isn't grounded.

Well if the 2021 iPad Pros come with USB4, TB3 and PCIe I will stand corrected. :)


An M1 Mac Mini as a media server is just gross overkill. I'm sure there is some fringe that will do it.

Seems more than a fringe based on these forums. ;)


There are probably more units allocated to cloud services ( LAN and remote : virtual machine hosting , on demand "test and build" systems, etc. )

Indeed. Which is why I believe we will see more powerful Mac Minis (with more unified RAM and more powerful SoCs) down the road to address those markets.


Adding another TB controller isn't going to help if still stuck with just two DisplayPort streams. The GPU subsystem will nee substantive improvements too. At least one more DP output stream.

Well there is the Apple "Lifuka" GPU waiting in the wings...
 
Seems more than a fringe based on these forums. ;)
This ENTIRE forum, like every single person, is fringe compared to the 100 million macs in use (unless there are 100 million accounts in these forums). So, any smaller percentage of “these forums” is even more... fringey :D
 
  • Like
Reactions: CWallace
So I downloaded the USB4 spec. I had to dig around a bit, but I found:
"A USB4 host supports 20G USB4 operation (Gen2x2) and optionally 40G USB4 operation (Gen3x2)."

So it seems like Apple is giving false advertising on USB4 "support". Either that or the M1 actually does support Gen2x2 and the specs are wrong... but I'm guessing it's the former.
I think 20G USB4 operation (Gen2x2) means the tunnelled packet stream type of operation (combines DisplayPort, PCIe, and USBx) which is like Thunderbolt.

The USB 3.2 Gen 2 x 2 (20 Gbps) mode is like USB 3.1 gen 2 but uses two lanes - the only traffic it contains is USB.

Chapter 9 of the USB4 spec describes USB 3 tunnelling over USB4. In 9.2.1 it says that "The Link Layer shall support Gen 2 Single-Lane (2x1) and may support Gen 2 Dual-Lane (2x2)."
This means that a USB4 port does not necessarily support dual lane USB3.
 
I expect the M2 to be different next year. It may have SMT-cores (disabled for the A15X variants), 128-256 GB memory addressing, and 4 TB4 ports. This is but the most logical choices for the bigger MBPs and Desktop Macs if Apple needs to to justify Apple Silicon’s legitimacy. The M2 may be the arch that Apple Silicon have designed for the desktop from the ground up.


The way Apple throws gobs of transistor budget at outsized L1, L2, and system cache, they are not likley to buy into SMT. SMT is more effective when the core is stalled waiting on getting some data back from far down on the memory hierarchy. What is Apple is doing is going through lots of gyrations that there are fewer stalls . That cranks IPC up and it also means that function units are kept more actively busy doing at least one piece of the "puzzle". Similar with the re-order buffers being large so that can find "something" to work on that might load sooner rather than later.

SMT can get more work done, but has some energy used trade-off to more to the "other" state. Or keep status of multiple inflight states. SMT is not really a problem space that phones have. Large data base scan/joins or lots of pointer chasing memory access give a much bigger target for SMT to get traction on. "Lighting up" more function units for as long as possbile will burn more power. That isn't the primary objective in Apple's designs. Apple wants to eek out more performance by expending the minimal amount of additional power. SMT makes far more sense for cores that are design specifically to be plugged into the wall all the time. ( quite doubtful even in the Mac space Apple is going to do that. Even their desktops are going to be about less CPU power consumption. Likely a "green , save the planet" marketing talking point. )




Evidently, where Apple is doing more SMT like stuff is in the GPU not the CPU. That makes more sense because dealing with "Mainstream" system RAM as the main pool for the GPU. Order of magnitude more cores that are being somewhat "starved" and so move them on to something else that hopefully has some coverage in the more local cache.


As for triple digit GB ram capacities .... errr decent chance that may be 2022. I suspect they would rather do that later. They are going to completely flush out these soldered down RAM solutions first across as much of the Mac line up they can get away with first. Only then , in space time, move on to something bigger. And leverage DDR5 (LDDDR5 ) increase in DRAM package memory density during 2021 to "kick the can" further out into the future





Having said that, by just looking at the M1’s features, it’s very likely to expect the 2021 iPad Pro to feature a TB3-capable port and 8GB of memory (since Samsung, Kioxia and Hynix have already began shipping the next-gen, higher capacity 16Gbit mobile memory modules late last year).

I suspect TB3 will be there on the iPad Pro SoC. It is questionable though Apple is going to open up that vector ( even with the IOMMU lock down abilities they have) in iPad OS. It means more drivers with more variability hooked to the kernel. As opposed to Apple written USB drivers that are not 3rd party extension "add ons".

An iPad Pro with USB 4 ( the only the required parts) would be step up but put the "Pro" between the iPad ( USB 3) and Macs ( TBv3 ). Less power required likely also.
 
Last edited:
Anyone get the feeling that these were rushed? They're more like proof-of-concepts. I guess I'll wait a bit to see what's in store for 2021. Or maybe 2022. Or I bet there'll be big improvements for 2023.
I feel like the release was perfecting timing for 2020 and pre-holiday.

Of course, technology is going to be more advanced one, two, or three years later .... It's all about the buyer's needs. Some people need an upgrade and don't want to wait for Gen 2 or Gen 3 years down the road. Just my two cents.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.