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

hajime

macrumors 604
Original poster
Jul 23, 2007
7,949
1,315
Hi, for the M2 Mini, if I upgrade the ethernet to 10gb, will it take up some thunderbolt 4 buses from other part of the system? Are all those thunderbolt 4 ports independent or not? What chipset is used in their 10gb ethernet?

How is the situation for M1 Mini?
 
I asked this same question back when the 2018 intel launched and never really got to the bottom of it. I suspect it doesn't impact TB4 bandwidth and if it does, I haven't noticed it on TB3.

Temps will be a little hotter. 10G will always run hotter than gigabit but again no instability.

10G chip is Aquantia AQC107
 
  • Like
Reactions: jagooch
I asked Apple and they have no idea!!!

Heat is a concern. I think some members said that M2 MacBook/Pro runs hotter. I wonder if M2 in the mini has the same problem and thermal throttle.
 
Thank you. How likely do those M2 use the same 10GbE controller? To be certain, we need to wait for somebody to get a M2 Mini with an upgraded ethernet to check?
 
I have seen a few threads of complaints with people getting the 10Gb upgrade, and when using it on a 1Gb network on CAT5e, it was reducing the link speed to 100Mbps.

Apple support seems to advise the users to switch to CAT6a (or higher) cable, which fixes it the issue, but it shouldn’t need CAT6a for a 1Gb link speed.

I don’t know if this is a HW or SW thing, nor do I know how widespread it is.

I was going to get the Ethernet upgrade on my next Mac, but the reports have discouraged me to.
 
Thank you. How likely do those M2 use the same 10GbE controller? To be certain, we need to wait for somebody to get a M2 Mini with an upgraded ethernet to check?
Does Marvell/Aquantia have an updated controller? They're pretty much the only supplier of desktop PCIe 10GbE chips...
 
So Apple does not make their own 10GbE chips and integrated into their own M1 and M2 series of silicons?
 
So Apple does not make their own 10GbE chips and integrated into their own M1 and M2 series of silicons?
No - they don't even make their own wifi/bluetooth and cellular chips yet, both of which are far more important than 10GbE controllers...
 
In the output from ioreg -i, Thunderbolt controllers in Apple Silicon Macs are separate platform devices integrated into the CPU using the IOPCIHostBridge class driver. Each one is a completely separate PCIe bus. The 10 GbE controller is under a separate IOPCIHostBridge.
 
  • Like
Reactions: Ruftzooi and hajime
No - they don't even make their own wifi/bluetooth and cellular chips yet, both of which are far more important than 10GbE controllers...
More precisely they don't make wifi/bluetooth for phones and macs (yet...)
AirPods and Apple Watch use Apple's BT/WiFi chips (either H2 in Airpods or W3 in Apple Watch).
 
Does it finally support TCP offload engine now otherwise throughput will take a hit like Mac Studio M1 Max when CPU is busy doing other things slowing down software TCP checksum calculation.

https://forums.macrumors.com/thread...sor-load.2344566/?post=31095544#post-31095544

Generically Apple has a massive amount of network functionality in their NoC implemented as "smart DMA". This can (at least according to a sequence of multiple patents, all of which are many years old) do all the sort of network offload you'd want, from removing heads and tails to checksum to compression and encryption. Because this is in the NoC it is available to any network client, from Bluetooth/Thread to WiFi to ethernet.

Now I am unaware of any explicit *proof* that this has been implemented, and I'm not sure how you could "prove" it. But as I say, the patent record is strong and makes sense (obvious improvements one year after the next, building on the previous implementation); and the idea makes sense (reduce energy by doing as much as possible, more every year, to cut the energy costs of on-going network chatter for iPhone and Apple Watch).

I'd say (for the particular issue in that thread) most likely the problem was a bug in the OS that has probably been fixed (if it was reported...) My experience in a different context (USB ethernet adaptors) has been that Apple has sometimes shipped a driver in the state of "works but not optimized" (presumably to hit some deadline) but a subsequent OS update gets everything right and you see performance jump by 2x or CPU usage halve or something similar.

Obviously, for example, if the OS and ethernet driver on some machine don't (somewhere between the two of them) make use of the smart DMA modifiers that perform this offload, then you will not see the benefits...
This is especially an issue for Macs that don't have a built-in ethernet port (eg MacBooks Air's+Pro's).
So who is providing the ethernet driver for this machine? If it's something that's well supported by Apple [obviously the Apple ethernet adaptor, but probably also some 3rd party brands like those sold at the Apple store], you should be in luck. If it's some random no-name brand, then you're at the mercy of their ethernet driver...

Here's an example of the issue:
 
Last edited by a moderator:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.