Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Can somebody give me the idiots guide between 8-10 Gigabit ethernet?
Is it quite literally a speed bump.
gigabit maxes out at gigabit speeds where 10 gigabit maxes out at 10gigabit speeds. It’s not about getting higher speeds while connected to the internet it’s about transferring data within a local network through a nas or something. 10 gigabit ethernet is a better choice than thunderbolt for data storage because the cabling and implementation is way way less expensive and, while the theoretical max speed of 10g ethernet is a quarter of thunderbolt’s max throughput, with 10g ethernet we’re still talking about 1.25 GB/s transfer rates.
 
I was just thinking about how powerful and power efficient an M1 powered server could be. The XServe was a turbine. I’m sure we will eventually see an M1 Mac mini supercomputer.
will they have IPMI? dual power? hot swap storage, raid 1?, no need to hook up and 2th mac to change storage card like on the mac pro?
 
  • Like
Reactions: Mainyehc
No need per se. M1 Mini has the same number of controllers as the outgoing model, so throughput is actually the same. Most docks/devices have a daisy chain feature, so in effect it's the same as previous.

If the controllers aren't' connected to anything then it doesn't matter if there are two embedded in the die.

If there are two controllers in the die and Apple only hooks up one x4 PCIe-v3 feed and one DisplayPort v1.4 stream then that second controller is "useless".

If could be cheaper just for apple to stuff the two controller subsystem in to the smaller die that the M1 uses because it perturbs the overall design less. Apple can just throw the added, "dead" die cost into the price if the system. They aren't going to loose any money on it if they charge for it anyway and don't deliver the functionality present. Apple could also likely power isolate the second controller and just flip it completely off it won't every be used on the die ( that will save power).

A larger M1 variant ( e.g. something like "M1X" ) could have a more decent iGPU display output subsystem that can feed four DPv.14 streams to the subsystem and another x4 PCI-e feed to the second controller. Those systems could have four ports but the SoC would be larger. ( possible attached to 2 more RAM packages so the overall package larger. Bigger die ( e.g., 120mm^2 --> 240mm^2 ) and twice RAM package footprint.

This way they sell lots more copies of the same two controller TB subsystem. That allows them to amortize the fixed costs ( R&D , validation , software drivers ,etc) over a large number of systems which leads to higher Apple profit margins.




P.S. If this driven by ' there are "two controllers" because there are two Intel Redrivers on the logic board' ... Those aren't controllers. But yes there seems to be some chatter that the Thunderbolt implementation in the M1 die is more complete that what is using in the M1 implementation. ( probably because it is also the A14X die with some features flipped on/off.) .
 
Last edited:
10G Ethernet REALLY is needed today. I'm still on 1000BaseT and it shows that files from my NAS storage are slower than the files on my USB3 connected SSD.

This matters on a desktop PC like the Mini or the iMac because these computers are more likely to be connected to network files servers and people are now used to the speed of SSD drives and want that from network files.

Also, these computers have such a tiny amount of internal storage then yu rally NEED external storage if you are doing things like working with media files (video, audio or photos)

Having 10G would make this a MUCH more usable and attractive product. In fact, the Mini would make a really nice file server if it had 10G Ethernet. One could buy a Thunderbolt external storage system and then the Mini could serve files over 10G Ethernet and WiFi.

As a server, the M1 chip could pay for itself just from the power savings. 100W of savings means 2.4 KWh saving per day and at 20 cents per KWh Over the five-year long life of the mac-mini that works out to save $876. That $876 is "real money" (assuming you really do save 100W by retiring some big Intell powered box.
 
Mac mini Pro coming? With space grey finish?

I would expect a bigger "M1X" standard configuration for the normal "better or best" Mac Mini and that 10GbE just be the standard configuration.

It is kind of ridiculous at this point. 1GbE was standardized in 1999. That some $900+ systems comes capped with last century technology is just plain lame. The standard wired connection should be faster than Wi-Fi because that is relatively easily doable.
 
Are there any here that have transferred over to real 10GB ethernet on their home network? If so what did it cost and what equipment would they recommend in lower price racket.
You can get a a quality 10gb managed switch from TrendNet (4 ports up to 5gb and 4 ports up to 10gb) for under $600. Intel cards for desktops are $200. While the budget switch is fine, mine has been running solid for over a year, don’t cheap out on the network cards. The intel x550 cards are tanks. Other cheap cards often fail due to heat. As for speed? A 4 bay 10gb NAS will get you speeds from 250mb to 500mb avg. with m.2 cache you’ll even hit 900mb a/s. 6 bay and especially 8 bay NAS is where you start seeing more full time 10gb speeds. Very sexy. 4 bay NAS start around $1200 for a good one, and then drive costs. Ironwolf drives are my choice.
 
Last edited:
  • Love
Reactions: iBluetooth
Are there any here that have transferred over to real 10GB ethernet on their home network? If so what did it cost and what equipment would they recommend in lower price racket.
Not full 10G except with SFP+, but I have a 2.5G switch (Ubiquiti UniFi U6-S24). Won't realize full 10G, but it allows 2.5G without using an adapter. Right now 10G switches are too expensive to implement.
 
Are there any here that have transferred over to real 10GB ethernet on their home network? If so what did it cost and what equipment would they recommend in lower price racket.

The 10GbE controller that Apple has used covers 2.5 and 5 also. 10GbE also means when costs come down later can also move to 10GbE. But between two or more systems in a more mainstream household context there are 2.5-5GBe switches now that are more affordable. ( like in range 1GbE was early 10-12 years ago.) and there are home NAS systems coming with 2.5-5 ports on them.

2.5GbE switch in the $100 Range.
https://www.qnap.com/en-us/product/qsw-1105-5t

~ $360 2.5GbE NAS
https://www.qnap.com/en-us/product/ts-253d

You don't have to be out $700+ to get to a network were have gone past the 1GbE limit. Home backups that finish faster without a much more expensive NAS/SAN set up.

but yeah the 10GbE switches have stayed stubbornly high ( the switch and component vendors just don't want to let go of those margins. And home infrastructure can have more limited (dated) cabling. )
 
Last edited:
  • Love
Reactions: iBluetooth
Ultimately Apple is going to have to make XServe's again with M-class chips, probably when Mac Pro is released.

While Apple have not made servers for 10 years, and there is less need for Apple server software now except Apple Caching Server which a Mac mini can do fine as you don't need redundant PSU or RAID there could actually be a decent market for Apple XServe's again using M1 for other server software that is unix based.

Big data centres want low power, low heat, and an M1 Xserve would tick the boxes, their old servers were nice bits of kit too. The question is though are Apple even interested ? or even realise the big potential in the server market for the M1 with the right price and support cycle as it is now very compelling proposition compared to a traditional x86 based server.
 
Are there any here that have transferred over to real 10GB ethernet on their home network? If so what did it cost and what equipment would they recommend in lower price racket.
I have a 10GbE LAN, basically between two rooms in my home. Consider them "server room" and "home office".

The sever room has a Netgear M5300-28G3 which has 24 1Gb ports, and two 10Gb ports. One of those 10Gb ports goes across the house to my home office with a Netgear XS708T that has eight 10Gb ports.

The big switch was about $300 and the smaller one was about $400. Note that the older/retired-enterprise 10GbE switches have pretty loud fans. So swapping those out is added cost/complexity.

Aside from faster file transfers, my interest in 10GbE mainly revolves around using headless machines. If you're going to Remote Desktop into Macs, the screen sharing can be improved with the extra bandwidth available.
 
  • Love
Reactions: iBluetooth
...
Big data centres want low power, low heat, and an M1 Xserve would tick the boxes, their old servers were nice bits of kit too. The question is though are Apple even interested ? or even realise the big potential in the server market for the M1 with the right price and support cycle as it is now very compelling proposition compared to a traditional x86 based server.

There are already multiple vendors there already with SoC that are specifically data center targeted. ( built in Ethernet , ECC RAM , etc. )

https://www.nextplatform.com/2020/0...silver-altra-lineup-128-core-mystique-kicker/

https://www.nextplatform.com/2020/03/16/marvell-cranks-up-cores-and-clocks-with-triton-thunderx3/


Some ex-apple folks jumped on bandwagon too.

https://www.reuters.com/article/nuv...s-240-mln-for-data-center-chips-idUSL2N2GL032

Amazon is doing their own for their cloud.

https://www.anandtech.com/show/15578/cloud-clash-amazon-graviton2-arm-against-intel-and-amd


(Microsoft Azure , Google , Oracle , Tencent , etc. other major cloud vendors are rolling out Marvell and/or Ampere stuff over next several months. ) MacOS on M1 doesn't compete with those Linux solutions with higher synergy with what is already there in data centers. M-series doesn't run anything else native. So any data center or cloud service that has some other standard hypervisor foundation is stuck not being a non-native hypervisior. That is going to go over brick rock in a pond of water in the data center market. ( Telling them "You can't run your favorite base software" ... that means you are done in the bidding competition. )

ARM has targeted this market specifically with their Neoverse solution. Apple doesn't have a huge lead here at all. Laptops? yes.(ARM largely ignored that, but is iterating on Cortex X now where de-focus on smartphones. ) But Severs? No. Amazon had deployed ARM instances before Apple even handed out the DTK. Marvell ThunderX is delivery on their 3rd generation solution. Apple isn't a 'first mover' in the slightest.


P.S. Apple's strict licensing restrictions on macOS gets them a relatively small niche in the data center / cloud services spaces. It is quite small. This only macOS native boot and sit on top of Apple hypervisior is going to make it a bit more narrow (and even more dependent upon the license to "drive" the business). The Mini and rack Mac Pro variants can more than fill the vast majority of that space with minimal "extra" R&D costs from Apple.
 
Last edited:
  • Like
Reactions: sos47
Are there any here that have transferred over to real 10GB ethernet on their home network? If so what did it cost and what equipment would they recommend in lower price racket.

Assuming the devices you're looking to connect actually have 10G Ethernet already, these are working great for me to connect my iMac to my Mac Mini server:

- MikroTik CRS305-1G-4S+IN
- Ipolex 10G SFP+ RJ45 Copper Transceiver, 10GBase-T Module (2x)

The 4 SFP ports on this mini switch give you room to add more than 2 devices, and the 1G port is great for adding onto your existing network, especially using PoE.
_____

If you'd prefer to have one device connect your whole network, and just need 2x 10G ports for 2 devices:

- Netgear Nighthawk GS810EMX


My current use case is really just to connect 2 devices for frequent large data transfers. Switching over to a bigger 10G switch (12x 10G ports or more) to connect every device in the home doesn't make sense yet, since most devices don't have 10G ports, and those switches are far too expensive either way.

Hope this helps!
 
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
There are already multiple vendors there already with SoC that are specifically data center targeted. ( built in Ethernet , ECC RAM , etc. )

https://www.nextplatform.com/2020/0...silver-altra-lineup-128-core-mystique-kicker/

https://www.nextplatform.com/2020/03/16/marvell-cranks-up-cores-and-clocks-with-triton-thunderx3/


Some ex-apple folks jumped on bandwagon too.

https://www.reuters.com/article/nuv...s-240-mln-for-data-center-chips-idUSL2N2GL032

Amazon is doing their own for their cloud.

https://www.anandtech.com/show/15578/cloud-clash-amazon-graviton2-arm-against-intel-and-amd


(Microsoft Azure , Google , Oracle , Tencent , etc. other major cloud vendors are rolling out Marvell and/or Ampere stuff over next several months. ) MacOS on M1 doesn't compete with those Linux solutions with higher synergy with what is already there in data centers. M-series doesn't run anything else native. So any data center or cloud service that has some other standard hypervisor foundation is stuck not being a non-native hypervisior. That is going to go over brick rock in a pond of water in the data center market. ( Telling them "You can't run your favorite base software" ... that means you are done in the bidding competition. )

ARM has targeted this market specifically with their Neoverse solution. Apple doesn't have a huge lead here at all. Laptops? yes.(ARM largely ignored that, but is iterating on Cortex X now where de-focus on smartphones. ) But Severs? No. Amazon had deployed ARM instances before Apple even handed out the DTK. Marvell ThunderX is delivery on their 3rd generation solution. Apple isn't a 'first mover' in the slightest.


P.S. Apple's strict licensing restrictions on macOS gets them a relatively small niche in the data center / cloud services spaces. It is quite small. This only macOS native boot and sit on top of Apple hypervisior is going to make it a bit more narrow (and even more dependent upon the license to "drive" the business). The Mini and rack Mac Pro variants can more than fill the vast majority of that space with minimal "extra" R&D costs from Apple.

However for the competing Windows desktop ... I somehow suspect that Samsung will be a bigger player than Qualcomm.

Samsung will try to emulate the process of evolving their own Exynos chips to support more powerful laptops and desktops for Windows or Linux for consumers or professional use, partnering with Dell, Lenovo, HP ... OR ... they'll manufacture chips for Qualcomm and thus do the same thing.

Let's see how long it takes Microsoft to get their Windows ARM working, 3rd party developers for big softwaer (Adobe, etc) and manufacturers to build laptops and desktops of all classes to support business and consumer lineups to match or beat the performance of Apple's M1 chips.
 
While Apple have not made servers for 10 years, and there is less need for Apple server software now except Apple Caching Server which a Mac mini can do fine as you don't need redundant PSU or RAID there could actually be a decent market for Apple XServe's again using M1 for other server software that is unix based.

Big data centres want low power, low heat, and an M1 Xserve would tick the boxes, their old servers were nice bits of kit too. The question is though are Apple even interested ? or even realise the big potential in the server market for the M1 with the right price and support cycle as it is now very compelling proposition compared to a traditional x86 based server.
Ultimately Apple is going to have to make XServe's again with M-class chips, probably when Mac Pro is released.
I don't really see Apple going back into servers. Having seen Apple server gui apps (Workgroup Manager, Server.app, etc.) & Xserves go the way of the dodo, I seriously doubt we'll see a new Apple server any time soon. I find that ever since the iPhone's first release, Apple's gone more consumer-focused with any enterprise-level stuff more of an afterthought.

However, I would like to see an Apple server. Considering the speed of the M1, plus the fact that Apple brings in billions of dollars in profits every year, it's definitely possible. Make the server utilities easy to install, configure, and maintain, and Apple could do a real number on the server world.
 
I reckon the intel space grey Mac mini will be replaced with a higher-end Apple Sillicon chip. This will be equipped with the 10 Gigabit Ethernet and will be the 'Mac Pro mini' that was rumoured earlier in the year (though I think it'll still be called the Mac mini, the only differentiation between it and the current M1 mini being the Space Grey colour and the specs).
 
I just ordered a M1 mini this morning. I almost canceled in favor of waiting for the 10G model... I have several systems with PCIe/NVMe storage along with a NAS array that continually hit a wall at 1Gbps. With 10Gbe copper switches still around $1000, I'm going to wait two more years. I'd be nice if they came with dual 1G ports for link aggregation, but then it wouldn't be a mini anymore.

I think Apple should lead the way with pushing the Ethernet standard and make 100/1000/10000 RJ45 ports standard. They didn't hesitate to jump on USB-C.
 
  • Like
Reactions: ikir
No need per se. M1 Mini has the same number of controllers as the outgoing model, so throughput is actually the same. Most docks/devices have a daisy chain feature, so in effect it's the same as previous.
The M1 has as many controllers as the intel but each one of them only supports a single port. So yes 2 controllers but only 2 ports.
 
Honestly think Base Mini Mac M1 could be worth a home setup.
Nice future hardware path.
Some nice entry configs -8/16 gb ram 256/512 ssd.
Very interesting.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.