I tried SO HARD to break this…

[Emily Young]

The M1 Mac Mini was an impressive machine with one major caveat – Apple removed the 10 gigabit NIC. We thought it was because they couldn’t make it work. We thought wrong… Or did we?

 

 

Buy Apple Mac mini M1 (PAID LINK): https://geni.us/jJPvr

[SorryBella]

Quote

I tried SO HARD

And got SO FAR

 

Thanks for the test video, was waiting for it since the story tease.

 

[Zmimgo]

Just now, SorryClaire said:

Quote

I tried SO HARD

And got SO FAR

 

but in the end

[Nathanpete]

1 minute ago, Zongohihello said:

but in the end

It doesn't even matter!

[Guest]

5 minutes ago, SorryClaire said:

And got SO FAR

 

[SorryBella]

1 minute ago, Den-Fi said:

 

Oh well, it was a shit joke afterall. Sorry, but i know you cant read this.

[Guest]

3 minutes ago, SorryClaire said:

 

Oh well, it was a shit joke afterall. Sorry, but i know you cant read this.

I was also joking.

[Egg-Roll]

I tried so hard to break this...

 

Have you tried using a hammer?

Like this:

[Nathanpete]

13 minutes ago, Den-Fi said:

I was also joking.

Whats the website where you can add the petting animation to images? I saw it months ago but I forgot. 

[Murasaki]

1 minute ago, Nathanpete said:

Whats the website where you can add the petting animation to images? I saw it months ago but I forgot. 

Yikes, meanwhile I've been doing these by hand in After Effects...

[BondiBlue]

Hey @GabenJr, is there a reason you guys always capitalize the word "mini"? It's stylized as Mac mini, just like Apple has done with the HomePod mini, iPad mini, and iPod mini. 

[Emily Young]

3 minutes ago, BondiBlue said:

Hey @GabenJr, is there a reason you guys always capitalize the word "mini"? It's stylized as Mac mini, just like Apple has done with the HomePod mini, iPad mini, and iPod mini. 

Honestly? This is embarrassing, but I just never noticed.

[BondiBlue]

1 minute ago, GabenJr said:

Honestly? This is embarrassing, but I just never noticed.

Yeah, it's something they do with the word mini. I guess they do it because mini = small, and their mini devices are smaller than the non-mini versions, like the iPad and the HomePod. 

[Splay80500]

1 hour ago, GabenJr said:

The M1 Mac Mini was an impressive machine with one major caveat – Apple removed the 10 gigabit NIC. We thought it was because they couldn’t make it work. We thought wrong… Or did we?

 

 

Buy Apple Mac mini M1 (PAID LINK): https://geni.us/jJPvr

Hi, just asking the script that you used in the video, that Anthony repurposed for the usb SSD's. Could we use the script so we could run tests on our own MAC system's  

[Spindel]

Hey great for Linus that he got his 10gbps port. 
 

But it is no mystery why Apple left it out initially. 
 

It is a niche need. 

[Musky C.]

5 hours ago, GabenJr said:

The M1 Mac Mini was an impressive machine with one major caveat – Apple removed the 10 gigabit NIC. We thought it was because they couldn’t make it work. We thought wrong… Or did we?

 

 

Buy Apple Mac mini M1 (PAID LINK): https://geni.us/jJPvr

I dont want to say anything, but he didnt even drop it once. Did he REALLY TRY TO BRAKE IT then?

[Ashleyyyy]

gotta love companies that you can trust with the numbers on their product page.

[caincha]

The tittle says it all: LTT gets a Mac on their hands and immediately tries to break it for clickbaits.

Oh well, get yourselves a PC to do that job instead why don't you…? 

[power666]

Want to nerf the 10 Gbit Ethernet port on that M1 Mac Mini?  Leverage the bandwidth reservation features of AVB of that Ethernet port and start moving around several hundred ultra high bit rate audio channels through it.  Once the AVB stream has been allocated, the rest of the system only gets access to the rest of the bandwidth, regardless if they are in active use.  The recent Marvell/Aquantia NICs support it and I use an ASUS XG-C100C card as a 16 channel output card on a classic cheese grater Mac Pro.

 

Interestingly enough enough, all the features are there in the M1 chip to leverage that 10 Gbit Ethernet port for M-JPEG or M-JPEG2000 (or JPEG-XS?) encoded streams to act as a video output.  This would be an interesting test of the hardware accelerated encoders of the M1 to see how low of latency it would be.

[geerlingguy]

@GabenJr - I appreciate the fact that you probably spent dozens of hours working on this setup, then had to call an audible probably mid-recording when that drive started acting funny. Nice work, and this was one of my favorite videos in recent times, just in terms of the effectiveness of illustrating the problem/solution, and teaching about common system bandwidth bottlenecks and discovering how the buses really work in spite of a lack of documentation from Apple.

[ari735]

I'm loving my M1 so far. It's not the Mini, but it's served me really well so far - glad to see they're beefing up the line-up.

[mariushm]

It's a shame the video was spoiled from the start with Linus saying that PCI network cards could not do 1 gbps because of pci limitations.

That's simply not true... PCI 32 bit 33 mhz can do 133 MB/s while PCI-X can go up to 533 MB/s (but the 266 MB/s 64bit 33 mhz was more common I think)

 

So basically even  at full 1 gbps, that's only 125 MB/s  ... even with overhead and other devices 110-125 MB/s was doable, or 90% or more of the advertised bandwidth.

Not fair to compare it with usb ports that may only get half their advertised speed when multiple ports are used at same time.

 

[power666]

On 5/14/2021 at 6:28 PM, mariushm said:

It's a shame the video was spoiled from the start with Linus saying that PCI network cards could not do 1 gbps because of pci limitations.

That's simply not true... PCI 32 bit 33 mhz can do 133 MB/s while PCI-X can go up to 533 MB/s (but the 266 MB/s 64bit 33 mhz was more common I think)

 

So basically even  at full 1 gbps, that's only 125 MB/s  ... even with overhead and other devices 110-125 MB/s was doable, or 90% or more of the advertised bandwidth.

Not fair to compare it with usb ports that may only get half their advertised speed when multiple ports are used at same time.

 

The bandwidth was there for 32 bit, 33 Mhz PCI only if the NIC was the only device on the bus.  The catch is that legacy PCI is a shared bus where that bandwidth is shared with other devices like on board audio or early USB controllers.  This is very similar in concept where USB host controller bandwidth is indeed shared.  The most notorious things were early SATA controllers which would share the same PCIe bus as the network card on lazy motherboard implementations.  Doing a network file transfer and storing it on that SATA drive?  Expect half bandwidth as the network card and SATA controller could not communicate directly to each other, thus data had to traverse the PCI bus twice to be put to rest on the disk (NIC to memory then memory to disk).  

 

PCI did have 66 Mhz versions and PCI-X increased clocks up to 133 Mhz and 64 bit widths.  The gotta is that the 66 Mhz version of PCI would drop to the lowest, most common speed on the bus which was often 33 Mhz.  PCI-X due to is increased width would often permit up to two 100 Mhz 64 bit slots or a single 133 Mhz 64 bit slot off of a PCI-X controller chip.  Most implementations of PCI-X were implemented as a point-to-point scenario so that all the bandwidth could be allocated to a slot, though this was not a hard rule of the spec.

 

Such issues today are 'solved problems' as PCIe is dedicated bandwidth per lane and only when aggregated through a PCIe switch or chipset bridge does bandwidth really become shared at some juncture.  Further more, there is an increase in device-to-device transfer support so that say a smart NIC can communicate directly with a NVMe storage device without actively involving the host CPU or constraining uplink bandwidth if they are on the same PCIe root/switch. 

[mariushm]

8 hours ago, power666 said:

The bandwidth was there for 32 bit, 33 Mhz PCI only if the NIC was the only device on the bus.  The catch is that legacy PCI is a shared bus where that bandwidth is shared with other devices like on board audio or early USB controllers.  This is very similar in concept where USB host controller bandwidth is indeed shared.  The most notorious things were early SATA controllers which would share the same PCIe bus as the network card on lazy motherboard implementations. 

Yeah .... doubtful.

A sound card uses maybe 2-5 MB of bandwidth, same for the USB (as most users don't use heavily more than a single port) ... simply put majority of users didn't have a lot of pci cards to saturate the bus, so even when other devices took control of PCI bus the network card could still get 100+ MB/s

I've been using gigabit cards since the socket A days and even then the sata controller was in chipset, not hogging the pci bus. I did have motherboards with extra sata controllers but all motherboards I ever had were made with chipsets that had controllers built in, not connected to pci slot.

 

Either way, nobody was complaining, 1gbps allowed up to 125 MB/s at a point where IDE hard drives were barely exceeding 50-80 MB/s.

Being able to get over 100 MB/s of transfer speed was still way better than being stuck at 12 MB/s (100 mbps ethernet) ... just like probably nobody would mind if they can't get over 850-900 MB/s on a 10gbps ethernet card because it's plugged into a pci-e x1 slot.

 

[zENjA]

For testing, I thought about connecting two external TB3 "GPU" Box, but put 40/100G Mellanox cards in them.

Yes, I know iperf is CPU bound on most systems I know at multi 40g or higher systems

May I see my DC Customers bringing these mini mac's in as a cluster... would be interesting