> The compact size of the Mac mini, which packs a powerful System on a Chip (SoC) into a tiny footprint.The energy efficiency of Apple silicon (M-series) chips, which allows high density without overheating or excessive power draw.
This really adds nothing to the article, and looks like AI fluff to me.
Combine that with there being a bold section in like every single paragraph, I'm going to assume yes
As to whether it was AI generated or not, who cares? It's useful information if you didn't know it already, and if those words came out of matrix math or someone non-technical with a BS in communications, does it really matter to you? Are you going hungry tonight because the money that went to creating those words went to Nvidia and not Sarah in Marketing? Sarah in Marketing might be out of a job soon, but her boyfriend has a good job that's not threatened by AI, so I hope she'll be fine, but I don't know. Is that the underlying worry here?
There is an emdash in the article though, you didn't think to call that out too?
> Scaleway’s solution to that problem was ingenious: embedding a Raspberry Pi module with each Mac mini.
(I realize this may be an artifact of a corporate style guide, but I'd much prefer "Our solution to that problem was embedding . . ." Both because the "was ingenious" doesn't add a ton and reads like puffery and because this is Scaleway's own blog and referring to yourself in the third person is grating.)
Simple... they're (likely) running something on the Raspberry Pi's that sets them up as USB gadgets, aka the Mac Mini "sees" a virtual keyboard and mouse. That's enough to manage remote provisioning.
To replicate that they'd need a KVM switch which doesn't have some weird edge case in how exactly it does USB-C switching, and it needs to be remotely controlled. A Pi is cheaper plus the failure modes of a Pi are more understood than the failure mode of some weird ass KVM switch someone cobbled together in China.
>/sys/power/s_stats is peripheral in terms of antico grego and a shallow-clientx .async language in thread_entry, arg
To do this, you will need a smart controller that switches which port it’s talking to.
Or you can stick a relatively cheap device on every mini and and connect it to the network.
Having a “controller” for every mini means you can swap single units in both hardware and software very easily. There’s a one-to-one relationship and you don’t have to deal with pairing.
Is video forwarding part of the product offering, or simply considered required for management functions?
Either way, I probably agree that a raspi per unit probably makes sense at a scale of a few dozen racks, but it would be interesting to do the math on when it would be price-efficient to have a 1:n management node scheme. I don't imagine there are many USB-C hubs that support being display sinks on the downstream ports (if that's even possible at all) but perhaps you could use an FPGA to synthesize a small ARM core with a bunch of native USB-C interfaces capable of doing it?
The Raspberry Pi 4 can emulate a USB keyboard and mouse, and there are inexpensive adapters that allow it to capture display output. You can also hook it up to a relay to cycle power for an external device.
You might want to look at Asahi Linux's Central Scrutinizer for some insights on the fancier stuff that can be done over the USB-C port.
This is not the image I expected to encounter under the title, “high density”.
Make those sleds taller and do three, maybe four per sled with a pair of large diameter fans. That’ll would be high density. This is medium at best.
Or find cable with matching length?
But I agree that for a big scale, this is a good solution. (cf: github)
Our way is quite efficient & we're able to quickly adapt to new HW gens
We're actually thinking about features which will that power button to good use in a near feature, stay tuned !
My point is the picture doesn’t show any details on the room or what’s outside the rack so it’s hard to know what’s optimal.
e.g. https://servermall.com/fr/sets/serveurs-blade-dell/?srsltid=...
Also keep in mind that these SLEDs are compatible both with the M2 & the M4. M2 require more surface area & cover the whole free space on the SLED
I'm also surprised they're touting the density of this solution— seems like the obvious thing would be to put the Minis on their sides. A 4U chassis has 17.5cm vertical space in it, and a Mac Mini is 17cm wide. With the Mini being 2in in height, that suggests 8-9 Minis in a 4U rack, vs 2 Minis in a 2U rack for Scaleway's trays.
EDIT: Here's a commercially-available solution that's 6/4U: https://www.mk1manufacturing.com/Rack-Mount-for-6-M4-mac-min..., you'd think it could basically be this but with the management plane behind or in front. And as others have said, making the management plane be more shared, so it's not 1:1.
https://cdn.prod.website-files.com/687e650a56916806eaaf8f62/...
You can find a bunch more detail in their related patent filings:
https://patents.google.com/patent/US20240397658A1/en?q=(%22m...
Sounds like 2 per 1U to me. I think it's power efficiency that limits filling a rack with typical 1U servers in most plans. The power efficiency here is probably not really impressive enough given so many power units, etc.
I'm pretty sure the drawers are 2U, but given that I don't know how they would get 96 nodes in a standard 42U rack.
I don't know if that's changed (they had odd pricing too, like Startup vs. Business, of which the difference wasn't clear), but aware. I hope someone has more success than I did.
It has been great. Good terraform provider and reliable service. I like their console, although the design feel very vaporwave to me.
Of course stuff can be better, but it is rapidly improving. The way they implemented grafana + user management was shit. But that’s fixed. Grafana still feels bolted on however. And login is a bit weird with their dedibox or whatever button next to the cloud offering. But no where near as confusing as aws is!
Also bumped on a bug in their terraform provider, found a related bug report, contributed some info and it was fixed within two weeks.
Quite happy so far. Running serverless sql, serverless containers. Secrets management and some iam config. No big stuff but quite sure it is capable to run a decently sized saas.
Glad to know you've got good experiences though.
I'll try to sign up again in a week or two.
There's an #apple-silicon channel there that the team behind the product monitors
With 60 minis per rack, and custom sled cases.
Going with AWS for cloud Mac Minis is the quickest way to lose a lot of money if you don't know what to do with it and to flush as much cash down the drain as quickly as possible.
20 years ago the company I worked for used a Mac-Mini for video transcoding, because there was some DRM issue we had to deal with, I don't remember the specifics.
https://www.apple.com/newsroom/in-the-loop/2025/10/shipping-...
I guess the Mac Pro Rack Edition technically exists, but that's needlessly huge (5U!) for a single node, ungodly expensive, and doesn't have server amenities like a BMC, so it's not exactly flying off the shelves.
What was your experience ?
From a user based perspective, using the remote desktop feature was a pain and I don't think any of our users were actually using it, the main use was : CI, AI training and bitcoin mining.
If their developer community grew a pair and made themselves heard, then maybe the billion dollar company would do their effing job and provide a proper rackable development platform.
Which they had many years ago, before they morphed into a company that builds telephones and furniture that occasionally functions as a computer.
Instead we pretend like extension cords and gluing Raspberry Pi's together is totally ok for professional purposes.
And actually the project was made to be faster on the market than AWS on their M1 offer, not to be industrialized at first.
They have the knowledge and it's existed in the past (Xserve) but they will never do it again because they design and sell furniture now.
Tell me they couldn't design a Mac Mini with holes for rack ears, built in BMC or serial console.... even if it was a specialty product people would pay a premium
Everything on this project is "hackie", it is not a solution for the future, the day Apple will change their license or break retro compatibility, the project will have no chance to survive.
So yes, the mac mini is absolutely not designed to be placed in a DC, a special product for that purpose would be better. So it's a risky bet from a technical point of view.
They technically still have the Rack Mac Pro, but that's such a half-assed offering that it's easy to forget it exists. It's a huge 5U chassis but if you look at the internals it's comical how little of the space is actually used for anything, because they kept the same layout as the Intel version which was designed to host multiple high-power GPUs, but dropped dGPU support in the Apple Silicon transition.
At 38 lbs and low starting price of $7,500 who can resist?
Unfortunately I would need another device active at all times to listen for commands and send that packet on demand. I was toying with the idea of using a Raspberry Pi connected to the router to perform that specific function, but I never got around to it.
Funny to see that idea scaled up to a server rack!
youngtaff•2mo ago
elcritch•2mo ago
Havoc•2mo ago
Don't think everything needed is included in the picture. Definitely additional cables for power and networking would need to be added
hinkley•2mo ago