hckrnws
https://www.jeffgeerling.com/blog/2025/minisforum-stuffs-ent...
> The strange CPU core layout is causing power problems; Radxa and Minisforum both told me Cix is working on power draw, and enabling features like ASPM. It seems like for stability, and to keep memory access working core to core, with the big.medium.little CPU core layout, Cix wants to keep the chip powered up pretty high. 14 to 17 watts idle is beyond even modern Intel and AMD!
Sigh. It's always something...
Nice device but the experience write up is more about distro choices than anything. It's quieter than the older units and it's harder to run 2 disk ssd raid because of some design choices. Is it faster? How many virtuals? What's the throughput if you use it for complex network related roles not offloaded to the microtik switching/routing kit?
Does FreeBSD work better?
Most likely not. This CPU has 3 types of cores. Heterogenous core is still a work in progress in FreeBSD and AFAIK they are targeting the Intel implementation first.
$599 seems like a lot to me. You can get numerous older, much more powerful Mini PCs (e.g older ThinkCentre Tiny series) or even a base brand new M4 Mac Mini for that kind of money.
Admittedly, the 10G interfaces and fast RAM make up for some of it, but at least for a normal homelab setup, I can't think of an application needing RAM faster than even DDR3, especially at this power level.
> even a base brand new M4 Mac Mini
A base Mac Mini (256GB/16GB) would cost me €720 while a Minisforum MS-R1 (1TB/32GB) would cost me €559 (minus a 25 euro discount for signing up to their newsletter if you accept that practice).
Price to performance the Apple solution may be better, but the prices aren't similar at all.
Upgrading the Mac to also feature 1TB of storage and 32GB of RAM, the price rises by a whopping €1000 to €1719.
Mac M4 have also the problem that you can't install whatever distro you want too, even if it is cheap. When they'll get out of support, there's no upgrade available anymore
Yes, I was basing my estimate more on the US price of $600, which is basically the same, if you buy both at MSRP. (However, that means the Mac does not have 10G)
I did not realize the EU versions were that much more expensive.
I do agree about the RAM/storage prices though. It's only worth it if you want the raw power, where the Mac handily beats this.
The MacMini will still be running in 5-7 years, reliably, and still have updates from Apple on the macOS side. It will still be running in 10+ years too should you keep it that long.
MinisForum makes disposable hardware. We used to use them for TV computers at work, and while they are cheap, they are fidgity with hardware and drivers, come with hacked-Windows Enterprise installed by default, and generally last for about 2 years before they hit the recycle pile.
559 vs 720? That's literally like a few coffees. I went to Amsterdam (assuming you're dutch) and I paid 5 euro for a coffee.
Go for the Mac Mini, the hardware incl thermal is also built exceptionally well. That's why you still have 20 year old Mac Minis still running as home servers etc.
If you're spending 170 euros on coffee then you're either abnormally rich or abnormally bad with money for a Dutchman.
Without the ability to upgrade either storage or RAM, a 256GB SSD with 16GB RAM is quite useless for a home server. Minisforum doesn't offer any options with that little RAM and storage it seems (you can pick between barebones and 1TB models).
The bare minimum spec for the Mac Mini sits at an interesting price point, but if you use it for any more than the bare minimum it'll be pretty restrictive with how memory-hungry macOS has become. No Linux support to speak of also makes for a rather mediocre home server experience.
One interesting part I found out of Apple's European pricing is that after currency conversion and subtracting VAT, the European price is still equivalent to $700, which is $100 more than they charge within the US. Looks like a 1/6th price increase is all you need for consumer rights!
I spend about ZAR 1200 (or 60 EUR per month) on coffee at home but who knows with all my cappuccinos. It's not really cheaper here in South Africa. But thanks, you made me look at my own coffee consumption now and it's always good to know!
Indeed macOS is a bit memory hungry but... unified memory, the sheer speed those chips can move data around is ridiculous. And macOS is a proper workstation Unix.
You're right - it's not ideal for headless. But there are ways. Still less painful than running Windows as as server.
my i5 7500t, 8gb ram 24/7 1l 'server' disagrees.
my atom, 4gb,1tb hdd bare metal ovh server also disagrees.
> That's why you still have 20 year old Mac Minis still running as home servers etc.
I often see statements like this made as if it's an exceptional characteristic of Macs. I've found that almost all computer hardware I buy has made it 20 years, though. Sure, a hard drive or something dies every once in a while, but most stuff gets retired because I just don't care to use it anymore, not because it doesn't work.
Exactly. Of all my hardware since 2003, which includes 5+ different GPUs that were mining and later training AI models almost non-stop the only things that stopped working and not just discarded for being too old/slow are 2 OCZ 2 SSDs which my guess would be had a bug in their firmware that caused a lockup.
> 559 vs 720? That's literally like a few coffees. I went to Amsterdam (assuming you're dutch) and I paid 5 euro for a coffee.
When someone says he drank a few coffees, I would never have guessed it was 32.
I feel like the "that's just a few coffees" metric is getting out of hand. By this metric, my current work laptop, purchased used from a local used reseller, was "a few coffees".
Also, I'm surprised how often on here I see people argue about price differences that are literally as I spend on entire computers.
And ironically, a special part failing and there being no replacement parts is more likely to happen on one of these NUCloids than a Mac mini.
So over the span of 20 years they’ll pay a multitude on these crappy computers than what the Mac mini costs once. May as well get a specced out Mac.
Isn't Apple also famous for not offering replacement parts other than replacing the whole thing and charging you accordingly?
In some cases yes, but having used Macs for decades and also working in companies with Macs for all the developers one thing is clear; these things don't easily break. Built extremely well.
(One exception being the GPU issues a few years back though on Intel MBP's)
But you can't run most Linux distros on Mac hardware without doing hacks
Depends on what you need it for. Love Mac minis but feature by feature the MS-R1 has more memory, ECC support and dual 10Gbe.
The ECC is cool. I don't need that for homelab servers though. But it's good to know and Minisforum is certainly a great offering.
Agreed. I can definitely see the Minisforum being far more cost efficient if you're mostly doing high speed networking transfers, while the Mac is more cost efficient if you need more raw power.
I just picked up a NAS - a ugreen dxp2800 - for £300. It has 2x nvme slots and 2x 3.5 bays. It’s x86 so if you don’t like the ugreen os you can change it.
It runs docker (supports docker compose) and vms and has the usual raid stuff.
They also do an arm version for half the price but I wanted the intel gpu for transcoding.
Does it take ECC RAM?
Not sure about the CPU performance being much more powerful for some shit-stained NUCs found on ebay, but one selling point for these minisforum machines are hassle-free dual 10G interfaces which are required for decent cluster performance - see ceph or proxmox ( with ceph ) or even kubernetes with, you guessed it - rook-ceph. Getting 10Gbit interface to work on ThinkCentre is possible, but not guaranteed to be reliable. This machine is perfect for such application and price point is not that terrible all things considered.
> Not sure about the CPU performance being much more powerful for some shit-stained NUCs found on ebay
The 10 GBit/s NUCs you find on eBay are enterprise-grade stuff: 10 Gbit/s hasn't really been a consumer thing. A used Fujitsu, Intel or Mellanox dual 10 Gbit/s bought on eBay isn't a "stained shit" that's "not guaranteed to be reliable". It's enterprise grade hardware.
(that said the machine in TFA looks nice)
I'd argue they're actually much better than some bargain-bin realtek cards, especially if you want raw performance.
They'll also probably work out of the box on whatever "server" distro you throw at them, which seems to be an issue with the machine in question.
However, GP was most likely talking about actual NUCs, not NICs. I mean, there wasn't a typo. The point probably being that the CPUs in some of those mini boxen are likely to be woefully undercooled, so performance may not be what you expect by just looking at the CPU model.
You can't get an ARM one though, only X86, which is mostly the point.
This is the only thing at a reasonable price with an Armv9.2-A CPU that is not a smartphone, but this Chinese CPU has various quirks.
An older but better ARM CPU with quadruple Cortex-A78 cores (Armv8.2-A ISA) is available for use in embedded computers from Qualcomm, rebranded from Snapdragon to Dragonwing. There are a few single-board computers of credit-card size with it, which are much faster than Raspberry Pi and the like.
Such SBCs are cheaper than the one from TFA and they are better for the purpose of software development.
The computer described in this article has the advantage of better I/O interfaces, the SoC has much more PCIe lanes, which allows the computer to have more and faster network interfaces.
If you want for an ARM computer to be a true high-throughput network server, then this one is the best choice. Nevertheless, for a true network server, a mini-PC with an Intel or AMD CPU will have a much, much better performance, at the same price or even at a lower price.
Using ARM is justifiable only for the purpose of software development, or if you want a smaller volume and a lower power consumption than achievable by a NUC-sized computer. For these purposes, one of the SBCs with Qualcomm QCM6490 is a better choice.
While a credit-card-sized SBC has only one Ethernet port, you can connect as many Ethernet interfaces as you desire to it (by using an USB hub and USB Ethernet interfaces), as long as the network throughput is not important and you just want to test some server software.
The Minisforum computer from the parent article has only 2 advantages for software development, the Armv9 ISA and being available with more memory, i.e. 32 GB or 64 GB, while the smaller ARM SBCs are available with 8, 12 or 16 GB.
Most people don't care about nominal difference in x86 vs arm. They care about cost, performance, efficiency, noise etc. Which applications run on the machine does matter.
The article never explained why the author wanted an ARM setup. I can only consider this a spiritual thing, just like how the author avoids Debian without providing any concrete explanations.
The usual reason to prefer ARM is efficiency, and the author's mention of replacing "power-hungry HPE towers" seeems to support that as a primary motivating factor.
This ARM computer has a much higher (3 to 4 times higher) idle power consumption than a mini-PC with an Intel or AMD CPU (e.g. an ASUS NUC), while having the same price and a much lower performance.
So in this case, the only valid reason to choose it is to have the ARM ISA for the purpose of software development.
This Chinese CPU is the only Armv9 CPU that is available in anything else than smartphones or expensive computers from Apple, Qualcomm or NVIDIA (or in even more expensive big servers). So there may be cases when it is desirable for software development, even if it has some quirks.
True. But as detailed in the Jeff Geerling article that was shared here in the comments, it has (at least at the moment) a rather high idle power draw, which seems to negate that, especially over time.
That is meaningful only if there is evidence to support that.
Mobile x86 processors used in mini PCs these days (as in 2026) are very competitive in terms of power efficiency. I wouldn't go for ARM just for that factor alone, especially without side-by-side comparisons of benchmarks.
> Most people don't care about nominal difference in x86 vs arm.
That's rubbish; even the people who don't care about ISA will care about stuff like power draw and software availability (although ironically arm seems distinctly worse in terms of power draw here).
But, I hope there are other people like me who will take a premium to avoid reading x86 core dumps, which is sort of like getting nails driven through your eyes. Yes, there's more software optimized for the chips; it is still bad code.
> Most people don't care about nominal difference in x86 vs arm.
"Most people" aren't on HN, either.
The # of ARM servers at cloud providers are growing, but the ARM server options are severely lacking for most.
I, personally, would like to see more ARM growth (and I think we're heading that direction anyway... look at NVIDIA right now). Buying ARM servers that help push ARM software development forward is probably a good thing, IMO, from that POV.
True. However, I've always noticed that ARM has less Linux support than x86, and the main benefits ARM is known for are typically performance/watt, running cooler, and less legacy support.
Since this server seems to have pretty average performance/watt and cooling, I can't really see much advantage to ARM here, at least for typical server use cases.
Unless you're doing ARM development, but I feel like a Pi 4/5 is better for basic development.
The performance is not average when compared with other ARM-based cheap computers, because it is high in comparison with them. It is also not average when compared to cheap Intel/AMD computers, because compared to such computers it is low. It could be called average only when averaging cheap ARM-based computers with cheap x86-64 based computers.
This computer uses 8 Cortex-A720 cores (and 4 little cores with negligible performance), which have a performance similar to the older Intel E-cores, i.e. Gracemont or Crestmont from Alder Lake, Raptor Lake or Meteor Lake. They are much slower than the recent Intel E-cores, i.e. Skymont or Darkmont, from Arrow Lake or Panther Lake.
So the performance of the whole CPU is similar to the 8-core Intel N300 (Alder Lake N) or Intel N350 (Twin Lake), which are found in various mini-PCs that are cheaper than this ARM computer.
Even so, the performance of this ARM CPU is many times greater than that of a Raspberry Pi and greater than of any cheaper ARM CPU. For greater performance, you must buy a more expensive smartphone, or a Qualcomm or Apple laptop or mini-PC, or a very expensive development computer from NVIDIA.
Linux support for ARM is inferior for end users of desktop 3rd party software. Everything else is provided by the repos. I doubt this person runs Signal or Spotify on those servers.
For those who don't need quite that much power I recently added an Orange Pi 5 to my own homelab, the RK3588 SoC packs an impressive punch for what it is
Similarly, a Beelink mini runs one of my Proxmox nodes and it's excellent. Literally sips power, too. I think I measured under 30w while under load. I mainly use it for my Plex instance given the N100 with QuickSync.
Does it run a mainline Linux kernel?
As of this past year (6.15+), most stuff you’d need for a regular desktop is upstreamed. Collabora has been working pretty hard on getting the chip mainlined, so it’s on a very good place compared to something like the Pi 5, which is not at all what the experience used to be in the past!
And despite their broadly similar performance, the RK3588s have significantly better power draw.
However I’m not sure of any of the rk3588 vendors that support both UEFI and have a full-size PCIe slot like the MS-R1 has.
RK3588 has many times better power consumption, but their performance is not at all similar.
8 Cortex-A720 vs. 4 Cortex-A76 means at least 3 times better performance for optimized programs.
Also for I/O throughput, this computer has far more fast PCIe lanes than RK3588, allowing many fast peripherals.
Funny, I just bought one of these last week. Agree with the article. Mine came with storage and Debian preinstalled. If you buy one from Amazon, keep an eye on price. I bought, then the next day the price dropped $150. Ordered another one and returned the expensive order.
You used to be able to just call Amazon and they'd refund the difference. Not sure if they do that anymore.
They did not. The rep told me to return it and buy another. :/
And fuck the planet, your time and the time of the logistics people.
This is so incredibly inefficient. Multiply by how many times this happens every day...
It's efficient ... for amazon ... because most people won't bother.
> Yes, while I use Fedora on my laptop, I also know Fedora is generally not a good option for a server.
Why is Fedora not considered good for a server?
It's a cutting-edge distro with 6-month release and 13-month support cycles.
Whereas Debian/Ubuntu have 5 years and RHEL/Alma/Rocky have 10 years.
I don't feel like this really answers the question thought, right? At least not at face value.
I could see the side of maintenance burden being a potential point, meaning that one would be "pushed" to update the system between releases more often than something else.
Typically you want stability and predictability in a server. A platform that has a long support lifecycle is often more attractive than one with a short lifecycle.
If you can stay on v12.x for 10 years versus having to upgrade yearly yo maintain support, that’s ideal. 12.x should always behave the same way with your app where-as every major version upgrade may have breaking changes.
Servers don’t need to change, typically. They’re not chasing those quick updates that we expect on desktops.
Yeah, and that's the take I assumed to hear based on what was said.
However, for something like ARM and the use case this particular device may have, in reality you would _want_ (my opinion) to be on a more rolling release distros to pick up the updates that make your system perform better.
I'd take a similar stance for devices that are built in a homelab for running LLMs.
Depends on what you're building an ARM system for. There are proper ARM servers out there; server work isn't the exclusive domain of x86, after all.
For homelabs, that's out the window. Do whatever you want/fits your needs best. This isn't the place where you'd likely find highly available networks, clustered or highly available services, UPS with battery banks, et. al.
I take it as no more than someone's personal opinion, since there is no reference provided whatsoever.
It's more maintenance due to its frequent release cycles, but it's perfectly good as a server OS. I've used it many times, friends use it.
You can't mess up the release cycle because their package repos drop old releases very quickly, so you're left stranded.
A friend recently converted his Fedora servers to RHEL10 because he has kids now and just doesn't have the time for the release cycle. So RHEL, or Debian, Alma, Rocky, offer a lot more stability and less maintenance requirement for people who have a life.
I'd also love to hear what folks have to say about this.
For myself I've had nothing but positive experiences running Fedora on my servers.
I think it's highly circumstantial. For example, my personal servers run a lot of FreeBSD and even though I could stay on major releases for a rather long time, I usually upgrade almost as soon as new releases are available.
For servers at work, I tried running Fedora. The idea was that it would be easier to have small, frequent updates rather than large, infrequent updates. Didn't work. App developers never had enough time to port their stuff to new releases of underpinning software, so we frequently had servers with unsupported OS version. Gave up and switched to RockyLinux. We're in the process of upgrading the Rocky8-based stuff to Rocky9. Rocky9 was released 2022.
Why is the power supply 2x larger than a Macbool Pro PS unit? Cheap? What about GaN?
Mac Mini/Studio has an integrated power supply, but other Mini PCs do not have the same luxury. It doesn't matter if you're Minisforum or HP.
Minisforum probably reused the x86 power supply for ARM. The x86 MS-01 and MS-A2 supports GPUs after all.
I'm not a hardware engineer, I've failed miserably in software engineering and now run a VPS host.
I was wondering why the PSU is half the size of the compute unit housing. 15 years ago, sure, but today it just seems cheap and lazy on part of whoever designed it.
Caveat: I'm frequently mistaken, always keen to learn and reduce the error between my perception and reality!
>I'm not a hardware engineer, I've failed miserably in software engineering and now run a VPS host.
I’m curious how hard hosting VPS as a business was to get off the ground? I’ve worked 5 years previously as a Linux sysadmin, but am getting pretty bored at my current job (administering Cisco VOIP systems). Think I’d rather go back to that
> but other Mini PCs do not have the same luxury
My Beelink Me Mini has an integrated PSU. Actually same with the EQR6 I got too.
I think that's why they were comparing to the MacBook Pro rather than the Mac Mini/Studio.
It's just cheap.
I have a personal ban on any hardware that isn't powered by USB-C. (Or if it's large I'll accept a C17 socket.) Either give me a GaN or I will get it myself.
Otherwise I'd probably have a few machines from this company.
This model accepts 100W USB PD input as well.
No ECC makes this a show stopper for me.
What speed is the PCIe slot? The Minisforum site talks about it being x8, but doesn't mention if its Gen4 or Gen5
slight tangent, but anyone had experience with running asahi on a m2 MacBook headless? I have a m2 air with a damaged screen id like to repurpose. Mostly want docker containers or something coolify adjacent
site down for me; here's an archive link:
Not sure I understand this distinction.
> I’ve always wanted an ARM server in my homelab. But earlier, I either had to use an underpowered ARM system, or use Asahi...
What is stopping you using Mac with MacOS?
It’s not great as a headless server.
With full disk encryption enabled you need a keyboard and display attached at boot to unlock it. You then need to sign in to your account to start services. You can use an IP based KVM but that’s another thing to manage.
If you use Docker, it runs in a vm instead of native.
With a Linux based ARM box you can use full disk encryption, use drop bear to ssh in on boot to unlock disks, native docker, ability to run proxmox etc.
Mac minis/studio have potential to be great low powered home servers but Apple is not going down that route for consumers. I’d be curious if they are using their own silicon and own server oriented distro internally for some things.
They fixed unlock in the last release:
"On a Mac with Apple silicon with macOS 26 or later, FileVault can be unlocked over SSH after a restart if Remote Login is turned on and a network connection is available."
https://support.apple.com/guide/security/managing-filevault-...
That's a bad implementation and seems like a bad afterthought
Thanks for the reply. I'm looking to replace my aging mini pc with a mac mini, so I'm quite interested in any limitations here.
The full disk encryption I can live without. I'm assuming these limitations don't apply if it's disabled. [Ah, I just saw the other reply that this has now been fixed]
I was aware of the Docker in a VM issue. I haven't tested this out yet, but my expectation is this can be mitigated via https://github.com/apple/container ?
I appreciate any insights here.
Apple containers run in one vm per container
Comment was deleted :(
The root of trust for Private Cloud Compute is our compute node: custom-built server hardware that brings the power and security of Apple silicon to the data center, with the same hardware security technologies used in iPhone, including the Secure Enclave and Secure Boot.
Granted, I don't know if it's really server oriented or if they're a bunch of iPhones on cards plugged into existing servers.
Most likely wanting to run Linux natively. Only M1/M2 can fill that role with Asahi, and still not with 100% hardware compatibility.
On the flip side, an M4 mini is cheaper, faster, much smaller (with built in power supply) and much more efficient. Plus for most applications, they can run in a Linux container just as well.
Thanks for the reply Jeff. This aligns with my understanding too. I'm close to purchasing a mac mini to replace my aging media pc. The core feature I want is to run microK8s natively, which I'm assuming the newish Mac containers will support.
anyone using proxmox on that?
> There is also one other perk: while the MSRP is $599, I got it for $559 despite a RAM shortage.
At that price, why not a mac mini running linux? I think (skimming Asahi docs) the only things that would give you trouble don't matter to the headless usecase here?
I don't think you can get an M1 or M2 mac mini with 64GB of RAM for $559
Oh, that's what I missed; you can get a Mac for that, but I missed memory. Yep, that's my mistake thank you for correcting me
Crafted by Rajat
Source Code