I just decided to give this a shot last night, I have no Idea how to use docker, so I have to figure that out this afternoon. Glad this video popped up, gives me hope
I've got a 2018 Core i3 Mac mini running as my home server. It runs Plex, an SMB file share, and a Linux VM. It also backs up my Plex Media Library and iCloud Photo Library to Backblaze. It's also a Time Machine Server.
You can put two 2.5" SSDs in a pre-2014 Mac Mini. I did that with my quad core i7 Mac Mini Server (yes, they made one for a couple years). I connected a larger storage unit to it via Thunderbolt 2 for additional storage, although SSDs are so cheap, I'm not sure I need it. I have 4TB internal storage for less than $160.
Same here - Late 2012 i7 Mini Server.. However, I have Ubuntu Server installed on it... When they come out with the M4 Mini, I will buy one for my daily use and relegate the M1 to my home server and take the i7 offline..
I think all these videos from linux bros are missing the biggest and easiest ways to fix all their problems. They just stop at "it's not like Linux." The host of this video is literally following linux instructions without doing any problem solving.
There's Asahi Linux that supports M1 to M3 Macs. It doesn't fully support every feature (especially on the laptops) but for a Mac Mini as a server, it *should* work fine.
Imo one of the best spaces for an m1 on a rack is its local LLM capability. Since ram is basically GPU memory, you get a really energy efficient llm machine that you can set up a reverse proxy and call anytime you want with a lightweight server.
I know that this video is one year old. But, using docker is very simple on macOS and instead of installing things like PlexMedia, JellyFin, or Sonar using binary, just use like you would on Linux using a docker image which will work like a charm.
I have been thinking about (I haven't tried it) to get a Mac Mini with 10 gig ethernet and connect it to Asustor Flashstor 12 with 10 gig ethernet (would this work with a plain CAT6E, crossover or would a hub be required?). The Asustor sets up RAID using SSDs and the Mac Mini as a server consumes it. The server application could be Plex, PostgreSQL, PostGIS or JupyterHub or another server app using Docker. To connect to the LAN use a dongle off the Mac Mini. One would have a powerful Mac Mini with a disk farm with a potentially huge capacity. Alternatively, one could use Flashstor 6 if there was a way to directly connect the 3.2 USB connections on each machine, but which is apparently not possible without ethernet (?). Avoiding 10 gbs ethernet would save a lot of money on both ends.
That would probably work, but why? All of your proposed server roles would run just as well if not better on the Asustor box as on the Mac. Save yourself a bunch of headache and cut the Mac out of the setup entirely. Or get one of those Thunderbolt to PCIe boxes and put a PCIe x16 to 4x m.2 carrier cards in it. Connect a bunch of super fast storage directly to the Mac to use for your server needs without involving another computer.
17:00 NUC like mini pcs do that at least this well, even with the same power draw. Intel QuickSync is really good and on 12th Gen, I've seen it pull over 20 1080p transcodes. 4k h.264 isn't that hard to transcode anymore. The battle test is transcoding 4k HEVC with HDR Tonemapping.
Seems like the best you can hope for is to run some arm based linux os in qemu and do all your nitty gritty stuff there, not sure what the networking situation would be like though. Though i'm guessing the idea here is to see much much macos can do out of the box.
Hi sir, this video has been extremely informative and you saved my a hassle in finding the software I needs. Thank you! Can you please tell me how did the system go like the power consumption and the ram configuration? Thank you sir🙏🏻🙏🏻🙏🏻!!!
What people don't see is that the M series mac minis are really energy efficient for what they can do You can run pretty demanding calculations while consuming only 15-20 watts and this 24/7, try this with a raspberry or any other mini PC
UTM (or well basically QEMU) is a thing. You can just run arm64 vm’s on there, or emulate amd64, but you shouldn’t do that because it’s so inefficient. So then just take a arm64 image from linux and run it as a vm.
I haven't heard of UTM when this video came out, but Its nice to use on my ARM Macs. I wish UTM and QEMU+KVM had better virtual graphics support without GPU pass-through for better desktop use, but thats a wish for the QEMU team that I've had for a while.
@@ElectronicsWizardry sadly enough there is no KVM on macOS, just QEMU. I understand why tho, if you want to stuff like gaming on the vm. Although I think on UTM the team added some more options than just standard QEMU. Also GPU passthrough would’t really work right now since there are no graphics drivers from Apple on Linux or Windows. The best you can do on Linux is openGL. Why isn’t there KVM in QEMU in this case? KVM isn’t available on macOS because it’s a Linux-specific technology that enables faster virtualization by allowing direct hardware access. On macOS, QEMU uses Apple’s Hypervisor.framework instead, which doesn’t offer the same level of low-level access as KVM. (This text is refined by chatGPT since I suck explaining these things in English and whilst typing).
I was thinking of something kinda different for my GPU thoughts. I want a virtual GPU that uses the host GPU and has a decent feature set, like like Vmware and Parallels have. Then I can have things like 3d acceleration in my guest VMs without setting up GPU passthough, vGPU or other tech for basic VMs. Its likely a niche want, but I like seeing good GPU performance when setting up test VMs. I know they have tried with stuff like virgl, but its never worked well for me.
@@ElectronicsWizardry I believe some GPUs in QEMU support 3D acceleration, but I’m not entirely sure. I mostly use VMs as servers, where a GPU typically isn’t needed (unless you’re working with AI or something GPU-intensive). As for GPU passthrough, you can’t passthrough the M1 GPU because there are no drivers for it, and I’m not sure how macOS would handle running without a connected GPU.
What would be nice would be for a group of Mac experts, like the group that does Open Core Patcher, to figure out how to run modern MacOS without a GUI. This would save a lot of memory. I wonder what the technical reasons are that block it from being implemented now.
That might be interesting to see for some uses cases and I’m curious how much it would help. I’d guess the gui doesn’t need that much ram as it could be swapped out if other programs are needed.
@@ElectronicsWizardryI'm speculating, but I would think in this case the memory is never really freed up. It remains used and unavailable for use as part of the buffer cache. On all computers, especially servers, you never want to see any actually free memory. If the programs being run don't need it, then it should be part of the buffer cache (and maybe others).
Agreed on using all the ram for cache as free ram is wasted ram. My thought was the ram used by the gui could possibly be swapped out without a performance penalty if the system was only accessed by the command line.
All I want is to use mac mini as file sharing, aka cloud service instead of paying monthly for all these clouds service. How to do this setup for as less work as possible, can we use external ssd and be awake continuously?
There are programs like nextcloud that make a cloud service like interface. Next cloud will be a bit of a pain to install on Mac OS but should be possible with php and a database setup. I’d try to use a different I if you can but it’s probably doable on a Mac. I don’t see an issue running these systems 24/7.
It's about the easiest thing I've had to do on a computer. I didn't even need to watch any youtube videos to figure this out, and I'm barely computer literate. I just went into System settings and searched file sharing, then followed simple instructions from there. Now I have all of my spare hard drives (2 SSDs and 2 HDDs for a total of 12TB) accessible by any of my laptops on the network 24/7. Also, screen sharing is kind of fun, too.
Got M1 mini, but using it as 24/7 server I'd find it lacking, especially since Mac RAM is so insanely priced. I got cheap 1L Lenovo i5-6500t model and with both internal slots and few CT's and VM's its pulling about 9-14w on almost idle and on moderate loads using like W10 on RDP something like 20W on average from the wall. Perfect for noob like me :)
I've been made aware that the M1 chip is pretty good for AI stuf like chat gpt and probably more. That kind of server interests me a little more just because their chip seems to be designed to handle it.
I wouldn't say that they are great for that use case personally. They are significantly better than the competition (at least what was out there 6 months ago and specially for the price). But you really do need a dedicated and expensive system for ai stuff to run locally. At the moment, I would recommend something which will cost around £2500 as the entry level and £3000-£3500 for a reasonable system. This system will include either [one nvidia 4080/4090 graphics card] or [two nvidia 4080/4090 graphic cards]. The vram jump to the 4090 is significant and important for current ai execution. I haven't built a system like that myself for ai and you will need to do further research into which graphics card to get and so on as one of my recommendations may not be suitable. Caution: please do not accept the word of most youtubers on the topic of AI in general. They know very little and often are either too excited or pessimistic. Both of which will mislead you and result in failure for the goal you had in mind. Unless you have a specific need/want/use-case or have a lot of money, I would stay away from running ai locally. The models available for offline and local use are not as good as the ones available via your web browser and that does matter. Also, the speed of running the prompts matters a lot more than we may think.We're humans, not computers. One notification on our phone or one phone call can wreak havoc on our productivity for the next few hours or days. One dialogue box opening at the wrong time and our idea is gone. We can't put things on hold and then resume them at will. I say these things because the chatgpt website has had a recent update which makes it look really nice and the 4-o version is really quick. And that matters a lot more than we think. That is the reason why we spend £1000 on a laptop as opposed to a £300 on a laptop even if we are able to perform our task on either machine. Because one machine will result in you doing the task, whereas the other one will have you leaving it at home, procrastinating, complaining, going off on tangents while being aware or unaware of why/what you're doing. So the local machine may not be that useful as the current software available for it is far from polished and unless you buy the best of the best hardware, the performance will hinder your chances of success when it comes to the system providing usefulness/value to you (in exchange for your investment). It would all still be worth it if the local models available were equally as good as the online ones or better but unfortunately, even if benchmarks say otherwise, assume that you are using chatgpt 3.5 if you're using offline models. Altogether, it's not worth it at the moment (unless you have a specific use case which has been shown to be useful (more than proof of concept)). RAG and agents are tempting but remember that things are simpler than they seem and the current execution, therefore, isn't great. If it was, there would be one singular polished application which would do it all for you seamlessly. The fact that there isn't is due to the fact that any such app would prove itself quite useless at the moment tbh. From my testing anyway. But it's good for developers and individuals like that to invest their time and money into. I tried quite hard to get a local system running but the models available were simply not good enough and I didn't want to invest £4000 into an additional system considering the limitations of those models. Privacy is the main reason I was interested in it. The second one being an uncensored and less aligned version (less aligned to the current alignment/censorship-goal). The censorship hurts the models massively in my opinion and the fake optimistic/positive tone of voice (and attitude) that the AI is limited to currently, renders it useless. Just as useless as a dishonest friend. Because that's what it is. It's not helping us, it's fulfilling someone else's wishes and using us in the process. Because it's doing it to us while giving us no indication of the fact that it is doing it. If the AI can't in a non-joking way call me an idiot, its usefulness is limited. Hopefully we will be able to get some ai chips or accelerator all in one systems which will provide excellent ai execution for low tdp and cost. But those aren't currently available to my knowledge and are also legally restricted for major/open distribution, I believe, due to the AI threat. I'm not an expert, this has been my limited experience so far. Bear in mind you will be able to find tutorials on how to do it, but not real examples of how the youtuber found it useful in their case with the entire example shown.
BTW, I could use homeassistant with UTM (free VM), I did a video about that a few months ago th-cam.com/video/1-rWFAuz8ls/w-d-xo.html, and regarding docker, I also use it but docker for macos is really limited! no USB passthrough, no macVLAN ...
Alot of the issues you ran into boiled down to you not knowing how to use docker... Honestly I was more curious how mac would handle being a nas. Personally I don't use samba anymore so this wasn't very helpful For context, im a Linux sysadmin/devops guy. So im pretty familiar with containers and happen daily drive an M1 mac for personal use.
Why would anyone waste money on a Mac Mini as a home server when you can buy a way more powerful, smaller, lower wattage mini-PC from Minis Forum or Beelink that destroys the Mac for half the price? The UM773 with 2TB of PCIe4.0 nvme storage and 32 GB of DDR5 RAM that destroys the Mac Mini in nearly all benchmarks can be had for under $600 including Windows 11 Pro license. For a home server, there is no reason to buy a Mac when the same function can be provided much cheaper & faster by competitors. Plus the Mac Mini is HUGE compared to the mini PCs put out by Minis Forum and Beelink.
Just because he can! It's not always about the money but about the possibilities and what resources you current have! Maybe he doesn't have any use for that M1 mini PC and doesn't want to sell it for whatever reason so he's gonna use it as a home server for the time being.
its probably because you have one lying around. I bought a mac mini m1 for £300 from ebay. Insane deal. If I want to use that as my home server (looking into it at the moment), then that's perfect. It's silent and low tdp with really high cpu and gpu performance for price and tdp. Plus the form factor. Not to mention, if you use alternative software to the ones shown in this video, it can be really awesome. For example, using the built in mac os network sharing which let's me easily access the mac mini and all the drives attached to it automatically. I can also access remote desktop and control the computer wirelessly while on the same network (at home) and assign tasks for it to do (as to save my laptop battery + keep it cool). That's all really intuitive and requires next to no set up. There are also really nice apps which like infuse and hazel, which you can look into, that provide awesome functionality without investing a huge amount of time into them.
Thanks for doing this video, i have interest because my brother is deep into the Cult of Apple. All other forms of computing he has banned. Apple really did make some great CPU for these systems. Im hoping other vendors will be coming out with similar cpus for Linux.
In my opinion, home server (mostly idle), I only need cheap one but fast. I got Beelink under $200, using SSD and samba, thats all. Raspbeey pi 4B with 1GB Nvme USB, for fast PLex Media Server: th-cam.com/video/mhqZH7sBdBM/w-d-xo.html
just build a hackintosh and have something 2x as powerful as a mac pro for less than half the price - even if you don't do this it is good to be aware of since this is the value proposition you are walking into - it is unambiguous and clearcut - apple sucks not just for servers but generally and all around
No this is not a security feature its a "this is not your computer thingie". At some point theyre gonna lock down macos as they do with ios. Apple sucks. I relly hope for ashi linux to save the mac
I just decided to give this a shot last night, I have no Idea how to use docker, so I have to figure that out this afternoon.
Glad this video popped up, gives me hope
I've got a 2018 Core i3 Mac mini running as my home server. It runs Plex, an SMB file share, and a Linux VM. It also backs up my Plex Media Library and iCloud Photo Library to Backblaze. It's also a Time Machine Server.
I'd be really interested to know how you did this... can you run a Linux VM with the M1 version too?
You can put two 2.5" SSDs in a pre-2014 Mac Mini. I did that with my quad core i7 Mac Mini Server (yes, they made one for a couple years). I connected a larger storage unit to it via Thunderbolt 2 for additional storage, although SSDs are so cheap, I'm not sure I need it. I have 4TB internal storage for less than $160.
This is always a good idea. I have an M1 I use my 2012 i7 Mac mini with twin SSD as a server works so well.
Same here - Late 2012 i7 Mini Server.. However, I have Ubuntu Server installed on it... When they come out with the M4 Mini, I will buy one for my daily use and relegate the M1 to my home server and take the i7 offline..
I think all these videos from linux bros are missing the biggest and easiest ways to fix all their problems. They just stop at "it's not like Linux." The host of this video is literally following linux instructions without doing any problem solving.
I vote for re-testing this one as soon as Linux fully supports M1 chips (aside from graphics, not really needed to test basic server needs)
There's Asahi Linux that supports M1 to M3 Macs.
It doesn't fully support every feature (especially on the laptops) but for a Mac Mini as a server, it *should* work fine.
Imo one of the best spaces for an m1 on a rack is its local LLM capability. Since ram is basically GPU memory, you get a really energy efficient llm machine that you can set up a reverse proxy and call anytime you want with a lightweight server.
I know that this video is one year old. But, using docker is very simple on macOS and instead of installing things like PlexMedia, JellyFin, or Sonar using binary, just use like you would on Linux using a docker image which will work like a charm.
Really interesting
Thanks for pursuing this topic
Love how through you are, keep up the videos man!
You could try Colima as a Docker Desktop replacement. It should perform better and it's also more simple.
Great video! I think I'm going to grab one of these and plug it into the living room tv along with all my hard drives
This system is being used as a home theater pc now and works great at that if you like macOS.
great job, I love your content!
this is so much easier to understand than Proxmox
I have been thinking about (I haven't tried it) to get a Mac Mini with 10 gig ethernet and connect it to Asustor Flashstor 12 with 10 gig ethernet (would this work with a plain CAT6E, crossover or would a hub be required?). The Asustor sets up RAID using SSDs and the Mac Mini as a server consumes it. The server application could be Plex, PostgreSQL, PostGIS or JupyterHub or another server app using Docker. To connect to the LAN use a dongle off the Mac Mini. One would have a powerful Mac Mini with a disk farm with a potentially huge capacity. Alternatively, one could use Flashstor 6 if there was a way to directly connect the 3.2 USB connections on each machine, but which is apparently not possible without ethernet (?). Avoiding 10 gbs ethernet would save a lot of money on both ends.
That would probably work, but why? All of your proposed server roles would run just as well if not better on the Asustor box as on the Mac. Save yourself a bunch of headache and cut the Mac out of the setup entirely.
Or get one of those Thunderbolt to PCIe boxes and put a PCIe x16 to 4x m.2 carrier cards in it. Connect a bunch of super fast storage directly to the Mac to use for your server needs without involving another computer.
17:00 NUC like mini pcs do that at least this well, even with the same power draw. Intel QuickSync is really good and on 12th Gen, I've seen it pull over 20 1080p transcodes. 4k h.264 isn't that hard to transcode anymore. The battle test is transcoding 4k HEVC with HDR Tonemapping.
Thinking about going this route for Homebridge primarily
❤my new favorite channel! cheers bro
I bought two Mac Minis from 2012 each to 100,-€, put in 2 TB ssd, 16 GB RAM und ubuntu Server. Each machine runs with 8 Watt in idle.
I have used it since launch and had zero issues with USB HDD bay
Thanks for the useful information
Seems like the best you can hope for is to run some arm based linux os in qemu and do all your nitty gritty stuff there, not sure what the networking situation would be like though. Though i'm guessing the idea here is to see much much macos can do out of the box.
I imagined that, I'd like to see the same review with Asahi Linux
Yea I’ll give that a try. I have heard many good things about the progress being made with Ahasi linux
Hi sir, this video has been extremely informative and you saved my a hassle in finding the software I needs. Thank you! Can you please tell me how did the system go like the power consumption and the ram configuration? Thank you sir🙏🏻🙏🏻🙏🏻!!!
What people don't see is that the M series mac minis are really energy efficient for what they can do
You can run pretty demanding calculations while consuming only 15-20 watts and this 24/7, try this with a raspberry or any other mini PC
Excellent video!
UTM (or well basically QEMU) is a thing. You can just run arm64 vm’s on there, or emulate amd64, but you shouldn’t do that because it’s so inefficient. So then just take a arm64 image from linux and run it as a vm.
I haven't heard of UTM when this video came out, but Its nice to use on my ARM Macs. I wish UTM and QEMU+KVM had better virtual graphics support without GPU pass-through for better desktop use, but thats a wish for the QEMU team that I've had for a while.
@@ElectronicsWizardry sadly enough there is no KVM on macOS, just QEMU. I understand why tho, if you want to stuff like gaming on the vm. Although I think on UTM the team added some more options than just standard QEMU.
Also GPU passthrough would’t really work right now since there are no graphics drivers from Apple on Linux or Windows. The best you can do on Linux is openGL.
Why isn’t there KVM in QEMU in this case? KVM isn’t available on macOS because it’s a Linux-specific technology that enables faster virtualization by allowing direct hardware access. On macOS, QEMU uses Apple’s Hypervisor.framework instead, which doesn’t offer the same level of low-level access as KVM. (This text is refined by chatGPT since I suck explaining these things in English and whilst typing).
I was thinking of something kinda different for my GPU thoughts. I want a virtual GPU that uses the host GPU and has a decent feature set, like like Vmware and Parallels have. Then I can have things like 3d acceleration in my guest VMs without setting up GPU passthough, vGPU or other tech for basic VMs. Its likely a niche want, but I like seeing good GPU performance when setting up test VMs. I know they have tried with stuff like virgl, but its never worked well for me.
@@ElectronicsWizardry I believe some GPUs in QEMU support 3D acceleration, but I’m not entirely sure. I mostly use VMs as servers, where a GPU typically isn’t needed (unless you’re working with AI or something GPU-intensive). As for GPU passthrough, you can’t passthrough the M1 GPU because there are no drivers for it, and I’m not sure how macOS would handle running without a connected GPU.
What would be nice would be for a group of Mac experts, like the group that does Open Core Patcher, to figure out how to run modern MacOS without a GUI. This would save a lot of memory. I wonder what the technical reasons are that block it from being implemented now.
That might be interesting to see for some uses cases and I’m curious how much it would help. I’d guess the gui doesn’t need that much ram as it could be swapped out if other programs are needed.
@@ElectronicsWizardryI'm speculating, but I would think in this case the memory is never really freed up. It remains used and unavailable for use as part of the buffer cache. On all computers, especially servers, you never want to see any actually free memory. If the programs being run don't need it, then it should be part of the buffer cache (and maybe others).
Agreed on using all the ram for cache as free ram is wasted ram. My thought was the ram used by the gui could possibly be swapped out without a performance penalty if the system was only accessed by the command line.
All I want is to use mac mini as file sharing, aka cloud service instead of paying monthly for all these clouds service. How to do this setup for as less work as possible, can we use external ssd and be awake continuously?
There are programs like nextcloud that make a cloud service like interface. Next cloud will be a bit of a pain to install on Mac OS but should be possible with php and a database setup. I’d try to use a different I if you can but it’s probably doable on a Mac. I don’t see an issue running these systems 24/7.
@@ElectronicsWizardry thanks
It's about the easiest thing I've had to do on a computer. I didn't even need to watch any youtube videos to figure this out, and I'm barely computer literate. I just went into System settings and searched file sharing, then followed simple instructions from there. Now I have all of my spare hard drives (2 SSDs and 2 HDDs for a total of 12TB) accessible by any of my laptops on the network 24/7. Also, screen sharing is kind of fun, too.
@@lawconic thank you
Oh, on quick glance of your thumbnail, I thought you were going to toss it in the washer.
Now THAT would be different.
Got M1 mini, but using it as 24/7 server I'd find it lacking, especially since Mac RAM is so insanely priced. I got cheap 1L Lenovo i5-6500t model and with both internal slots and few CT's and VM's its pulling about 9-14w on almost idle and on moderate loads using like W10 on RDP something like 20W on average from the wall. Perfect for noob like me :)
I've been made aware that the M1 chip is pretty good for AI stuf like chat gpt and probably more. That kind of server interests me a little more just because their chip seems to be designed to handle it.
I wouldn't say that they are great for that use case personally. They are significantly better than the competition (at least what was out there 6 months ago and specially for the price). But you really do need a dedicated and expensive system for ai stuff to run locally. At the moment, I would recommend something which will cost around £2500 as the entry level and £3000-£3500 for a reasonable system. This system will include either [one nvidia 4080/4090 graphics card] or [two nvidia 4080/4090 graphic cards]. The vram jump to the 4090 is significant and important for current ai execution. I haven't built a system like that myself for ai and you will need to do further research into which graphics card to get and so on as one of my recommendations may not be suitable. Caution: please do not accept the word of most youtubers on the topic of AI in general. They know very little and often are either too excited or pessimistic. Both of which will mislead you and result in failure for the goal you had in mind. Unless you have a specific need/want/use-case or have a lot of money, I would stay away from running ai locally. The models available for offline and local use are not as good as the ones available via your web browser and that does matter. Also, the speed of running the prompts matters a lot more than we may think.We're humans, not computers. One notification on our phone or one phone call can wreak havoc on our productivity for the next few hours or days. One dialogue box opening at the wrong time and our idea is gone. We can't put things on hold and then resume them at will. I say these things because the chatgpt website has had a recent update which makes it look really nice and the 4-o version is really quick. And that matters a lot more than we think. That is the reason why we spend £1000 on a laptop as opposed to a £300 on a laptop even if we are able to perform our task on either machine. Because one machine will result in you doing the task, whereas the other one will have you leaving it at home, procrastinating, complaining, going off on tangents while being aware or unaware of why/what you're doing. So the local machine may not be that useful as the current software available for it is far from polished and unless you buy the best of the best hardware, the performance will hinder your chances of success when it comes to the system providing usefulness/value to you (in exchange for your investment). It would all still be worth it if the local models available were equally as good as the online ones or better but unfortunately, even if benchmarks say otherwise, assume that you are using chatgpt 3.5 if you're using offline models. Altogether, it's not worth it at the moment (unless you have a specific use case which has been shown to be useful (more than proof of concept)). RAG and agents are tempting but remember that things are simpler than they seem and the current execution, therefore, isn't great. If it was, there would be one singular polished application which would do it all for you seamlessly. The fact that there isn't is due to the fact that any such app would prove itself quite useless at the moment tbh. From my testing anyway. But it's good for developers and individuals like that to invest their time and money into. I tried quite hard to get a local system running but the models available were simply not good enough and I didn't want to invest £4000 into an additional system considering the limitations of those models. Privacy is the main reason I was interested in it. The second one being an uncensored and less aligned version (less aligned to the current alignment/censorship-goal). The censorship hurts the models massively in my opinion and the fake optimistic/positive tone of voice (and attitude) that the AI is limited to currently, renders it useless. Just as useless as a dishonest friend. Because that's what it is. It's not helping us, it's fulfilling someone else's wishes and using us in the process. Because it's doing it to us while giving us no indication of the fact that it is doing it. If the AI can't in a non-joking way call me an idiot, its usefulness is limited. Hopefully we will be able to get some ai chips or accelerator all in one systems which will provide excellent ai execution for low tdp and cost. But those aren't currently available to my knowledge and are also legally restricted for major/open distribution, I believe, due to the AI threat. I'm not an expert, this has been my limited experience so far. Bear in mind you will be able to find tutorials on how to do it, but not real examples of how the youtuber found it useful in their case with the entire example shown.
Is it possible to use remote management and use the mac from a linux pc over the network?
I think the remove management in mac os is VNC or ssh, so it should be simple to remote into the macOS box with a linux system.
BTW, I could use homeassistant with UTM (free VM), I did a video about that a few months ago th-cam.com/video/1-rWFAuz8ls/w-d-xo.html, and regarding docker, I also use it but docker for macos is really limited! no USB passthrough, no macVLAN ...
Apple have listened to you. Apple have introduced an update Mac mini M4 with a redesigned chassis.
I think the new design is generally a big improvement. Looking forward to seeing how it performs.
Alot of the issues you ran into boiled down to you not knowing how to use docker... Honestly I was more curious how mac would handle being a nas. Personally I don't use samba anymore so this wasn't very helpful
For context, im a Linux sysadmin/devops guy. So im pretty familiar with containers and happen daily drive an M1 mac for personal use.
Why would anyone waste money on a Mac Mini as a home server when you can buy a way more powerful, smaller, lower wattage mini-PC from Minis Forum or Beelink that destroys the Mac for half the price? The UM773 with 2TB of PCIe4.0 nvme storage and 32 GB of DDR5 RAM that destroys the Mac Mini in nearly all benchmarks can be had for under $600 including Windows 11 Pro license.
For a home server, there is no reason to buy a Mac when the same function can be provided much cheaper & faster by competitors. Plus the Mac Mini is HUGE compared to the mini PCs put out by Minis Forum and Beelink.
Just because he can! It's not always about the money but about the possibilities and what resources you current have! Maybe he doesn't have any use for that M1 mini PC and doesn't want to sell it for whatever reason so he's gonna use it as a home server for the time being.
its probably because you have one lying around. I bought a mac mini m1 for £300 from ebay. Insane deal. If I want to use that as my home server (looking into it at the moment), then that's perfect. It's silent and low tdp with really high cpu and gpu performance for price and tdp. Plus the form factor. Not to mention, if you use alternative software to the ones shown in this video, it can be really awesome. For example, using the built in mac os network sharing which let's me easily access the mac mini and all the drives attached to it automatically. I can also access remote desktop and control the computer wirelessly while on the same network (at home) and assign tasks for it to do (as to save my laptop battery + keep it cool). That's all really intuitive and requires next to no set up. There are also really nice apps which like infuse and hazel, which you can look into, that provide awesome functionality without investing a huge amount of time into them.
Thanks for doing this video, i have interest because my brother is deep into the Cult of Apple. All other forms of computing he has banned. Apple really did make some great CPU for these systems. Im hoping other vendors will be coming out with similar cpus for Linux.
In my opinion, home server (mostly idle), I only need cheap one but fast. I got Beelink under $200, using SSD and samba, thats all. Raspbeey pi 4B with 1GB Nvme USB, for fast PLex Media Server: th-cam.com/video/mhqZH7sBdBM/w-d-xo.html
"Taking a loot" or "Taking a look"?
Ooops. Thanks. It’s been fixed now.
just build a hackintosh and have something 2x as powerful as a mac pro for less than half the price - even if you don't do this it is good to be aware of since this is the value proposition you are walking into - it is unambiguous and clearcut - apple sucks not just for servers but generally and all around
Thanks all the way up on this video. This was very helpful. I will have to watch this a few times and take notes.
No this is not a security feature its a "this is not your computer thingie". At some point theyre gonna lock down macos as they do with ios.
Apple sucks.
I relly hope for ashi linux to save the mac