Just Install it on the Host
We dig into the NVIDIA RTX Spark and what it means for local AI, revisit the Proxmox-vs-Unraid debate through Alex's Perfect Media Server Part 3, and do a deep dive on Jellyfin's strengths, rough edges, and the upcoming version 12.
What we cover
- The NVIDIA RTX Spark announcement at Computex and NVIDIA’s plans to own both the data center and the desk, as well as the local LLM space in general.
- Alex’s Perfect Media Server Part 3 video and whether Unraid or TrueNAS….or OpenMediaVault are potentially the better choice.
- Jellyfin June update, and why running Jellyfin, Navidrome, and AudioBookshelf as separate containers might be better than chasing a single holistic solution like Plex.
First up, we cover NVIDIA’s announcement at Computex of the NVIDIA RTX Spark; a laptop platform combining a Blackwell GPU with an ARM CPU and 128GB of unified memory. Adam walks us through the hardware, we debate the MacBook Pro parallels, and Alex lays out what he thinks the announcement really signals: NVIDIA wants to own the edge just as much as the cloud. That conversation spills into local AI for the enterprise, with Stephen explaining what it actually looks like to roll out AI tools in SMB environments, and Geoff puts on his lawyer hat (yet again) and shames Microsoft for their horribly complicated terms of service.
Next, Alex tells us about the release of Version 3 of his Perfect Media Server Part 3 series video. We go into the core design choices (and why some viewers were upset about it). Adam makes the honest case that Unraid would have handled almost all of it without any CLI, and the panel hashes out the real question: is building the hard way still worth it when perfectly good managed alternatives exist?
Finally, we do a discuss how Jellyfin June is going with Alex. We talk about what Jellyfin genuinely does well, where it still has rough edges, and how a stack of focused containers (Jellyfin for video, Navidrome for music, AudioBookshelf for audiobooks) might be a better answer than any single holistic platform.
Topics: NVIDIA RTX Spark, Computex, Windows on ARM, local AI, LLMs, Microsoft Copilot, data privacy, Proxmox, MergerFS, SnapRAID, Unraid, TrueNAS, Docker, Ansible, Jellyfin, Navidrome, AudioBookshelf, Plex, Tailscale, self-hosted media
Links
- Nvidia
- Perfect Media Server 3.0
- Jellyfin June Check-in
Transcript
Alex: Hey, welcome into another episode of Bitflip, the show where we take a look at the pragmatic side of infrastructure and the wider conversation around technology. I’m your host, Alex. 00:00
Adam: I’m Adam. 00:08
Stephen: I’m Steven. 00:09
Geoff: And I’m Geoff. 00:11
Alex: So first up today, we’re going to dive into some Computex news. NVIDIA had a busy week last week talking about local AI and the NVIDIA RTX Spark. Which one of you added this to the doc? Was it you, Adam? 00:12
Adam: That was me, yeah. I found this one 00:24
Alex: All 00:25
Adam: pretty 00:25
Alex: right. 00:25
Adam: interesting. So what they actually announced here was a laptop version of this little sucker. So they’ve been behind the scenes working with Microsoft on the, what did they call it? The RTX Spark, which is basically a Nvidia CPU, 20 cores, ARM-based cores. I think they pulled that from MediaTek. I think they partnered with them on the 00:26
Geoff: Yes. 00:49
Adam: actual core design for the CPU. And then on the RTX side, it’s Blackwell for the GPU. And from what I’ve been reading, it’s an equivalent of like a 5700 as far as performance on that side. So it’s a beefy little sucker. 00:49
Alex: A 5700 watt though. 01:05
Geoff: Do 01:07
Adam: Oh, 01:07
Geoff: you mean 50-70? 01:07
Adam: yeah. Oh, did I say 50? Okay, yeah, it’s a 57. It’s a 5070, like the RTX 5070. 01:09
Alex: Oh, wow. 01:14
Adam: Mobile, 01:16
Stephen: That’s quite the 01:16
Alex: That’s 01:17
Stephen: thing. 01:17
Alex: not 01:17
Adam: just 01:17
Alex: bad 01:17
Adam: to be clear. 01:17
Alex: at all. Oh, okay. 01:17
Adam: Yeah, 01:19
Geoff: But 01:20
Adam: yeah. 01:20
Geoff: you missed the best part, 128 gigabytes of unified memory. 01:20
Alex: Yeah. This is NVIDIA making a statement. I mean, actually, AMD made the statement last year with the Strix platform. And so actually, I think this is probably NVIDIA’s way of trying to play catch up. But what’s most interesting about this for me is that it’s not just a GPU. It’s also a CPU-capable architecture, like it can do normal instruction sets and run a normal kind of ARM-based operating system. 01:25
Adam: And to that point, actually, the way that I frame it is that it’s very much, I think, trying to compete or one up the Mac kind of unified memory system on a chip type of architecture. Yeah. 01:55
Alex: That’s true. That’s where it started, isn’t it? It was 02:10
Adam: Yeah. 02:12
Alex: the M1, the Apple Silicon M1 chip. 02:12
Adam: Yeah, I think 02:15
Alex: It 02:15
Adam: that’s 02:15
Alex: was too 02:15
Adam: the closest 02:15
Alex: powerful. 02:15
Adam: parallel, I would say. And what’s to be determined still is how is the battery life going to look on this machine. I’ve heard already, though, when they’re putting it under some performance stress testing, it does ramp up those fans pretty hard. So that’s a differentiating factor already from like a MacBook Pro. But it could be really interesting. And I think Microsoft is kind of, they’re specifically going after the MacBook Pro. If you look at their design for the new Surface laptop, I believe it’s a 15 inch, it’s not 16, but it’s got a mini LED display. It actually gets brighter than the MacBooks. It looks like a MacBook, haptic touch pad, really nice keyboard on that thing. They’re kind of gunning for them. Now, The possible Achilles heel, we’ll see how this plays out, is it’s Windows. It’s Windows on ARM. 02:17
Geoff: You really missed the other Achilles heel. We haven’t talked about pricing yet. 03:14
Adam: Oh, that’s true. Well, 03:19
Stephen: Oh, I didn’t 03:19
Adam: wait, 03:20
Stephen: know 03:20
Adam: wait, 03:20
Stephen: they 03:20
Adam: wait. 03:20
Stephen: released 03:20
Adam: But 03:20
Stephen: it. 03:20
Adam: if you get 128 gigs of unified memory in a MacBook, you’re going to pay about the same. So you’re 03:21
Geoff: Oh, no, 03:28
Adam: paying 03:28
Geoff: no, no. 03:28
Adam: out the nose. 03:28
Geoff: I’m not saying it’s not unreasonable. I’m just saying this is not for the faint of heart. The DGX Spark, which is the desktop equivalent, starts at $3,500. So I imagine this is going to be… Probably the $2,500 price range if I had a guess. 03:30
Alex: But that is kind of irrelevant. I mean, yes, the price is important because it determines ultimately who’s going to buy it. But I think what’s more interesting about this entire product is the signal that it sends from the top, the top being Nvidia. Their bread and butter is selling incredibly expensive bits of silicon to data center customers for tens of thousands of dollars for a single GPU. And here they are building a computer capable or an architecture capable 03:48
Geoff: Thank you. 04:22
Alex: eventually, maybe not today, but in a few years, capable of invalidating or even augmenting that data center strategy. They want to own both places. They want to own the cloud compute and also what’s on your desk, which I find fascinating. 04:23
Adam: It’s a different angle to come at too. They’re going for the high-end pro first, which we haven’t seen that from the ARM perspective, right? Specifically with some of these, like the Qualcomm’s, they kind of claimed that when they came out the gate, oh, this is going to be the equivalent of a high-end x86 offering from Intel or AMD. Didn’t really pan out, honestly. Um, so I’m interested to see how this compares. Uh, but I think again, so that’s the whole, the software is going to be the linchpin is what I think the hardware, I have no doubt that hardware is kick-ass. But 04:41
Geoff: Oh yeah. 05:15
Adam: Windows on ARM, so the question is, did the trillion dollar company NVIDIA pushing in this direction and Microsoft wanting this as well, is that gonna get more movement, more investment into getting native ARM ports over on the Windows side? 05:16
Alex: I thought that was a solved problem by now, isn’t it? I 05:36
Stephen: Yeah, 05:38
Alex: mean, 05:38
Stephen: but 05:39
Alex: we’ve 05:39
Stephen: there, 05:39
Alex: had Arm on… 05:39
Stephen: yeah, there is an ARM version of 11, like Windows 11. 05:41
Adam: For sure, 05:45
Stephen: We know 05:45
Adam: but 05:46
Stephen: this, 05:46
Adam: that’s 05:46
Stephen: right? 05:46
Adam: not what I’m saying. What I’m saying is that a lot of it’s been still running through compatibility layers to 05:46
Stephen: Okay. 05:52
Adam: get the x86 software to run on that platform. 05:52
Stephen: Sure. 05:57
Adam: A lot of things have moved. You get your browsers and a lot of what people use on the day-to-day for sure, but there’s still quite a few things and games have been kind of a linchpin on that side too. But there was some interesting, when I was watching the reviews of these units, they were showing Cyberpunk and some interesting games running at seemingly full speed without any hitches and whatnot. So I wonder how much they’re taking advantage of. I can’t remember what that, that translation layer that steam created 05:57
Alex: Proton. 06:29
Adam: for. Hmm. 06:29
Geoff: Proton. Oh, you’re thinking the one they’re going to have on the Steam frame. 06:30
Adam: Yes. 06:35
Geoff: Oh. 06:37
Adam: It’s a box 86. 06:37
Geoff: I will look it up while you guys are talking. 06:38
Adam: something like that, but I’m interested to see how they’re doing that. Once people get to peek under the hood, it was pretty much a closed box kind of review or all of the reviews that I saw seem to be that way where it’s like, okay, here’s what you can do with this box. And here’s what you can’t do, which is mostly can’t do on that list. So, but it looks compelling. 06:40
Alex: I wonder if… 07:01
Adam: I mean, from a design perspective, it looks just as compelling as a Mac book. And, and the one that I’m really looking at here is specifically the Microsoft surface. That’s not even speaking to all the other vendors. Cause everybody kind of, all the majors hopped on board. We got a ASUS MSI. I think HP might be on that list too. Yep. 07:02
Geoff: Dell, 07:21
Adam: Dell. 07:21
Geoff: Lenovo, all 07:21
Adam: Yep. 07:22
Geoff: of them. 07:22
Adam: Everybody, everybody’s jumping on. 07:23
Alex: I wonder if this is going to be an attractive option for your customers, Stephen, in MSP land who are on the small to medium end of business spend and don’t necessarily want to sink their teeth into an Anthropic subscription for tens of thousands of dollars a year or an OpenAI or Gemini subscription for the same. And they want to just have this stuff locally. I mean, some of your customers, I’m sure, have local compliance needs as well. You know, like stuff can’t leave the country or can’t leave the building or like depending on the sector they’re in. I wonder if these kinds of local devices that they just shove under a desk somewhere like the Spark, you know, that we have now or indeed a laptop. Are they compelling? Yeah. 07:26
Stephen: So first off, I don’t have 08:10
Geoff: Thank you. 08:11
Stephen: a ton of clients that want to do local AI models, mostly because it’s not a simple installer where they double click a thing and an application loads and it’s instantly working, right? 08:13
Alex: yet. 08:24
Stephen: Claude or any of the other ones, you literally just install the application and you can get going even without any subscription. You just start typing and it just starts taking all your info in, right? So first off, I’m excited for things that are coming out like this, mostly because it does show major hardware vendors trying to give you the ability to do local. And so that kind of makes me wonder if NVIDIA and stuff is actually going to start building a product out for being in your own building. Like, am I going to be able to download a Claude that is actually like the local Claude edition? I don’t know. But if the major vendors are pushing this way, one has to assume that people will try to make money that way. Maybe it’s a fraction of the price that a normal Claude version would be. I don’t know. But to your question, anyone that is doing… any sort of AI stuff right now from my customers’ perspectives, and that’ll be like medical field and whatnot. They just rely on what is available from, say, like their 365 instance. because you you sign up for their their co-pilot which be it good or bad if you’re paying the license and whatnot in the end user agreement they basically say that they’re not training on your data and so with certain loose loose morals on that situation i think most people are willing to just say yeah, that’s perfectly fine. I really want the AI thing, and I don’t necessarily worry too much about it. I don’t have any lawyers, and so I don’t want to kind of go down that avenue because I think, from what I understand, lawyers can be a little pickier about what leaves and what doesn’t. So, yeah, 08:24
Alex: I’m just 10:17
Stephen: I 10:18
Alex: waiting 10:18
Stephen: don’t… 10:18
Alex: for Geoff to interject at this point. 10:18
Stephen: Yeah, 10:19
Alex: Objection! 10:20
Stephen: I’m waiting for him to jump, right? 10:20
Geoff: Lawyer hat, lawyer hat. 10:21
Stephen: But, yeah, Geoff, I mean, feel free. What do you think on that? 10:23
Geoff: So it’s actually interesting because I don’t want to get too much into this, but my employer recently came out with an AI policy. And we’ve been, because again, judges and all that are trying to work out AI. I mean, AI is everywhere. You have attorneys using it and claiming, oh, I didn’t know that it could hallucinate and make up cases and all that. And so I’ve been using it for work, but I also have to keep in mind confidentiality, where I’m putting things. If I’m putting in contracts or stuff like that, I usually go and I’ll substitute out names of parties and things like that. So I’m not putting, because I’m using Claude right now for my own personal subscription. However, my IT department is actually working on using Microsoft’s Copilot and basically running it on-premises and giving access through a web portal for everyone in the office. And that will hopefully be, again, it’s local So it shouldn’t give up confidentiality, but I have not gotten a chance to use it yet, so I’ll be curious to 10:26
Alex: Are 11:35
Geoff: see how that goes. 11:35
Alex: we trusting Microsoft to not, you know, pinky swear, you know, we 11:35
Geoff: Okay. 11:39
Alex: won’t leave the building? Microslop of all companies. 11:39
Geoff: I actually have read Microsoft’s Terms and Conditions when it comes to AI because I had questions from… 11:43
Alex: Oh, well, let’s hear it because I think you’re probably one of the three people on earth that has. 11:49
Geoff: Oh, dear God, Microsoft is one of the worst companies when it comes to their terms. It is literally, it’ll be like, here are the terms and conditions. Here are these 17 other agreements and things that are incorporated, but not all of them are incorporated in this other document. They are absolutely the worst company I have ever dealt with in terms of contracts. There is nothing explicit in their contract that says that they will train on user data. There’s also nothing explicit in the contract that says they will not train on user data. 11:53
Alex: Which means they will. 12:18
Geoff: there is enough language in there that I think they’d have a hard time doing it because there is language on there about you own your data and things like that and they can use it for usage data and stuff like that so there could be some data used for training but your actual like hey can you tell me about how to make a bomb that should not be trained 12:22
Adam: Back to the laptop. I have a question for Steven on this also, given 12:46
Stephen: Go on, 12:49
Adam: that 12:50
Stephen: yeah. 12:50
Adam: the majority of your customer 12:50
Geoff: Bye. 12:51
Adam: base, if not all are running windows, is this, 12:52
Stephen: Yeah. 12:55
Adam: would this be something that would be compelling to you at all for, for your own personal use? 12:55
Stephen: That laptop, certainly, if I wanted to spend that much money on a laptop, then yeah, I think it would be pretty good. I’d use a Mac every day, and that’s just my normal thing. And so inherently, I’d really love a Mac that had all of the RAM, so I could try this stuff. But certainly from a PC perspective, if it had really nice hardware, I’d love to try it for sure. 13:00
Adam: but it’s not compelling specifically because of the customer base doesn’t, doesn’t register. 13:23
Stephen: No, not for my usage. No one pretty much is going to spend that much on a laptop that I sell. Like right now, understandably, your pricing is a bit different than my pricing, right? And so you say $2,500 and that ends up being $3,500 here. And so most businesses that I deal with don’t want to spend $2,500 on a laptop. They want to spend $1,500 Canadian. And that’s at the high end because they’re just going to cycle them out every few years. Right. So, no, I don’t have a huge, a huge go for it. But the exciting part for me is what it will mean for local LLM in the future. 13:28
Alex: Couldn’t they have made it something easier to say? 14:08
Stephen: LLM. Yeah. 14:11
Alex: L-L-L-L-L-L-L-L-L-M. 14:12
Stephen: Yeah. 14:13
Geoff: Yep. 14:13
Stephen: A hundred percent. Yeah. 14:14
Adam: Cool. 14:15
Alex: Always. 14:15
Stephen: Let’s 14:15
Adam: But 14:15
Stephen: just trip over it every single time. Right. 14:15
Alex: Yeah. 14:17
Adam: before we move on, here’s just one last thought is, will Linux come to these boxes? Because they are open 14:17
Alex: Yeah. 14:23
Adam: bootloaders. 14:23
Alex: Definitely 14:24
Stephen: 100%. 14:24
Alex: has to. Yeah. 14:24
Adam: Maybe not 14:28
Alex: All 14:28
Adam: the 14:28
Alex: right. 14:28
Adam: surface. 14:28
Alex: That was a short answer, wasn’t it? Yes, it will. I don’t see how it can’t because it’s the first class citizen for containers, for a lot of the model development happens on Linux. And on the server side in the cloud, it runs on Linux. So why wouldn’t you just want to port that whole cloth locally? And the Linux desktop is better than ever. So why not? 14:29
Geoff: One last thing. Sorry to keep doing this. Facts. That was the Valve runtime thing. I looked it up. 14:54
Alex: Okay, yeah. 15:00
Adam: Thank you. 15:01
Alex: All right, housekeeping time. Let’s just remind everybody where they can get in touch with us and support the show. You can go to bitflip.show to find out more about us and our wonderful patrons who support the show. You can get in touch with us at contact at bitflip.show. Speaking of patrons, we have our first terabyte tier supporter. So a big thank you to Badnet Mask for supporting the show. And, you know, if you want to get your name in the show and be preserved for perpetuity, or in perpetuity, I should say, please support the show, a terabyte tier or greater, and you’ll get a shout out every episode. Big thank you to everybody that has supported us already. There are quite a few of you. It’s been quite a pleasant surprise how many. So please, we would love to, I don’t know, keep bringing this show to you without, you know, paying for it out of our own pockets. You know what I mean? Please help us to help you keep bringing this show. And a big thank you from me. Big thank you from the other three as well to everybody who’s a patron. So you can head over to bitflip.show slash support to find out about our patron tiers. 15:02
Geoff: And we did have our first behind the scenes episode where we diagnosed some networking shenanigans. So if you’re at all interested in learning more about the show and kind of just watching us do our thing, by all means, please subscribe. 16:13
Alex: Yeah, 16:27
Adam: attempt 16:27
Alex: we did. 16:27
Adam: attempted to diagnose the network issue. 16:28
Geoff: Yeah, true. We didn’t quite 16:30
Adam: Let’s 16:32
Geoff: fully, 16:32
Adam: be real. 16:32
Geoff: other than, other than figuring out Steven’s internet is the problem. 16:32
Adam: Blame Canada. 16:37
Alex: I mean, an hour and a half of us running Iperf tests. I don’t know how interesting that was, but anyway. 16:37
Stephen: But I feel like it goes 16:43
Alex: All right, 16:44
Stephen: both ways, right? Because if I try to write to you, then all of a sudden you guys are the problem. 16:44
Geoff: No. 16:50
Alex: but there’s three of us. 16:50
Stephen: That’s true, yeah, sure. 16:51
Alex: And we don’t happen to have fancy data center rooms. You know, now I’m saying. 16:54
Stephen: Yep. 16:59
Alex: All right, now I nearly killed myself this weekend doing the Perfect Media Server Part 3. Am I allowed to talk about that yet? 17:01
Stephen: Sure. 17:07
Adam: Please do. 17:07
Alex: No. 17:09
Adam: I want more. I watched 17:10
Alex: I set myself 17:11
Adam: it. 17:12
Alex: up for that one. Yeah. Well, about 10 years ago, I wrote the very first Perfect Media Server article, as it was back then, on the Linux server blog. And the idea was just a showcase of free and open source technologies of how you could build a media server, a network attached storage server, using nothing but off-the-shelf free software and here we are 10 years later somehow and I’m still broadly speaking running the same stack and I made a video about it so it’s at the core it is Proxmox and I managed to make half of the internet come out in hives apparently by talking about running containers directly on the host and not running everything in a virtual machine and all the rest. It’s like, guys, it’s just Linux. It’s just Debian that also happens to be a capable hypervisor. There’s one guy here literally in the YouTube comments that says, what is the purpose of this? Just go with pure Debian. It also seems you installed Docker directly on the host. This is just wrong. And I’m like, okay, sure. Why? Why? Is it wrong to install Docker on top of Linux? 17:12
Geoff: No, 18:31
Alex: I mean, that’s what I did. 18:31
Geoff: no. We had this conversation kind of in the first episode because I do the same thing where I run Docker in because there are some things where it’s just easier to run Docker on the host like scrutiny. I don’t want to have to create a VM, pass all my disks through just to run scrutiny to VM. That is complete overkill. 18:32
Alex: and quicksync if you need any kind of 18:55
Geoff: Yep. 18:56
Alex: hardware acceleration for things like scripted or for what’s the other one uh frigate thank you yes if you have the coral tpu like there’s a there’s a bunch of reasons why you might want stuff directly on the host to which the next reply is we’ll just run everything in an lxc and i’m like okay been there done that tried that it doesn’t end well 18:56
Geoff: I don’t have a problem with QuickSync. The Coral TPU was an absolute PETA. I have tried passing that thing through to an LXC and to a VM, and it just does not play nice. 19:18
Alex: Yeah. I mean, my rationale for it really is that Proxmox is a base, right? If you want to experiment later on, you can. But the purpose of a guide like this is to just get people out of the ground as quickly as possible. It’s already a 90 minute video for crying out loud. going through how to partition your disks manually, going through how to install basic packages and edit config files and all the rest of it. And it’s like, this is already for most people just going to be, by the time you’ve opened up a terminal, you’ve already lost 95% of people. And the whole stack is built around Snap Raid, MergerFS and Proxmox. And by the end of it, I’m sort of left there thinking, and this isn’t a paid advertisement for Unraid. 19:31
Adam: Oh, I was 20:18
Alex: I 20:18
Adam: waiting for it. 20:18
Alex: promise, this isn’t. I have no affiliation except Adam’s a friend of mine, obviously. But I’m sort of sat there thinking, all of this stuff that I’ve learned over the years now, I could just throw it into an LLM, couldn’t I? And ask it, how do I partition a disk? And off goes the agent and does it for me. Do I need to know how to do it myself? I think so, because that’s just how my brain works. But then I look at an Unraid product that just handles most of that for me. Yes, it does it in an opinionated way. Same with TrueNAS, same with OpenMediaVault as well. They’ve all got opinionated ways of approaching these things. But I’m in a tough spot with it, really, because I’ve used this system myself now for a decade, and it just works pretty flawlessly. But, I don’t know. I just, I look at it and I’m like, is it right for most people? Is it perfect? No, not really. 20:20
Geoff: Bye. 21:20
Adam: I was listing in my mind as you went through each different factor that you were applying to the system and Unraid actually did everything. I couldn’t, 21:21
Alex: Yeah. 21:33
Adam: I can’t identify a single thing. that you set up that Unraid doesn’t do for, from specifically a media center perspective. Okay. Understanding 21:33
Alex: Yeah. Yeah. 21:42
Adam: that you’re, you have Proxmox as your base, and that’s going to give you a lot more opportunities for tinkering on things like LXCs and such, and having a enterprise level hypervisor, no argument there. But for this specific purpose, I was like, It’s such the easy button for this. I could have had this set up in 20 minutes easily with no CLI. So it really just matters. I think it matters what your, what your goal is. If you really want to understand all these technologies top to bottom, by all means, go, 21:42
Alex: Yeah. 22:18
Adam: you know, watch the move, watch the video, 22:18Alex: The movie. 22:21
Adam: go through step by step. 22:21
Alex: It basically was a feature 22:24
Geoff: It 22:25
Alex: length 22:25
Geoff: was. 22:25
Alex: epic, wasn’t it? Yeah. 22:25
Adam: It was of that level of quality. I’ll give you that compliment. There you go. 22:26
Alex: Thank you. 22:29
Adam: But if you just want a media center set up that will allow you to do a lot more tinkering and a lot of their other options as well, Unraid is a great option for that specifically for the media center. Also the limitations around performance from the array. Aren’t really an issue. Uh, and in fact, like the flexibility of the array is why you went with merger of us in a similar sense. Right. 22:31
Alex: Yeah, 22:52
Adam: But, 22:52
Alex: exactly. 22:53
Adam: um, I prefer the real time parody checks also over the snap 22:53
Alex: I think, 22:57
Adam: rate. 22:57
Alex: when I think about where I was 10 years ago, Unraid was version 6, maybe? And 22:58
Adam: Hmm. 23:05
Alex: you guys 23:06
Geoff: Bye. 23:07
Alex: have made a lot of improvements to the product since then. Like, you’ve added multiple parity disks, which I don’t think was a feature back then. You’ve added ZFS. You’ve added GPU pass-through, when I left, was barely, barely, barely a thing. And so it’s, you know, I put this stuff in place, like I say, 10 years ago, and I’ve made a wiki about it. I’ve written a dozen blog posts and all the rest of it. But I felt like this video had to happen as a matter of record to say, look, this thing exists. You can do this using nothing but free and open source software if you want to. Should make a follow-up video first. I think I’m going to which is like an FAQ because I’m seeing the same trends come up in the YouTube comments as well of like Why didn’t you just do this? Why are you doing it this way and all the rest of it? And in that video, I think I will have to address the elephant in the room of Why don’t you just push the easy button and play to pay? I don’t know 50 bucks 100 bucks whatever it is for Unraid 23:07
Geoff: Thank 24:02
Alex: and 24:02
Geoff: you. 24:02
Alex: and just get on with your life To which my answer would be I’ve built a nice career on the skills. I learned building that thing so I 24:03
Stephen: I think it’s more than that, right? Because you have done it this way for a long time. You’ve gone through many iterations. I always rib you about using a hypervisor in Just because at the end of the day, if you’re using a hypervisor in an actual enterprise environment or for business, you’re probably not the sort of fellow that’s actually going to install Docker on there. Just nod your head that you won’t because we’re friends. But in your situation, it’s literally a home lab or it’s your home media server. It’s just being used as a base to be able to build upon. And so that’s perfectly fine. But the reason I think that you’re gonna always build it the way that you’re building it is because you’re used to understanding and having the full control in the backend without a GUI. Now, 24:10
Alex: yeah 24:59
Stephen: that said, you are a Mac user and you’re commonly thrown into a walled garden. but you still complain about it, but you really can’t change. Right. But with, with your media center, you do have the option to not be in the wall garden and you 24:59
Alex: there’s 25:16
Stephen: don’t 25:16
Alex: no 25:16
Stephen: like 25:16
Alex: good alternative 25:16
Stephen: GUIs. 25:16
Alex: there’s no good alternative to the mac desktop for my needs where i’m doing video production all day every 25:18
Stephen: Right. 25:22
Alex: day and 25:22
Stephen: Right. A hundred percent. 25:23
Alex: like linux on linux for that use case i’m sorry Geoff 25:24
Geoff: No, 25:27
Alex: is 25:27
Geoff: no, I will cede media creation, not Linux’s strong suit. 25:27
Alex: Windows, 25:32
Stephen: Right. 25:32
Alex: nah-uh. I’m going in and out of airports and all around the place, and I still have not found a Windows laptop that just opens and closes and wakes and 25:33
Stephen: Sure. 25:43
Alex: sleeps every time, first time, connects to the Wi-Fi first time every time, and I can do airdrop from my phone when I’ve got footage. It’s just… But on the server, there are plenty of options. So there’s Unraid, there’s TrueNAS, there’s OpenMediaVault, there’s ZimmaOS, there’s PerfectMediaServerStack, whatever you want to call it. There’s bCacheFS, which we haven’t even looked at yet because it’s kind of… I don’t know. I feel like BcacheFS had its moment with getting accepted into the kernel and then ended up getting it yeeted out of the kernel like… Yeah, it’s 25:43
Stephen: But that’s 26:22
Alex: just an interesting 26:23
Stephen: my 26:23
Alex: one. 26:23
Stephen: point. 26:23
Alex: And there’s no one right way to do these things. This is the other thing about Linux, right? 26:24
Stephen: Right. 26:27
Alex: There’s always five different ways to skin a particular turkey, you know? 26:28
Stephen: And so that’s 26:32
Geoff: I 26:33
Stephen: my 26:33
Geoff: mean… 26:33
Stephen: point, right? You like it that way. And even though there are really great alternatives, that’s why you’re going to keep doing that, because there isn’t a thing pushing you over to using an Unraid or whatever. 26:33
Alex: Mm-hmm. 26:46
Geoff: I think the other thing to kind of keep in mind, Alex, is I still use Ansible to set up a lot of the stuff in Proxmox. And I don’t think that’s necessarily as doable with something like Unraid. I’ve never really played with Unraid, so please, Adam, correct me if I’m wrong, but… I like certain buttons tweaked. I like certain things certain ways. And with Ansible, I can go in and say, okay, go ahead and change this setting, this setting, this setting, this setting. And I do it once, and presumably it’s going to work forever unless they change the settings. 26:47
Adam: We 27:18
Geoff: With 27:18
Adam: actually 27:18
Geoff: Unraid, 27:18
Adam: use 27:18
Geoff: again, 27:18
Adam: it. 27:18
Geoff: I like a GUI, but I don’t want to have to go in every time I have to click 17 different buttons to when I can just click one button, run a script, and it’s done for me. 27:19
Adam: I happen to be the head of quality also for LimeTech. And in fact, we do deploy our QA lab with Ansible. But here’s the thing. You guys are all Docker Compose fans. 27:28
Geoff: Yes. 27:44
Adam: And Unraid still does not have Docker Compose natively in it. So that would be one thing that I didn’t consider until just now. So if you guys, when you’re talking about your Compose stacks and being able to do it declaratively for an entire stack, that part, that would be a differentiating factor. I’ll give you that one. 27:44
Alex: For now. I mean, 28:04
Stephen: For now. 28:05
Alex: there’s probably 15 plugins to add it, though. 28:07
Adam: That is true. 28:09
Stephen: Or you just do what I do and you just went into the terminal installed it anyway. 28:09
Alex: Exactly. Now, I’m still neck deep in the middle of Jellyfin June. That was the reason that I finally got off my ass and put out Perfect Media Server Part 3. I needed something to run Jellyfin on that wasn’t my primary Jellyfin instance to start messing around with it. So I’ve been looking at clients and doing live TV and all that kind of stuff. And it’s been really interesting. I’ve actually reached out to the core Jellyfin developers, and I’m hoping to line them up for an interview either on this podcast or on my YouTube channel. I’m not sure if they’re willing to do video or not, so that depends on… couple of things so we’ll see but it’s jellyfin is pretty interesting it plays video like movies and TV shows I would say is pretty much a solved problem with jellyfin the rest of it yeah some rough edges 28:15
Stephen: When’s the music episode, Alex? 29:08
Alex: well notice I didn’t include music in that list 29:11
Stephen: Hmm. Hmm. 29:14
Alex: because I’ve ended up spinning up a Navidrome 29:15
Geoff: To me, I hear you, and actually one of our listeners wrote in and made similar points about Jellyfin. The developers seem to be focusing on improving the UI and making some improvements to how things are running and the speed of things and optimizations and all that, but they haven’t worked on some of the core features. I kind of agree with him. Particularly, he raised the point of being able to transcode a download so you can have a smaller version of the file. full quality image of whatever movie you want. But when you go on an airplane, do you really need to have it in 4K max 29:20
Alex: No. 29:58
Geoff: bit? No. But there’s no easy way of doing that from Jellyfin, and that would be a great feature to have. And it’s been one of the number one requested features, and it’s just been sitting there for 10 years now. 29:58
Alex: But that’s because Jellyfin is led by volunteers. 30:11
Geoff: Yes. 30:16
Alex: So if you want that feature, go build it. I mean, in the age of agentic AI, there’s no reason why you couldn’t go and build it, Geoff, hypothetically. 30:17
Geoff: Hypothetically, but then would it be merged into the main line would be the question. 30:26
Alex: I actually 30:33
Geoff: I don’t know. 30:33
Alex: don’t know what Jellyfin’s policy on AI contributions is. 30:34
Geoff: But I mean, again, I go back to the going back. Yes, it can do movies and it can do TV shows. A plus. I have no issues with it. I agree with you that movies or books and audio books and music and stuff like that are definitely not its forte. But to me, that’s fine. There are other products that do just as well 30:39
Alex: yeah 30:59
Geoff: for that. Correct. 30:59
Alex: it’s um it’s not a holistic solution 31:01
Geoff: Correct. 31:03
Alex: like like plex is although plex is only good at audiobooks because of the prologue client i should say uh which prologue has added audiobook shelf support along with and i’m i’m trying to look for it on my phone i added a another audiobook client which i forget what it’s called 31:04
Geoff: Did you start with 31:23
Alex: uh 31:24
Geoff: an A? There was one that I saw that I’ve been testing. 31:24
Alex: Yes, that’s it. Absorb, that’s the one. And this is a really slick client for Audio Bookshelf. And I think just Audio Bookshelf, if I’m not mistaken. So we’ve got more options now than ever across the ecosystem. And honestly, it kind of follows the Unix philosophy of just do one thing and do it well. And so if we’re okay with Jellyfin being just movies and TVs and shows and audio bookshelf handling books and Navidrome handling music, is that really such a bad thing? I mean, it’s so trivial to spin up these containers these days. 31:28
Adam: I don’t think it’s bad at all. That was why I was asking you guys, is music a deal breaker? Are these things deal breakers? Because to me, they aren’t. I actually kind of like the separation of duties for some of those things. So Jellyfin works just fine, really. They only… I wouldn’t even call it a gotcha. The only challenging bit for me with it is the remote access part still doesn’t have that discoverability where I can just give it to mom. That’s the only, that’s the 32:09
Alex: Yeah, 32:39
Adam: only linchpin. 32:39
Alex: it never will. 32:40
Geoff: Nope. 32:40
Alex: It never will 32:41
Adam: Nope. 32:41
Alex: unless you add some kind of client. And the clients need to be aware of how to phone home or phone your server, however that is. And I made a video for work about this. And it still presupposes that you can install TailScale on your client device, which is fine until you get to a PlayStation 5 or you get to, I think there is an experimental Tizen port of TailScale actually at the minute. but yeah it’s rough unless you open ports in your firewall which you know with the CVE jungle there is out there these days as we touched on a couple of episodes ago I don’t know it’s not really not really something I want to do 32:41
Adam: Or you get fancy with pangolin or something along those lines. 33:23
Alex: you can do that yeah 33:27
Geoff: Yeah. 33:28
Alex: yeah yeah you absolutely can 33:28
Stephen: You guys aren’t setting your parents up with like a site to site VPN over a really nice unified gateway. I don’t understand. 33:30
Geoff: No. 33:37
Alex: To be honest with you, man, I ran Plex for my family for a very long time and in the end, I love my family dearly, but all it created was friction. If I want to take my server down to do maintenance, and this was early Alex. This was sort of five years ago. This was before I’d really internalized the idea of home prod. As a father now, I genuinely understand the importance of home prod because when kiddo needs to be parked for a couple of minutes whilst my wife’s teaching or something like that, we just need it to work. there can’t be there really can’t be downtime and i don’t want as a nerd the fallback to be oh we’ll just open up disney plus or youtube because that it’s just a cesspit on youtube for kids so um for me i think i i’m just fine keeping it to myself honestly and it’s 33:37
Adam: Hmm. 34:31
Alex: not the most uh friendly of of stances but It is what it is. If you want to access it, I’ll give you access over a VPN, but the batteries aren’t included. You’re going to have to do a little bit of legwork. It’s just 34:31
Adam: I 34:46
Alex: the price of entry. 34:46
Adam: would miss the, so there’s like an interaction aspect to the sharing that I enjoy. So me being able to tell my brothers, plus like you, Alex, I live very far away from my family, most of them at this point. But me being able to say, hey, I just watched this. It’s on Plex. You should all watch it. Let’s share notes and like have that fun interaction around it is something that I think I would really miss if I wasn’t sharing it out. 34:48
Alex: Yeah, that’s true. I suppose it depends how close you are with your family or friends, I guess. All of my friends run their own servers, so… You know, 35:17
Geoff: Yeah. 35:27
Alex: even my sister’s husband has started doing it, you know? So it’s a tricky subject. Version 12 of Jellyfin is coming out real soon, and they’re skipping version 11, or at least it would appear that they are. They’re actually changing their naming scheme. So they’re currently on 35:28
Geoff: Oh. 35:45
Alex: 10.11. So that’s currently version 11. And so they’re moving 35:45
Geoff: Oh. 35:50
Alex: to version 12 and just dropping the prefix. 35:50
Geoff: That makes, I was trying to figure that one out and I could not wrap my head around it. Thank you. 35:53
Alex: Some major performance improvements coming in that one, we are led to believe. And a release candidate will be along any day, but I hope for my own sanity it doesn’t happen in the middle of Jellyfin June. 35:59
Geoff: Well, hopefully not with any major breaking changes and stuff like that. 36:09
Stephen: Can 36:13
Alex: Yeah, 36:13
Stephen: you imagine it changes 36:13
Alex: yeah. 36:13
Stephen: everything that you’ve been talking about? 36:14
Alex: Oh, my goodness. Yeah, well, it may very well do that. Anyway, we’ve got a whole bunch of clients here in the show, Doc, but I’m conscious of time. So I think we’ll do a client deep dive probably in a future episode and come back to Jellyfin next time as well a little bit. Maybe talk about live TV because the World Cup starts tomorrow as we record. So I need to get my live TV sorted because I want to watch the England game that’s happening soon. So, yeah, 36:16
Geoff: I mean, 36:42
Alex: we’ll see how that goes. 36:42
Geoff: speaking of clients, if you have a favorite client, why don’t you contact us and give us your suggestions? I mean, we’ve got ones that we use, but I would love to hear what our listeners are actually using. Okay. 36:43
Alex: [email protected], please. All right, well, I think that’ll probably wrap us up for today. Big thank you to everybody that listened so far. You can find us at bitflip.show. As I say, contact at bitflip.show. Big thank you to our patrons over at bitflip.show slash support. Thank you very much for listening. I’ve been Alex. 36:53
Adam: I’ve been Adam. 37:13
Geoff: I’ve been Geoff. 37:14
Stephen: And I’ve been Stephen. 37:15