The Offset Podcast - The Offset Podcast EP029: ARM In Post Production Part 2

Episode Date: April 1, 2025

In this episode, we continue our discussion about ARM-based computing and its impact on postproduction. If you missed part 1 in this series, be sure to check that out, as we cover a lot of helpful background information. In this show, some of the specifics discussed include:How ARM will factor into our next workstation purchasesDo we think Apple will deliver on a ‘beast mode’ Mac ProWill Apple Silicon ever support discreet GPUsHow ARM Linux systems are very appealing from a hardware point of view but have software challengesIs widespread adoption of ARM an inflection point for cloud-based workstations for the massesA side effect of cloud-based workstations might be faster feature/bug fix deploymentWhy the future of workstations is really excitingBe sure to check out our library of episodes, and if you liked the show, be sure to like and subscribe to the podcast wherever you find it.

Hey there, and welcome back to another episode of the Offset podcast. And today, we're continuing our conversation about arm processors and the future of arm in the post industry. Stay tuned. This podcast is sponsored by Flanders Scientific, leaders in color accurate display solutions for professional video. Whether you're a colorist, an editor, a DIT, or a broadcast engineer, Flanders Scientific has a professional display solution to meet your needs. Learn more at flanderscientific.com. All right. Well, Joey, we couldn't get enough, man. We couldn't get enough in part one. Everybody, welcome back to another installment of the offset podcast. I am Robbie Carmen with me, as always, is my partner, Joey Deanna. Hey, Joey, how are you, man? Hey, welcome back. Yeah, man, welcome back. So we obviously had a lot to talk about, about Arm in general, and decided to break this up into two parts. So they're a little easier to digest in one sitting. And so we are continuing on talking about Arm.

ARM processors and the future of ARM in post here in part two. We had a lot of good stuff we covered in part one. But in part two, I want to get a little bit more into some of our prognostication, where we kind of think our next, what our next computer choices might look like, things like cloud computing with ARM and stuff like that, and kind of, you know, prognosticate a little bit about where we think things are going. And of course, we're not, you know, we're not fortunate. tellers, but I think we have a pretty good idea with connecting some of these dots where things

might be going. And I think this is going to be an exciting year for Arm and for our industry using Arm as well. So let's talk about what I kind of consider the what is the next computer question? Because we've been talking about this a lot. You know, what is my next workstation? It's something that I think about a lot because, you know, that main workstation is making me all my money. It is one of the most important tools in my tools. And, you know, you've moved from the Intel Mac Pro to the latest and greatest Mac Studio, and it's been performing really well for you. It actually performs better than my Intel Mac Pro in a lot of workloads.

And it's a really good primary computer. But I'm looking at the workstation purchase as a, you know, seven to 10 year long investment. And I want to get the most professional, the most rock, solid, reliable. performant incredible machine that I can just beat on all day every day. And I don't think Apple is quite there yet with their product line. Obviously, they do have a ARM Apple M-series Mac Pro now. It's essentially a Mac Studio in the bigger case. The only thing you get is some PCI slots.

But because it doesn't support discrete GPUs, there's not much use for that. You get it for your decklink. I'm not sure. Yeah, decklake, maybe network card or a, you know, a V&B card or something like that. To me, here's the fundamental question. We've already mentioned that Apple and using SOCs, they scale up by quantity, right?

So you get the fastest SOC you can get, then you make two of them. You interconnect them. Now, at some point, you make it such a big, big unit that even your little Mac Studio can't dissipate the heat or can't provide it enough power. That's where I think the next generation Mac Pro could excel

because that chassis we know can deliver, I think it's like a 1500 or a 2,000 watt power supply in that. It can deliver tons of power and it can remove and dissipate massive amounts of heat. So right now, between the Mac Studio and the Mac Pro, buy the Mac Studio all day long because the Mac Pro is no faster, right? they're not using that chassis to anywhere near its Macs. Yeah, it's something they already had the R&D and the investment in.

Why retool if you don't need to just drop in the chip? Here's the question for Apple and me. And this is what it comes down to. Are me and Apple going to stay workstation friends? Because like you said at the very beginning, we go with what is the best tool for the job at the time. I've gone from Linux. I've gone from Windows.

N-T to Irix to Linux to Windows again to Mac. Could be Linux again, could be Windows again, whatever. Whoever's making the best workstation is where I'm going to go every time. That's what brought me back to the Mac, was that 2019 Mac Pro. Now, what's going to keep me on the Mac is if Apple does the following. They take some of these big, beautiful SOCs that they have and they somehow get four of them interconnected together and that thing puts out enough heat and takes enough power that they actually really need to put it in that big Mac Pro chassis to make it work.

That will be a monster machine. The fundamental question that you came up with, I think, is will they keep on using their own GPUs on the dye? Or will they make their own external GPU make the die's CPU only? and it's like I have tons of CPU and tons of GPU. I think that configuration is unlikely. I don't think we're going to see a Mac with discrete GPUs probably ever again.

I don't think Apple wants to be in the business of making PCI cards. I don't think they want to be in the business of selling upgrade kits. They want to sell you a box that you use until it's done. Let me interrupt you a second because I had one thought about this. I tend to agree. I think that they have too much R&D and investment in the architecture that they've designed to have it for a relatively niche product.

Yeah, they cannot reinvent the wheel for the Mac Pro. Right. You know, they can't pour in millions and millions of dollars to something that's going to sell thousands, maybe tens of thousands of units versus the millions of units that they sell for their other machines. So I agree with that part. I will say, though, that we've been thinking about this moveover

in maybe some clouded, slightly biased fashion, and that is one, that somebody else doesn't move the needle, right? So so far, Qualcomm and Snapdragon, that's been mobile products. That's been like, you know, like the laptop behind me. That's been for those kind of things. You know, Microsoft, Qualcomm, like, there hasn't been like a consumer push yet into more desktop workstationy type computers in a major way. I mean, they have a little bit by those companies.

They've kind of ceded that to, we'll get to the, this in a minute. They've kind of seeded that to Nvidia, who's part owner and arm and other, you know, more data center focus type of machines. So that could be, that could push Apple a little bit in that direction. But two, I think we've been thinking about it in a traditional

form factor. Like, so when the Macs Studio came out, I think besides the arm stuff, it was like, wow, that is a weird design, right? Like in principle, it was like, that is a weird box. It's a big square, not a little square. And it's not a workstation size. It's not a rack. And it's not a

racking. So I think there is potential for, you know, I think the current Mac Pro is a 5U chassis, right? You know, and it has a lot of space in it for all the things we talked about, you know, PCI slots and, you know, RAM and all that kind of stuff. I think it is possible that they could reinvent that chassis a little bit to take advantage of something that's like, hey, maybe we go from, maybe we lose the PCI cards, but we go to a two U or a one U even, you know, likely, probably not, I think they don't want to have to reinvent the reinvent the wheel for a niche product. But that kind of thing is always within their

Yeah. So the question is where do they go next? Do they build a bigger Mac studio? Or do they build an Epic Mac Pro? Personally, I think for most people, the bigger Mac studio would be a wonderful choice. For me,

I prefer a level of professionality in a work computer that is above the Mac Studio. And just to be clear, Just to be clear, because I know this, but our audience probably doesn't. Joey has some very strong opinions on things like thermal performance, connectivity placement, things of that nature. And I own four Mac studios in various places at home and at the office and stuff.

And it's a little bit of a hack job, right? It's like, okay, fine, to get them in a rack mount. I had to buy a case for it to rack mount them. Okay, to get PCI slots. depending on Thunderbolt expansion. Oh, but wait, I want it rack mount. Well, that's even more. So to get it to operate like a workstation is a little bit of a hatch job.

But again, Apple might not go down that route. They might just say this is not worth our R&D in time to please Joey and Rob. And I think we're out of that market. I think that's a likely outcome, honestly. I'm holding out hope that they come out with a big bad Mac pro that meets my standard of workstation. But I think it's unlikely that they. will. And just to go into a little bit more detail, you know, Arm while being not new, as we've said, is new to MacOS and is new to a lot of the fundamental software that we use. I don't mean

just DaVinci Resolve or Adobe or anything else. Just about any modern software is built on piles and piles and piles of libraries that handle certain parts of the code, whether it be file I.O, whether it be dealing with strings, whether it be dealing with times and dates. There's thousands of libraries that all had to be kind of ported and optimized and compiled for Mac on Arm. That's why it took some time. Oh, it's still, it's still in transition for a lot of people. Yeah.

Exactly. And what, what I have said anecdotally, and I'm going to, I'm going to say it here, get ready, internet, to be mad. but I don't think that the ARMACs in terms of workstation level reliability are really ready for prime time in the software department and what I mean by that is they crash more they need to be rebooted more just little intangible things I feel like I've used them long enough now

that I can say yeah my Intel Mac is definitely more stable than any of the ARM Macs and to be clear, this is a huge generational transition. It will get there. And I'm really, really nitpicking here. Well, I think, I think the one thing that Apple in particular, and Qualcomm and Vydea might have some of this, but they're not as vertically integrated through the whole ecosystem as Apple is. I mean, Apple does a lot of tricks, you know, kind of things under the hood to kind of get some of this stuff to work a little bit better. You know, when we've seen it over the years with their phones and how the phones and the mobile devices work.

They're porting some of that to this. So I mean, I, I am more of a risk taker, edge use case, first adopter, see if it breaks, kind of thing. And I've just, anecdotally, I've been really pleased with the performance, right? I've been really pleased with the acoustics. I've really been pleased with the thermals and just the overall performance. But, I mean, I will, I'm not kidding you that I have the same lust factor. over raw power, you know, that you have and a lot of people have.

And I frequently find myself at 11 o'clock at night sitting on the couch and my iPad, my arm processor in my hands, looking at, you know, like super micro sites and building, you know, quad, quad CPU, you know, 96 core beasts and stuff like that. I get that raw power. So that exploration, I think it's worth noting, does have some legs, right? And I think that over the past five, six, seven years, the highest end buyers, that is the data center buyer, has also made a little bit of a transition slowly but surely, moving away from Intel, moving away from AMD, going to integrated SOCs that we're seeing from companies like Nvidia, for example, right? And I said earlier, Nvidia, slightly complicated owner of the arm architecture or whatever, and there's been some legal stuff going on there. But my point is, is that we have seen that.

And now we're starting to see that filter out the other way back to the workstation market. So I discovered this company, System 76, not an endorsement of System 76 at all. Just my Google foo found them. And let me tell you some specs on this box real quick, because this is like on paper, this is drool worthy. And this is everything you'd want a workstation to be. So starting with the processor, it's the ampere ultra up to 128 cores three gigahertz. That's a arm processor that that company is coming out for workstations, which is great. 512 gigs of RAM.

Oh, and guess what? Dedicated GPUs from Nvidia. So if you want an A6,000 or something in there, here's the only thing about it, man, is that it's running Ubuntu. And I don't mean that as in a negative way. I mean that it is Linux on ARM. So all the components are there. But guess what? Not a lot of software developers are making creative applications for Linux.

running on Arm and that's where we're going to get stuck, I think, yeah? Exactly. So here's where we get to, you know, the other big, big, big elephant in the room is what does this mean for big workstations and for cloud workstations? We keep talking about what's our next computer do we buy? It might not be a computer at all. It might be an instance in the cloud because for all the reasons that we said Arm is a good performer, it's an even better performer for data centers. You have this company, Ampeer, that makes these gigantic data center CPUs that have 64, 128, or more arm cores. That means in one rack unit server at Amazon, they can split those 128 cores into a boatload of virtual machines. Right. So for the amount of the density of power that gets put into each server is so much higher with these arm CPUs that you can get cloud compute, cloud workstations, cloud servers cheaper if you opt for arm. And that's across

the board with all cloud providers. You can choose these days between arm instances and Intel instances and the arm instances are cheaper. What does that mean? It means let's look at all the pieces to this ecosystem, right? We've got the CPU. Arm, wonderful. Linux runs on it great. Result. Runs on Linux. Great. GPUs. We've got Nvidia GPUs that can be virtualized. So you can take one gigantic GPU and split it up into a little GPU assignments

per virtual machine. Guess what? Linux supports that. Works perfectly. Resolve supports Nvidia GPUs. It just doesn't support Nvidia GPUs virtualized with an arm CPU on Linux. So all of these little ingredients are all there in the ecosystem. They just haven't been put together.

Imagine with me, if you will. Blackmagic comes up with a version of resolve that runs on Arm Linux and uses discrete or virtualized GPUs. You now have the perfect equation for a fully cloud-based workstation offering that will be have the power of the discrete GPU. Have the efficiency and the cost savings of Arm and run the same DaVinci Resolve you've always run. We've priced out virtualized workstations in the cloud running Windows, invidia, and Intel, and it can be done, and they actually run very well. It is just really price prohibitive right now. If my imaginary dream of a fully cloud workstation comes to fruition on the software side, I think the cost could be at the point where it is, you know,

a good question. Do I just pay a monthly. fee based on usage for a workstation in the cloud and storage. And I just own a small thin client that connects to my decklink, uses remote monitoring, and connects to my mini panel uses the black magic remote panel. Yeah, I think there's something to it. If we adopted or, you know, people who wanted to adopt that, you'd be cutting edge, bleeding edge, or however you want to say it at first. But I think clearly everything is moving in that direction, right?

you know, you only have to look at the, you know, the virtualized, you know, cloud NASS systems like, you know, Lucid Link and all that stuff. People are moving to that space anyway, want location independence, want, you know, decentralization of where the stuff is, you know, all that kind of stuff. So I do think that makes a lot of sense. I'm with you on the exploration of that. I'm a little hesitant for all of the intangible feel-good reasons to do that, right? I mean, like, you know, like, I love having. the monster workstation. That's why I'm keeping my 2019 until the very last breath. But it's even more simple than that. It's like, oh, well, something's wrong. Like, let me just go hit

restart, right? In a, you know, I guess there could be a tier of that kind of service where it's like it's dedicated to you and to you only. But, you know, the models for a lot of cloud is that you're sharing resources with other people. So like the ability to be like, hey, let's just restart the box. probably not existing in the same way that it is now when the physical piece of hardware is right in front of it. Yeah, but I think it's, like you said, I'm imagining some bleeding edge stuff. Right, right.

I think this is going to be the space to watch at NAB this year, is cloud offerings of various software that are optimized for, you know, we're probably going to see different marketing terms for it. We might see cloud native or like a Amazon, dedicated instance to this or other companies might kind of make their own service offering with their own branding. But at the end of the day, the combination of highly efficient armed CPUs, virtual GPUs, all in the cloud with kind of Linux putting it all together and the software being ported to that,

I think we're going to see more movement on that than just from Black Magic. And to be clear, I'm speculating about what Black Magic is going to do. I know nothing about if they're going to come out. with ARM Linux versions or anything like that. Reading the tea leaves, it sort of makes sense, right? I mean, Apple. Yeah, they have all the disparate parts, right? Yeah.

It runs on Linux. It runs on Arm. It runs on Nvidia. It just doesn't do all those three things together yet. And also, I mean, think about it this way. You know, from a distribution point of view, getting people up to the later, like there is a level like Apple has for that kind of thing over, from the developer's point of view, of greater control over the product, right?

So right now, you know, you have a universe where people are like, well, I'm using everything from Resolve 12 up to Resolve latest and greatest, right? And every dot release in between. And so for a lot of like those emergency kind of like, you know, bug fixes, that kind of stuff, the efficiency of pushing that out to controlled cloud instances versus having to having to distribute that kind of thing and hope people download it is a big, you know, a big boon for developers because they can squash, you know, the complaints and bug issues much faster than they could buy the traditional. Part of it is almost going back to the systems days of old where, you know, you bought a flame. That flame came with a $70,000 octane because that was the computer it ran on and that's what it was designed for. Right. So there was one qualified set of hardware for that.

And that's obviously gone completely out the window with, you know, the way modern creative applications work. And in a lot of cases, that's for the better. You know, you can use them on your own hardware. You can do your own configurations, whatever. But if this move to arm industry wide kind of allows more virtualized solutions, you can put the ideal hardware configuration in a config file and just boot it up on Amazon. Yeah. And I mean, I think it's a little scary.

Honestly, I think it is scary to people to pull control and to have, you know, essentially a service that you don't know how it's connected. You don't know what the rack looks like. you don't know what's next to it. You don't know who's touching it. Like all of those kind of, I get that part. It is, it is a little weird and scary. But it's not, we're making us,

so, you know, people are getting more comfortable with putting their data in the cloud, right? And working in cloud-based storage. So, like, I think it is, you're right. It's only a matter of time between these truly virtualized, you know, situations become more commonplace. There's always going to be the edge case of use, though, right? You know, the air-dap systems and the place that, you know,

can't have that connected. I'm going to live in that edge case. probably. Right. You know, there's a lot of reasons compelling either way, but I do think, to your greater point, it will become more common. I do think NAB this year is going to be a fascinating read. I think

the worldwide developer conference from Apple and VDiA's annual conference, all these things, I think over the next year are going to have some clarity on where we're going with this stuff too. Yeah, and again, if Blackmagic does make this hypothetical, Nvidia capable Linux resolve, that

works great in the cloud, it would also work great on monster arm workstations that you could buy for your own setup. You know, to kind of wrap this whole thing up, I know we've gone really all over the place here, but Arm has been a great transition for the industry. I'm very excited about it. It's bringing in more competition in more areas than we've had in a very long time, right? The fact that Apple is making GPUs that for our purposes at least are rivaling the big Nvidia ones, Well, guess what? That's just going to make

Nvidia make bigger and badder GPUs. And now Qualcomm's getting in the GPU game with their Snapdragon laptops that run resolve really, really, really well. No, I agree, man. And I think, you know, I think really what this comes down to is people look to Apple as a trendsetter, even though they tend to be late to the game most of the time,

but a trend center in the way of getting it out to the masses, right? And I think a large part of this for our work and our workflows is going to be how much tolerance does Apple have to create relatively niche products just to say they have it in their product lineup? And that's a big question mark, right? Because if I was, if I was a bean counter at Apple, I'd go, guys, this makes zero sense, right? Like we sell millions and tens of millions of iPhones and iPads. Great. Let's focus there. Sure, we'll make laptops because people like to be mobile and like to use laptops. Okay, cool. we'll appease the creative guy by having the studio.

But like the workstation market is not Apple's game, right? And so they- Honestly, I was surprised when 2019 came out. Right. After the disaster that was the 2013, I was surprised they not only came back into that game, but came into that game with such a superlatively good offering. So I'm not convinced that Apple is the place, like you,

I'm not convinced Apple is the place to watch for that next big development. I hope it is because I like a lot of things about that ecosystem. Yeah. But, you know, I think that I think it is encouraging that, you know, over the past couple years, there really hasn't been an alternative for people like us, right, unless you wanted a, you know, a less capable Snapdragon system or something like that. But now that we're seeing this development on, you know, Linux arm, uh, Windows based arm systems, like, okay, cool.

Like, we're now seeing the benefits of the SOC and the architecture and now we're going to get some choices. it's to be seen whether that's from, you know, Apple, Nvidia, whoever, to develop that next great thing that, you know, our industry will adopt. I have hopes, but we'll see. Yeah, very cool. All right, well, Joey, always a good on conversation. I appreciate your knowledge and history of this.

By the way, for the low, little price of $15.99, you can take a tour of Joey's Computing Museum in his basement. You just have to get some Maryland and buy some tickets. But no, in all seriousness, I think this is a good discussion. Again, we're not, you know, there's a lot to be to be understood about the manufacturing of these chips and the design considerations, a lot of information out there on the web. I encourage you to take a look at it. Some of it's technical and kind of hard to read, but there is, this is at the forefront of our industry right now. So it's definitely time to get educated on it. So for the Offset Podcast, I'm Robbie Carman.

And I'm Joey Deanna. Thanks for listening. Thank you.