 So we're here at the Lunar Connect. Hi, so who are you? Hi, I'm Robin Rantawa. I work for ARM at Cambridge in the UK. I'm responsible for system software architecture, specifically the safety track. Safety? Safety. That's right. I have my friends over here from System76. This is the System76 laptop. System76 is cool as far as I know because they do hardware that's built for Linux. And I got them here because everyone wants to see ARM hardware and ARM software. And we want to see ARM development happen on ARM machines. So what's the probability of that happening? Why don't I let System76 explain that? Yeah, sure. So I'm the engineering manager at System76. My name is Jeremy Solar. I direct all the engineering activities. We've been investigating ARM for quite some time and we actually have an ARM device that we've already sold. We have the Starling Pro, which is our ARM server, and we actually did an ARM netbook a while ago called the Starling. What we're looking for in ARM is a high-performance desktop class processor, something that could be utilized in a laptop or desktop. We're looking for a replacement for U-class and H-class Intel processors. And we think there are a lot of promising alternatives here. David's been doing investigative work on that front. So you have some leads? Yeah, we've been looking at several SOCs here. And we've also been discussing a lot of the challenges and things we would like to see from SOC vendors and various other people here at Lenoero in order to help OEMs like us be able to offer fully supported ARM machines on Linux. So how completely awesome would it be if this succeeds? If there's an amazing cool ARM laptop, then everybody in ARM is going to use one? Yeah, I think it's fair to say that a lot of developers, specifically kernel developers, there's a better incentive for them to do development for ARM on ARM if you actually have a laptop or a desktop workstation that's actually using ARM technology. I mean, you've heard all of the commentary from Linus and others that you guys need to up the game by providing ARM hardware that can be used for ARM development and this is probably one way of doing it. So I think the guys at Lenoero are very much hoping this project works. So what would be the challenges in making that work? You need to make a PCB design, you need to test it like crazy to make sure it works and everything? For us, making the PCB design is one of the easiest things because most ARM CPUs provide a very obvious pinout and specifications for mechanical placement of the CPU package. The issue for us on the Intel side, we can't even get that information without going through an NDA process and significant sales work to demonstrate that we can meet certain order requirements. So for ARM CPUs, it's a lot easier to get the technical specifications but what we need from vendors also is a guarantee that this hardware is going to be continuously supported through updated Linux kernels and through updated versions of Ubuntu for many years. Currently, we've been seeing issues with the ThunderX processor which is what we're offering on our ARM server. We're only able to deliver it with 16.04 because the vendor has not been updating support and Cavium has not been updating support in newer kernel versions and making sure it continues to work without regressions. I'm very excited by the latest trend at the ARM marketing department where they're talking about matching Core i5, Core i3, comparing them that way. That's going to be exciting when some chips come out with A76 that are super powerful and stuff like that. But many of the ARM chips so far have been either for the smartphone or for a few coming from the server but for laptops we need something kind of in between, right? Yes, I agree. I think like the... and Jeremy might know more than me over here but fundamentally the trend has been even in the Intel space to actually aim at the desktop workstation, right? And then there was a market that was created for more mobile desktops which happened to be laptops and in the beginning the packages for desktop workstations were the ones that were being in a sense fitted into this form factor, right? There's no reason why a similar trend couldn't happen with ARM for example, right? I mean I personally don't think there'll be a specific line of processors that are aiming for laptops but I think there should be enough wherewithal to take something that's aiming for the enterprise market perhaps or the high-end mobile market in terms of processors and systems around it that can be repurposed much in the way Intel did for laptops. So without saying any secrets about the potential leads you might have but maybe for example there are some processors that are kind of for networking or there are maybe not the Thunder X2, it's maybe too big for a laptop, right? But something else that could be out there and then adding through the PCI and having graphics and everything? Yes, so the latest Cortex-A processors as you referenced in the marketing blurbs they've made it very clear that they're in the same performance envelope as incumbent Cortex-Cora 5 implementations from Intel. You can extrapolate the performance envelope and I would say that in the next two to three years there'll be options available which are even better for this class of market, right? So without trying to get any secrets or anything and you can't be saying that on behalf of ARM but it'd be nice if ARM was able to provide the support to make sure this works, right? That's why they heard at Lenaro, right? Because I wanted to tap into this really hot trend around ARM on ARM at this connect and the connect before this and actually try and get these guys in so that they can share their insights because they do commercial grade really classy laptops not with ARM right now but they have a whole load of information they can share with us, right? There's also a very wide range of what constitutes a laptop. We saw everything on the Intel side from 25-watt TDP processors all the way up to about 95-watt processors like what you'd find in the servo here. What's this? This is the System76 servo. This is a couple-year-old model. Yeah? It has a socketed desktop i7 processor. Socketed? Socketed. This is actually literally a desktop processor in a laptop. And how thick is this? It's about an inch and a half thick total. You could totally fit a Thunder X2 in there. Yes, yes. You could activate all the cores. The idea would be exactly that. On our 13-inch Galago, we may have something similar to an extremely high-performance mobile processor like the Kirin that recently came out but then moving up to the Works class, this class we really need something in between the Thunder X2 and the highest-end mobile processor. This is an H-class Intel processor, and it also has NVIDIA graphics. Can we open and see the... So you have super nice keyboards and everything? Yeah, I have a backlit keyboard. So we can see the backlit there. How nice is the keyboard? How do you engineer that? It's a very nice keyboard. It comes from a company called Chikoni. And we are able to choose keyboards and ask for design changes. This is a pretty good one. And how about the mouse? This is synaptics. So it's best in class? Yeah, yeah. It's standard on most laptops. And displays are good? Yeah, of course. So you get any mats, like people like mats, right? Or what kind of displays do you do glossy to? Yeah, yeah. Our 13-inch Galago has a glossy high-DPI display. We also offer a 4K display on this model, and on the Servo and the Nobo as well, which are larger models. We have a number of different options. You can go visit our website, system76.com, to see them all. But what we're really looking for from ARM, we absolutely could have a Cavium Thunderex in its glass. Is that like for sure? Isn't it too hot? The Cavium Thunderex looks like it has a big, what do you call it, cooling system over there. We put i7-8086's in this. We put 5 GHz Intel processors in there with 6 cores. And 8 cores, actually, with the next platform job. We'll have the 9900K, which has soldered internal heatsink and can go up to 5 GHz boost and has 8 cores, 16 threads. So is it the same class of keyboard, different one? Same class of keyboard. It's a slightly different keyboard, but it's fairly similar. This one supports in-key rollover. It also has the ability to change colors, and it has three different zones. So it's an RGB keyboard. The color stuff is for gamers, no? Oh yeah, for gamers for sure. I leave it off because it uses power obviously to run your backlight, but yeah, you can set it to full RGB. I don't know if I would give a demo of that right now. I probably could. You can have some apps with some interesting notifications with the colors and stuff maybe. What's your, what's called, impression of these guys? What are they really good at? Are they making the coolest Linux laptops in the world or? So I'll give you my honest opinion. I always assume that anybody who's trying to design hardware for Linux would actually have a fairly, not very polished product. But look at this, right? This is pretty good stuff. The reason it's pretty good is because these guys know what they're doing. I've had a chat with these guys about certain bits of the technology they use for doing firmware, maintenance, image builds, deployments. And they seem to have put in a lot of thought around security, deployment securely, those kind of things. So they know what they're talking about. And he's also the creator of the Redox microkernel written in Rust, so he really does care about security. What is that? Right, so we're going to have a talk about this tomorrow, but do you want to do this? Yeah, Redox is a microkernel operating system written primarily in Rust, which is a new language designed around security. It is designed to be extremely reliable while still supporting POSIX applications. So where would that run in this kind of hardware? Where is it for? It can run as a desktop on something like this. I can boot it up pretty easily. I was just there. I keep closing with it, so I have to log in every time. So how much of the hardware here is completely open? Sometimes, for example, as far as I heard, Intel has some kind of... I don't know if it's called back doors or something. People don't want to have this. They want to have full control over everything. Yeah, so there was a lot of talk in the past year about the Intel management engine and the security vulnerabilities that were found in it. So for quite a long time, we didn't do anything with the Intel ME, but once the vulnerabilities were being disclosed, very quickly we finished with our new firmware update infrastructure that we'd already been working on for some time and deployed firmware updates for all of our laptops to disable the Intel ME by default. Nearly two dozen models. So two dozen models for the past, I think, four years received firmware updates that disabled the Intel ME. And we did this using an open-source firmware update infrastructure. We are developing more open hardware components, and we'll be releasing those very soon. It'll be a huge release. This is one of those components. This is a backplane. It contains an AVR microprocessor that is the embedded controller. The firmware for that is completely open-source. It contains a completely open-source hardware design for the backplane. It also has fan control and power button control. I'll flip it around and you can see the SATA ports. And we're also doing a model that is... Let's see if I can pull that up. Give me about one second. Is this a secret? No. No, it's not a secret at all. We're going to be releasing this very soon. This is SAS, so we can do U.2 as well. This was a bitch to route, by the way. But once routed, we're going to have this in our highest class product. We'll have U.2 supporting extremely high throughput. SAS is a specific kind of storage that people have in laptops, so nobody has that in laptops. This is not in laptops. This is for our desktop. We're beginning our hardware efforts with desktops since there's a lot of overlap in terms of what you need on a desktop. You also need to be able to do on a laptop. But there's additional things on a laptop that it's easier to learn how to do this on the desktop and release all that in open hardware in the meantime and then apply all of that with a few other things to building laptops down the road. For laptop design, what we have going on is we have an embedded controller framework that we've developed that we're using on the desktop boards that you've just seen. That framework allows for battery control. It allows for control of back lights, keyboards, and all the peripherals that are common on a laptop. What that will allow us to do is to embed that design onto a motherboard with whatever CPU we end up deciding to use and then basically get a modular laptop going where you have a chassis controller that controls the laptop functions such as the power button and keyboard and touchpad and battery, and you have a separate module that controls the functions, and they communicate over actually USB-C. By doing it in this method, you can swap out either of those components. You can move your computing component to a different laptop, or you can replace your computing component very easily with the same laptop chassis. Is there any chance that the ARM ecosystem is more secure than these bugs and stuff that Intel has had with those backdoors and all that stuff? Let's say if they choose an ARM processor that doesn't have graphics, then all these ARMs, they can always do some kind of PCI graphics that could fit in a laptop. So to your first question, I think it's very hard to in a quantifiable way say whether one architecture is more secure than the other. I think what's happened is in light of recent events in this space, in the security space, all architecture designers are going back to the drawing board in a sense of making ways of making their architecture more robust, more secure, and ARM is going to be at the top of the pack, just like the others. To your second question, having an open source option for GPUs is an interesting question in the ARM space because of all of the legal wrangles associated with the IP around GPUs. But I like what System76 are thinking of, which is like, they understand GPUs very well, they're already using NVIDIA GPUs. They understand busing protocols and architectures like USB-C, PCI Express very well. I see no reason why they could basically compose a desktop workstation or a laptop design where they use PCI Express and whatever well understood GPU architecture that they're comfortable using with that. Because when you specialize in safety, that means that's a big topic for you then, the whole safety. That could be in these kind of devices too? It could, although the focus for me is actually in so called safety critical domains and those are more closely associated with things like automotives and robotics and healthcare and those kind of things. I would say one very important part of security for us and security for users is being able to get security updates, being able to get the latest version of an operating system continuously through the life that the system will actually be used for. So for us that means that the SOC that we choose would have to have long term supports that it works with every version of the Linux kernel from the time we start selling it through the time that the hardware dies. You want to get a picture of these? Oh, they're using right here, 76 in the wild. What do you think about it? I really like this. It's got everything. I really like all this port in here. What do you have on the ports right here? You have a real port, a USB 3. I'm here on the other side. You both have both a headphone and a microphone jack. Another USB 3 and you both have display port and HDMI. So you have you can choose Type-C maybe 3 speed maybe and a gigabit internet and SD. Most people don't have that. Why? I don't know. Do you have many, many customers? Who are your customers? Are they people that just like Linux and are hobbyist? Or are they serious people? A large number of our customers are serious people. So a majority of our customers are businesses that need to buy large amounts of Linux laptops for developers and entire teams. Then we have maybe 40% of our customers being individuals or people who order through our site outside of our sales channel and aren't gathered into that kind of enterprise business group. And those are people who are passionate about Linux and want something that works out of the box. So it could be I'm not going to say CIA but some kind of top security firms. We have government customers. We have universities, businesses all using Linux. Because they want to be able to know what's happening in their laptops? They want some assurance that they'll get support around Linux. Whereas if you buy like a ThinkPad such as many of the ARM people have you don't get support from Lenovo based on or from IBM based on that ThinkPad's support to Windows mainly, right? Whereas we provide a lot of effort into first making sure that it works correctly out of the box with Linux and then providing technical support for both whether you have problems and whatnot. We'll get you going, we'll help you figure things out and if customers do report problems to our technical support side we on the engineering side hear about it very quickly and address any issues that do come up. So you have customers at Lenovo, at Red Hat, at Suzy, all these places and they kind of help you or do you have many flavors of Linux? So we for over a decade exclusively offered Ubuntu on our systems. Today we offer both Ubuntu as well as our own Ubuntu based distribution called PopOS. We provide support for many other Linux distributions so we sell laptops with Ubuntu or with our Ubuntu derivative PopOS but we support the laptops after they've gone to the customer on many different distributions and there are maintainers for the code that we write for supporting our laptops for Arch, for Fedora, things like that so it's easy to get the driver support. I mentioned support, support is a big deal. It's not easy to do support and you have a feeling that these guys are able to do all the support? Of course, they've been doing it for a long time and they ship to 60 countries I think so you know like you can't run an operation at that scale unless you actually thought through the ramifications for support and engineering right? So how many people in the company? We have 24 people now we just hired a few more people and we're scaling up operations because we're moving manufacturing to our location in Denver, Colorado and that operation is going to start manufacturing desktop chassis and assembling desktops which presently our desktop chassis come from China. So what is the change to do it locally in Denver? It changes, we have complete control over the design and how it's made and we're able to actually reduce our costs by using more automation or quality by having automated mechanisms for ensuring quality. So how old is the company now? The company was started in 2006 I believe. But there's a potential that things could explode and you could be the next Dell? I'm not sure that we'll ever get to that point unless Windows dies a horrible death. It could happen. But that's our major struggle and continues to be the struggle of Linux and that's to really conquer the desktop and to figure out why people keep moving to Windows even though they have ads in the start menu now.