 So we're right here at the Nenaro Connect and you are just getting a whole bunch of, what's going on here? These are the motherboards for the new developer box that has been produced with SocioNext and Lenovo and Gigabyte. So who are you? I'm Daniel Thompson. I work at Lenovo. I do support and solutions engineering. Support means I'm going to be helping people use this and solutions means taking on small projects to make it more awesome. So right here we're getting the unboxed, so can you carry it up? Yeah, so this is the SocioNext board. Yes, so it's a PC-like, it's basically, 1998 was the last time somebody made a PC-like device with an ARM processor in it. So this is a big difference. So the last one I know of was the RISC PC which they stopped production in 2003. And then right here it's out of the, can you show it around? Can you explain what's on it? Oh, okay. This is a micro-ATX board and under this heat sink, here is a SocioNext LSI. And for the DIM, for the Nenaro 96 board we use, we put only one UDIM. But you can put four? Yeah, maximum fourth lot can be used. So how much run total max? Max, 64 GB. We've been running them here at Kinect with 32. 32, is it okay? Yeah. Is it enough for doing some stuff? Yeah, it was enough, yeah, we could put everything in a RAM disk and let it really run. And this is PCI? PCI is a 1, 2, 3, a 4M, 4M, 6T LAN. So what's the difference between the small and the big ones? What could you do in different ones? So that was 1 and 1 and 4 I think, actually. Yeah. Those are one-lane, those are four-lane. It's a 16-lane slot. So you can plug a graphics card in and it will drop down to the right amount of bandwidth. But technically it's hard, it's a four-lane PCI slot. Let's check out one of the boxes you've been working on over there. So right here there's an awesome desktop machine right here, right? Yes, we've got two of these boxes. These are not the boxes it will ship in, these are just the boxes we're doing engineering in, but it will ship in a PC-like box. This is a 24-core ARM Core DEX 853 desktop PC. Yes. Which will be great for developers. It will. It's there to develop arm-on-arm. So instead of everybody cross-compiling, in fact people still call it cross-compiling. They're so convictuated to saying I'm going to cross-compile my kernel that I've seen people compile on this board this week and still claim to be cross-compiling even though they're not. They're not cross-compiling anymore. It won't be a thing of the past. It won't be a thing of the past, it will continue on. I'm not going to claim that it will. It will revolutionise the world, but it's going to be slow. And your colleague right here has been doing some stuff? Yes. What's he been doing? So that's a lot of firmware work. The nice thing about this board is that it does feel like a PC. It's running the same EFI frameworks that you would get on a regular PC. So when it boots, it shows you what I think of secretly as the BIOS menu. It's not been a BIOS menu for 10 years, but I still call it that. You can see the BIOS menu. You can use your grub boot litter to select what you're booting. So it makes it very, very natural and very normal. Very boring, if you get right down to it, but boring has been necessary. We will get exciting afterwards, but right now we want to be as boring as we can. You want to get it out, you want it to just work. And you got the first one just very, very recently. As I understand it, it was Thursday. So it's Thursday today. I think the Pink and Place machine was assembling it. I think it was a week or two weeks, but one week. One week. So yeah. And you've been able to bring it up, basically, right? So you were at the, like there was a keynote, was it yesterday? The day before? Yes. You're already showing everything, like some stuff just working. Yeah. You're uploading YouTube videos. Yeah, yeah. The YouTube video is great. I mean, it's called Ramsay. Is it hardware-accelerated YouTube video playback? It wasn't hardware-accelerated in the demo, no. So it was? No, definitely not. It was not? No. But it still plays the video. There's no video accelerators on here. Obviously, when you plug in a graphics card, any acceleration that we have Linux drivers thought on the graphics card will be exploited. But that's usually quite, it's not really necessary much on modern things. What's the board right here? Between this and that. This is just the board separately? That is one of the... So Cinex concept was to create massive, massive scalar systems. So they had, I think they were targeting 1024 cores in one washing machine. And that's the board that sits inside the washing machine, scaling everything out. So they have a big fabric and they plug lots of those boards in. So what the developer box is, is one of those compute nodes that you can have at home, both to do fairly... You can do some interesting parallel algorithms with 24 cores. You start to hit the bottlenecks at 24 cores that you used to see in supercomputers five, ten years ago. How's it compared in performance with Cortex A72, if you have like quad core A72 compared to 24 core A53? We're still finding that out. My gut instinct was it's equivalent to about eight A73s, but that's pure gut, it's not been measured. So that was my gut feeling. So there are plenty of interesting boards in the army ecosystem now with four A73s. The Macchiato bin is one, it's an awesome board. And it will be interesting to see how they compare. I don't know. But then it's a whole bunch of smaller cores, maybe it could be very useful in terms of power consumption running a whole bunch of stuff in the cloud, right? Yeah. It depends what kind of workloads... It depends at five watts, which is astonishing. I mean, at that point you start having to tune the system. The core uses so little that you need to find low-energy power supplies, you need to find low-energy RAM, because otherwise you're spending five watts on your CPU and 20 watts on your power supply inefficiencies. So this one is all loaded up with... it's connected with stuff up here or... This is a GPU right here. Yes, that one's been plugged in. That's actually, I think, the card that we ran the demo on earlier in the week. That's NVIDIA. That's NVIDIA 1X. It's designed for Bitcoin minus, because it's single-lane. Bitcoin and GPU. How about the radian? Did that potentially work? It could potentially work. We haven't... Well, we have plugged one in, and we were struggling to get it to run, but that's because it was warm. We just picked up the fries. We've no idea if it's even going to run on a PC at the moment. But, yeah, yeah. So what kind of stuff have you been doing otherwise? You've been working together on this kind of stuff? Well, I do the splash screen, which was very important. I help kind of get things together. I'm not one of the key firmware developers, but I'm so excited by this that I volunteered to come in and help with the Sprint. So the Sprint started on Saturday. I was personally so excited by the product I wanted to come in and join on the Sprint, which means I'm a bit of a fraud, because many of the other people here have been working on this for six months. It's the genesis of... The Enterprise guys have been working on Enterprise ARM at Leonardo for four years, maybe. So this is the genesis of a lot of lot of work in other parts of the company, and I'm just joining in the ride at the end. But I'd like to say it's so exciting to have this on the desk that I'm charging into it. So I'm going to take it home and work on the recovery. It's significant, it's exciting, it's special. It's a new form factor. ARM has been in phones, it's been in Chromebooks, but this is a new form factor. This is more like... This is the final design, right? It's going to be more like this size. Yeah, so this is slightly smaller. It's hard to see on the screen, I suspect. Can you show the ports and stuff? So there's USB in front. So yeah, we've got... These might get blocked off, I don't know, in the final product. So this is USB 3. It's got two USB 3 at the front and up here, two at the back. Obviously, because we've got PCI, this one's got extra USB slots in. So we can get all sorts in there. And how fast is the ethernet? One gigabit. One gigabit? Yeah. Single, I think you'd say. We will market a single. So this one is a... One is a networking port. The other is part of the interconnect fabric. So the main CPUs can't use it. So there may eventually be interesting use cases for it, but the initial stuff that comes out in December, we will be marketing it as single. Nice. What else is going to happen with this? What do you think is going to happen? Is everybody in there going to get one? It's an open source thing. So I think it's a bit old school. What's going to happen is we're going to see this on lots of developers' desks, and developers are going to write software on it. And that means I don't know what's going to happen. I know what I'm going to do. I'm going to go home and start working on the recovery OS. So there's an EMMC on there, which is not normally there on a PC. You can do interesting things like that by creating a recovery operating system that's not in the main hard disk. So if you think about the PC, modern laptops come with a recovery partition that you can accidentally damage if you destroy your GPT table. We won't have to do that. There's an EMMC on there. We can put the recovery into the EMMC. So that's what I'm going to play with when I get home. But the exciting thing is I don't know what anybody else is going to do when they get home. Can you sit down just one second? Yeah. This is very exciting what Social Next is doing, right? Yeah. They've made an interesting product. It's a really interesting concept. Running little cores, the one gigahertz cores, and trying to get massively parallel is an interesting approach. I don't know if anybody else is doing it. If they have this DDT, they call it to connect them all together somehow and get huge, some combined kind of performance. The point is that these are the cheap boards that I get to play on. I'm afraid I shouldn't get to play on the big massively parallel ones, and I have no idea how they can perform. But there's a very interesting architecture. So yes, this board is for benefit of the ecosystem as a whole because having PCIe on there means that we'll be maturing PCIe on ARM. We'll be maturing various of the EFI firmware. Everybody will start seeing ACPI on a machine they can buy really cheaply. So Lenaro's been working on ACPI for a long time for Enterprise Systems, but they were all systems that cost tens of thousands of dollars. We're now going to start having a board where we get to experiment with this one for a slightly smaller scale. Will you invest in the thousand-dollar for a system? That's what I'm told. I'm an engineer, and I will get one on expenses, and I'll be very happy. Cool. And how soon is mass production? How soon will everybody come back? It can be available in December.