 And here the NXP, we have the IMX8 right here. Hello, so who are you? Hey, I'm Arnaud Vanen-Bosch. I'm in charge of the IOTMX in terms of business and marketing for the automotive application in Europe. So the IMX8 is a very exciting new chip, right? Yes, exactly. And here, the one of the example that we're showing here you is not only doing everything about infotainment and cluster in here, you can have, I mean, some neural networks for it only directly. And here is really for the new technology when we're looking at artificial intelligence and not to have something which is pre-programmed. So neural networks, so it's like AI kind of stuff? Exactly. So we have learned the GPU to recognize a few signs. I mean, and we just send them in 40,000 signs. And from there, afterwards, they can recognize any type of signs directly. And what we're trying to do is to rate at least 97% of accuracy. So actually you have three IMX8, right? You have IMX8, 8M, and 8X. What's the difference? So the 8 would be really for the ECOPD solution that we'll show you just afterwards. The 8M is really much more for the application and consumer. And the 8X will be in the lower end devices, much more on a digital display, digital radio system. And right here, we're looking at the 8, which one? The 8-quad mark, the biggest device that we'll be able to reach. Is that 8X? That's 8. 8. So that's the biggest one right there? Yeah. With all these functions right here. This is a 64-bit arm, right? Yes. Let's go. 64-bit, what are the cores you have? So what do we have here? I mean, we have 453, 272, plus we have accelerators up to 128 gigaflops in terms of GPU. We have VPU, and up to, I mean, we are decoding H.265. One of the key things that we're showing here is we can have four displays, 1080p, I mean, running concurrently. Here we have two different domains, and safe domains for the cluster, and an input and domain, with all you have, like Android or whatever, you've got the car play. So two separate domains, which is very important in the core, because you cannot mess up from the application, I mean, compared to everything when you're driving. So all the information, driver information from the cluster, which is ACI compliance. So you have Android, SOC, you have two drivers. Exactly, exactly. And that's one of the beauty of the thing, because if one crashing, it should not have any impact on the one which is not the infotainment domain, but the one which is the driver information. And one of the key thing that we're doing here is the full separation of the system, with really clear separation in the domain, which makes the full system very safe. And so there's no need to have two SOCs. Do you just have one? Do everything. One SOC that will do the entire cockpit inside the car. As I said, here you've got the infotainment system, you've got the HUD, you've got the cluster here, and here you've got the map, for example, I start the route. So this is your infotainment system, you've got the route. Here you've got your cluster, and in the HUD, you've got exactly the same information that's going here. And then you can still play some infotainment. Everything, all the four displays are totally independent, and driving independent information to the four displays. So it's not just the exit cord, there's also four Cortex-M4 in there? So you have two Cortex-M4 inside, so it's 453, 272, and two Cortex-M4. And the Cortex-M4 can be used in different ways, either for safety mechanism, or for additional audio preprocessing, additional cameras, fields, or whatever. Let's go back to just right here. Also that, you know, it's been trained. It can recognize the stop sign. So stop sign here, we just train them to recognize stop sign, but it's not only that, the other thing that we want to do is, for example, face recognition, drowsiness detection, all this type of specific algorithm that will not have any safety impact, but still giving some information to the driver that something's happening. So it's really in addition to all the safety domain that we are developing on some other elements within NXP. And if I look at here again, it says 128 gigaflops. That's a very powerful GPU as Vivante. Yes, that is a Vivante, the GC-7000 Vivante, which is a new one, and which we are giving, I mean, not only, I mean, what we have done here is, we have separated in two GPUs, so we have two different GPUs, and that's what we're showing here, is one GPU will really doing all the neural network, and another GPU is doing all the different camera feed. So there's also GPU compartments. Exactly. And you have all kinds of power management in terms of, you can choose to only power sound chips, not others? Exactly, so what we're doing, so exactly we'll have different power domain on the chip. For example, the cortex will be different power domain compared to the GPU, so you will be able to scale the power of the GPU independently to the CPU. What you will be able to do, for example, if you want to shut down everything and just keep the cortex in for the low power mode, you can do that. Which means that the system, you can put in sleep mode the entire system and wake up very, very easily with a very low power consumption. Is that exclusive to you? Nobody else has done this before? I mean, it's something that we are working a lot on this low power mode because when you have this type of system, especially when you start to separate them in different domains, you need to have this type of different power mode. And that's very, very important. That's one of the unique feature that we have, these two different domains, totally separated. Totally separated? How do you power each one? How does it connect? And you say, I only want to power them four and nothing else. How does that work? So you have a small system controller unit that would make this type of decision, which is pre-programmed and the customer have also access to program it in order to make sure that the different assignment, for example, the peripherals are assigned to the different cores. As well as the partitioning of the memory. So we are partitioning the IOs, peripherals, as well as the memory. And afterwards you have different ways to modulate that through a system controller. So maybe this one, this demo is maybe only using the A53 or maybe it's using the S72, but you can choose. Exactly. You can use some, you can use the other one. Or you can virtualize the Cortex A because everything which has done with the ARM Cortex has been pretty well done in terms of virtualization. So it's good to have kind of an app version on top in order to have good load of the balance in order to have a good power dissipation or some good power consumption. And this is GPU compute under Vivante. And the GPU compute, so here what we want to do is really the GPU only is taking care of the 3D. And after we have some additional 2D cores which will do all the 2D effects, meaning the different layer that you are blending together, how you put them in a different area on the display. So this is all the thing that we have a specific 2D cores and the 3D cores we only do 3D. If I look over here a little bit more, it says this 4K H265 encode decode, right? Both decode and this 24 bits audio, lots of things people want to have. What kind of DSP do you use for? It says High Performance DSP. So we have the High Performance DSP from Tensilica. And this is, I mean, running at the 666 megahertz which makes about 2.5 giga-max available for audio post-processing, which is more and more important in the automotive domain because you can have Activeness Constellation which is running on it without any taking on CPU MIPS. And it says here Full Chip Virtualization with hardware isolation, yeah. What is that about? So that's what I was explaining for the other one. You can isolate two domains. You can isolate the domain for the infotainment versus the one which is for the cluster, which is safe domain. And it will be physically separated because you will have the two GPU, the two display processors which are totally separated on which you will not share anything in terms of domain and each OS will address its own GPU and its own display processors. So when you talk about this, you talk about automotive, but what else is gonna be the target? I mean, the thing that when you look at the health, health is very important because you will have some infotainment but you need also a secure domain because you don't want to that the infotainment domain to mess up with the healthcare of the people. So that's one of the key point that you will have. Any type of transportation in the train or in the plane, this also you need to have a safe domain versus an infotainment domain. Here, another way to make it that we're showing here is for example, you have a domain for everything but I would say neural network, but you don't want to have that, I mean messing up with your another domain. So that's the type of thing that we want to do is really when you need to have two separate domain running on a single SOC, all this type of application can be running on this type of SOC. But it's big little? So if we have this A53 and A72 in a big little possibility. And heterogeneous multi-core, this is with the, specifically when we have the M4, on this M4 you can free RTOS, AutoSAR, whatever, which will make I mean the low power mode versus the Cortex A53, A72 that will do the high level OS. And is long term support? Long term support as usual, automotive 15 years on JVT, industrial 10 years on JVT, as usual that's what we're doing. So people hopefully will be watching this video in 15 years and it's still gonna be irrelevant. Still be irrelevant and still be supported with the, from our supportive. So when we're talking about self-driving cars and all that stuff, is this something else? So this is part of it. This is not the self-driving car because it's not necessarily, I mean it will not do any decision-making, I mean which will really have an impact on the driving but it will helping because for example, it will take care of the image stitching thanks to the GPU that will do a very smart stitching of the image. And this stitch image will be passed afterwards to the vision coprocessor that will do all the analytics in order to take a decision at the end. So this is part of it but just one single piece of it. And what's next with iMac State? So you announce something, you showed it and now it's a chip out. So for the Adotimix 8, we'll have this one, I mean you can see it's working. We have the first sample. We have the new samples in the next few weeks. Everything is already in the fab and that is happening right now. So within this year, we'll have all the different silicon already available for the customers. Next year will be all the qualification. And development boards and everything. Everything will be available this year. Well.