 Hi. Welcome to CES 2020. I'm here at Texas Instruments. I'm one of the business managers for our Jacinto 8S processors and today we are going to talk a little bit of our Jacinto 7 devices. So this is a new platform family and really the first two areas that we are introducing is our DRA8 SoC family which is really intended for our gateway inside the car as well as our TDA4 family of SoCs more intended for our 8S portion of the car. So it's high performance and also real-time right? Absolutely and these devices actually will sit in the car and will be doing a lot of the analytics real-time. So it's an ARM Cortex? The Jacinto 7 family showcases an Cortex A72 as well as our 5Fs. In addition to a number of other compute cores which not limited to the TI DSP, C7X DSP, our MMA and several other accelerators that really help achieve that real-time performance for these applications. So is this targeted for the self-driving cars or? So you can see here that really a lot of the 8S applications kind of bucketize this into two groups. You've got smart parking and then smart driving and you can see here we've got two demos that here we have the TDA4 SoC that's showcasing a front camera. It's a smarter driving application. It's able to identify objects on the road and classify. Is it a car? Is it a road? Is it a pedestrian? Right? Is that for a full self-driving system or just for enhanced driving? So you know a lot of it is getting the enhanced driving. So you a lot of systems today have very limited 8S functionality. If you look at high-end cars versus low-end cars there's a pretty big gap in terms of what type of 8S functions are available. And so features like this allow our tier 1s and OEMs to really kind of bring those features across an entire fleet of cars. And so that's what kind of what you're seeing here are some of these functions that now can be enabled based on the just into 7TDA4 class family because we offer the scalability for very high-level processing compute towards kind of more a higher levels of autonomy as well as more cost-affording affordable. It's even become kind of mandated to have this kind of backwards facing smart cameras now because it's for safety and also very important. Yes, if you look at, you're right, safety is a big piece of this. If you're backing out from a parking spot is there a pedestrian behind you if you're coming out of a grocery store? Is there a person kind of moving a cart? So these types of features are very, very useful. A lot of customers are really demanding this across more cars. But it's not just for backwards and reversing. Also in parallel parking situations. And if you move to other countries we're starting to see more valet parking for example where the car is actually parking itself even without a driver there. That's so cool. What are you showing here? This is a demo that kind of shows a... The chip is there? This is the EVM platform right here. The chip is right underneath... This is development kit? Yes, this is the TI development platform. That's correct. And if you look here on the screen this is 8 megapixel capture data and doing multi-class object detection on the road. And then from this data it's actually taking that and then building a localization map. So it's taking this data along with GPS and IMU CAN data and then being able to actually do a full localization. It's doing all of this on a single SOC. And if you can see here passively cool. There's no fan. So if you touch it here... Nice. It's not really warm. And so how much for example does the DSP do? Or the other accelerators that TI is famous for? Right. So algorithms like this that you see here really kind of take a fusion. A lot of customers are fusing both computer vision as well as deep learning and putting that together. And if you look at kind of what these algorithms require it's they're very very math heavy. So these accelerator blocks that we have inside of the chip really allows a lot of these math heavy computations to happen in parallel very fast and very efficiently. And that's what really enables kind of these solutions to be a lot more power efficient as well as get kind of the performance level that you need for low mid and high-end systems. Does it have GPU also right? It does also have a GPU. And it also does a bunch of stuff on the GPU. Right. So for example here's an example of a car rendering here where we actually this is part of what we offer in our SDK development kit today which is in some example starter code. What does this do? Assimilate it like a car? It is assimilating. So you can see that this is kind of showing four cameras stitched together to create this 360 degree surround view as well as a GPU based rendered car. And as the kind of the car is moving you can kind of see how the image gets recreated and stitched up here in the screen. But this code is offered part of the standard SDK so that customers can really get their applications running a lot faster. It's possible for you to do a full self-driving car like this with this solution. This code will continue to scale so algorithms get better and our customers demand more. We'll be able to meet the needs of the market. What is over here? Okay so yeah this one right here actually shows this again based on our TDA for EDM right here and what we're showing is eight 2 megapixel cameras all shown off of a single chip and the chip also has as one of the accelerator blocks and ISP and all these are handled by a single chip on the EDM. Hi so my colleague just showed you how Jacinto solves ADAS applications. My name is Katrina Twazen. I'm from Texas Instruments and I'm going to talk to you about how Jacinto helps solve vehicle compute systems and gateway systems. So what are you holding there? So this is the new Jacinto device. This is our EVM. This is available on TI.com. So development kit? Yes. And what is a compute gateway? So as more data, there's more and more data moving through the car as we move to self-driving vehicles so that means that instead of tiny MCUs moving data throughout the car you're going to need big powerful system on chips or SOCs. So this is what a vehicle compute system is, is multiple compute systems communicating to each other in a domain. So what Jacinto is addressing is integrating external features typically found in a vehicle compute system into one chip and using and routing data throughout the vehicle compute system. So does that simplify the car also? Yeah, it simplifies both the car, the board design and also saves on bomb cost. And what is running right here and there? What's showing here is what we had previously talked about in this diagram. So this is three compute SOCs, in this case three Jacinto devices talking to each other through our integrated PCIe switch. So they're routing data via PCIe using our PCIe switch. And Jacinto is the first SOC in the market to fully integrate a hardware PCIe switch into the device. So that means you can scale, you can have as many as you want to accelerate things or? So you can scale from just talking from one SOC to another SOC or if you want a really big powerful compute platform system you can talk to multiple SOCs so anywhere from two to four.