 I'm Ryan, I work for ARM, I'm a developer community specialist. I work mainly on the Mali developer.arm.com website. So that's where you can go and you can get a lot of free resources, developer guide tutorials, all on the Mali GPU. So how do I do my labs? Ah, probably all of them I expect. I have no idea how many in terms of physical number. There's a lot of things you can optimize using this. Yes, definitely. If you use applications correctly, if you know how to target specific ARM and Mali hardware, you can do that through a lot of GPU compute, for example, if you're doing OpenCL, which is something we're showing over here, or if you are looking at things like Vulkan, which is the new API from Kronos. So you have a Vulkan integrated right now on the Exynos 8.890, this is 8.53. That's right. And there's Samsung M1 in before, so it's a big little core configuration. So M1 is the big core from Samsung, so that's why they're our own CPU cores. Yeah, custom big core, now pairs with the A53 little core. And then we have the GPU, which is the Mali T880. That's got 12 cores of graphical processing. Can you show me this half? We're showing this, which is an ARM internal demo. So the demo is actually our Vulkan technical demo. You can see it's running through at the minute, or we can also just go and control it. So you can see it's a whole cityscape. We can see all the bandwidth and performance being measured there. It's all running in Vulkan. So this is a proof point of the Vulkan API running on ARM Mali mobile devices. With Vulkan, you're able to get a lot closer to the hardware as a developer, so you can actually tell which bits of hardware are being activated to run different parts of your application. And by doing that, you're able to control much better battery and performance. So you can get performance when you need it, but also you can get great amount of battery performance. So you're actually quite efficient as well. It's like more granular, more detailed API, so you're not like just putting a bunch of GPU resources and stuff, but you can control it better. Yeah, exactly. You've got a lot more control as a developer when you're using Vulkan. You have traditional APIs like OpenGL, ESI, OpenGL for desktop. And the GPU computer is super busy with that. Yeah, GPU computer is really taking off. You'll have heard about ARM's acquisition of Apacle. If you've seen it, there's a lot of videos aligned for IVG as well. So they're here at TechCon as well. But also we have an abundance of partners for computer vision, so one of which is Luxoft. And tied in with that is a lot of deep learning and deep neural networks. So in this example here, we have a deep learning occurred offline. So there was a database of about 1,000 objects. And as you see here, we're now running the application just on a phone. And as it goes through, we're doing a detection of what these objects are as the camera sees them. So it says as the multi is done. Yep, exactly. And now we're going golf balls. So that's referencing the algorithm, and it's running here with OpenCL, which I mentioned for computer vision. And it knows it's a mouse, it's quite confident. Yes, it's very good. Isn't it awesome? Yep, and if you look at the CPU here, so we're currently running the algorithm just on the CPU. You can see a load of about 58% there. If we change now to the GPU, change to the GPU, we can see a drop down to about 16-20% CPU utilization. That's awesome. Is this ready? So this is demo running on Armour Marley natively. There's also an application from Locksoft, which is already available to download from the App Store. So you can see it using their cloud system for the image recognition as well. Yeah, I think so. We're really interested in engaging with ecosystem partners. We have a whole massive ecosystem for computer vision, gaming apps and browsers, even emerging technologies like clamshells as well. We're very committed to supporting any app in the use case. Yeah, just like any use case with Armour and Marley, we want to see it used in the most efficient way. So you've got, in the case of a clamshell, we want to make sure they're picking appropriate CPU-to-CPU combinations. We offer a whole suite of IP there. And making sure that those apps and games can be optimized for that different use case. They're full-factor.