 So we're here at MediaTek and here's a tricluster CPU architecture. It says 30% less power, consumption for Facebook, for Temple Run, Skype is nearly half the power consumption. So this is inside Android, this is power consumption. Is it real? Huh? It's real. Compared to what? It's a comparison between these two architectures, this is a conventional big detail, this is a new architecture in X20. So is this, this is ARM Cortex A5753 octa core? Oh it's based on the same CPU. Same CPU? Same CPU. Alright, so how soon is this in the market? In the end of this year. And it has a 64 bit tank core, so it's quad A57, quad A53 and quad dual is a 272 and a 453 and the other 453. Ah 453, 453 and 272. So there's a medium 53, the higher clock speed or? Yeah mid is higher, it's 2GHz and the little one is 1.4GHz. So why is it the best solution to have a small and big 53? You already do this before? Big 53 and small 53 right? Yeah. It's not big little, is it big little also or no? You mean? ARM big little. Is it similar? Is it using ARM big little technology? You mean X20? Yeah X20. It's modified from Big Little. Actually we have 3 clusters. Big Little is a pure cluster. Yes and this time we do 3 clusters. So is it exclusive technology, MediaTek is developing this? Yeah, yes. We develop by ourselves. So what more can you show here? It has a CAT 6 LTE built in. So it's higher performance. Is it possible to play all the 10 cores at the same time? Yeah, yeah. So here you're showing also some other things about the advantage. Light, medium, heavy? Yes. It's the same concept from the auto industry. So shifting gears? Right. So it's coming at the end of this year? Yeah. Before the end of this year? Yeah. And what is the core pilot 3.0? The core pilot is technology. We control the CPU and the GPU together. The core pilot is 1.0, 2.0 and this time we migrate to 3.0. Is it heterogeneous multi-processing? The heterogeneous? Yeah, yeah. Is similar to inside this? It's enhanced. 2.0 is the heterogeneous. And 3.0 is even more? Yeah. It's a tri-cluster concept. All right. So what are you showing here? It's a conventional big little and this is a tri-cluster. And you could see the performance line in the middle, the difference between these two curves. The orange one is the power reduction. All right. So to develop this is using the same code a little bit as a big little. Linux and Android optimized in a similar way. So how do you do that? You have a smart scheduler, right? Can you develop all this? We have done some hardware design to face switch between each cluster, face switch and some scheduling to optimize. Is this possible to do this automatically? Or you have to tell the application developer to change the apps? No, no, no. We don't do it. Base level. What do you call it? We will download it in the button. Transparent to the application. All right. Cool. So you have already tape out or not yet? How soon? Is it testing the chip? We are going to tape out soon. Tape out soon? And so far you're very confident it will work, right? Yeah. But you will do some tests and then after you will chip maybe. Cool. For sure. And there we have mid-minute max of difference in Gmail. Very low power consumption.