Sunchip shows some of their latest products including their all-in-one AR system based on Rockchip RK3288, RK3399 board for digital signage. $58 Amlogic S912 TV box, $26 RK3126 HDMI Stick, $200 Apollo Lake Box. 360 panoramic camera. $7 single Smart Wi-Fi Plug, $11 for dual Wi-Fi plug.
Actions S900 is a Quad-core ARM Cortex-A53, the Actions S500 is a Quad-core Cortex-A9. Actions also shows some other products for video smart speaker, smart door bell, smart kids robots, Actions also provides their MIPS based chipset for Bluetooth speakers, here also they are launching their new ATB1103/ATB1109 ARM Cortex-M0 system for their upcoming Bluetooth and perhaps also low cost smart speakers platform.
Allwinner Smart Home solutions including their chipsets and development kit for the R18 which I also filmed here the $26 (169rmb) Allwinner R16 powered Xiao' ai Classmate Smart Speaker Mini by Xiaomi to be delivered in over 5 million units quantity, smart speakers by DIngDong. The Allwinner R6 enables even lower priced smart speakers with a maximum of 2 microphone in the microphone array for devices that will be closer to the user. The Allwinner B288 enables E Ink e-reader supported by DongDong which is China's leading online book store. The Allwinner MR100 powers the fully open source drone project by Kudrone. The Xiaomi Drone uses the Allwinner R16 too. The Allwinner FC1600 powers the Niutingting Early Learning Machine by Benew which reads stories for kids for 2-6 years old. Allwinner T3 for car dashboard and smart mirror, T8 for in-car entertainment. The Xiaomi Robot Vaccum Cleaner uses the Allwinner R16 and uses lasers to calculate distance to walls and navigate to vaccum and mop the floor.
Allwinner V5 chip uses the HawkView 5.0 image processing engine, which integrates 2D/3D intelligent noise reduction, frame width dynamic synthesis, sharpening enhancement, specific color enhancement and other image pre-processing technologies, combined with various types of image scene requirements, for each frame of the picture it applies real-time fine detail recovery. This chipset is for the "real" 4K action camera and 4K-enhanced security camera solutions, where smart object tracking can enhance the resolution only of the moving objects to save on bandwidth. Also the Allwinner V5 encodes using H265 to provide for about the same quality at about half the bitrate compared with H264.
The new Allwinner VR9 is quad-core ARM Cortex-A53 processor with an independent 32KB L1 I-cache + 32KB L1 D-cache, shared 512KB L2 cache, with MaliT760 GPU and support for OpenGLES 3.2 and OpenCL 1.1. Allwinner VR9 is the first SoC optimized for VR, with as low as dedicated 20ms low latency acceleration module portal, dual-engine direct drive dual-screen system, panoramic 6K visual effects with support for expansive camera positioning, up to 6K30 H265 and 4K60 HEVC/VP9 video decode. Allwinner's hope is that this SoC can be compatible with the Google Daydream platform for All-in-one VR device usage.
Emdoor is one of the largest design houses in Shenzhen, here they talk about their support for the new Allwinner VR9 quad-core ARM Cortex-A53 processor with an independent single core 32KB L1 I-cache + 32KB L1 D-cache, shared 512KB L2 cache, with MaliT760 GPU and support for OpenGLES 3.2 and OpenCL 1.1. Allwinner VR9 is the first SoC optimized for VR, with as low as dedicated 20ms low latency acceleration module portal, dual-engine direct drive dual-screen system, panoramic 6K visual effects with support for expansive camera positioning, up to 6K30 H265 and 4K60 HEVC/VP9 video decode.
Allwinner releases their new ARM SoCs for 64bit and 32bit cost effective tablets, featuring their new Allwinner A63 Quad-core ARM Cortex-A53 with Mali-T760MP2 GPU. This video was filmed at the Allwinner press conference event in Shenzhen where they had a lot of their industry partners, design houses, factories, OEMs, software partners, to see and to hear about the latest of what Allwinner has been working to do for them. The new Allwinner A50 is Allwinner's mid-range Quad-core ARM Cortex-A7 with Mali-400 GPU with support for DDR4 and LPDDR4 RAM, better TLC based Flash memory. For the very entry level Tablet market, Allwinner presents their new Allwinner A33G which is limited at a lower 1.35Ghz clock-speed with a lower resolution display support. While these 3 new ARM Processors by Allwinner are presented as their newest Tablet processors, they are also compatible with all sorts of other devices which could be Robots, interactive Smart Speakers with displays, Smart Rearview Mirrors, Smart Mirrors, Laptops, Desktops, and more. Allwinner also talks about their participation in the Open AI Labs which I also filmed here.
Using IBM Power9 and Nvidia Volta GPU technology, the Summit system is a proven AI pioneer expected to deliver in excess of 200 Peta flops of performance, likely making it one of the world’s most powerful supercomputer.
For decades we have been able to take advantage of Moore’s Law to improve single thread performance, reduce power and cost with each generation of semiconductor technology. While technology has advanced after the end of Dennard scaling more than 10 years ago, the advances have slowed down. Server performance increases have relied on increasing core counts and power budgets.
At the same time, workloads have changed in the era of cloud computing. Scale out is becoming more important than scale up. Domain specific architectures have started to emerge to improve the energy efficiency of emerging workloads like deep learning
This talk will provide a historical perspective and discuss emerging trends driving the development of modern servers processors.