Opinionated - China Chipping Away to Semiconductor Dominance

Bussard Ramjet · Apr 28, 2018

qwerrty said:
chinese alternatives to finisar, lumentum, lclaro, neophotonics

these companies are making high performance 100G/200G and 400G optical transceivers for next-gen data centers

accelink
innolight - backed by google
hisense broadband - backed by huawei
eoptolink
o-net - optical subcomponents

Except these are not the only components required. Also, there are also a lot of software systems required where China is usually even weaker.

And then there are components where America has near monopoly. Like FPGA for example.

qwerrty · Apr 28, 2018

Bussard Ramjet said:
Except these are not the only components required. Also, there are also a lot of software systems required where China is usually even weaker.

And then there are components where America has near monopoly. Like FPGA for example.

fpga is no big deal. there are a lot of companies making it or making chips for it like: deephi, horizon robotics, bitmain, cambricon, alibaba, tencent, baidu, nudt; etc..

Bussard Ramjet · Apr 28, 2018

qwerrty said:
fpga is no big deal. there are a lot of companies making it or making chips for it like: deephi, horizon robotics, bitmain, cambricon, alibaba, tencent, baidu, nudt etc..

Umm.. really....

Please tell me where?

Deephi makes its NPUs.
Horizon Robotics is making edge computing ASICs for image and video analysis.
Bitmain, is again making some AI ASICs.
Cambricon again is making AI ASICs.
Alibaba makes nothing as yet, but is developing its own Ali NPUs.
Tencent, same as Alibaba.
Baidu, sources it FPGAs from Xilinx and uses them to make its FPGA solution called XPU.

In fact the companies that actually do make FPGAs in China like Gowin haven't been pointed by you. But all FPGA vendors in China as of now are really small, and have products for very low end applications.

qwerrty · Apr 28, 2018

Bussard Ramjet said:
Umm.. really....

Please tell me where?

Deephi makes its NPUs.
Horizon Robotics is making edge computing ASICs for image and video analysis.
Bitmain, is again making some AI ASICs.
Cambricon again is making AI ASICs.
Alibaba makes nothing as yet, but is developing its own Ali NPUs.
Tencent, same as Alibaba.
Baidu, sources it FPGAs from Xilinx and uses them to make its FPGA solution called XPU.

In fact the companies that actually do make FPGAs in China like Gowin haven't been pointed by you. But all FPGA vendors in China as of now are really small, and have products for very low end applications.

all those ai chips are designed to work like fpga

j20blackdragon · Apr 28, 2018

Bussard Ramjet said:
And then there are components where America has near monopoly. Like FPGA for example.

Historically, FPGAs have been slower, less energy efficient and generally achieved less functionality than their fixed ASIC counterparts.
https://en.wikipedia.org/wiki/Field-programmable_gate_array

As a result, FPGAs are approximately 3 times slower, 20 times larger, and 12 times less power efficient compared to ASICs
https://pdfs.semanticscholar.org/1ceb/8fedc4269403e682ea89205369bdbf74919d.pdf

:lol:

Bussard Ramjet said:
Bitmain, is again making some AI ASICs.

qwerrty · Apr 28, 2018

j20blackdragon said:
Historically, FPGAs have been slower, less energy efficient and generally achieved less functionality than their fixed ASIC counterparts.
https://en.wikipedia.org/wiki/Field-programmable_gate_array

As a result, FPGAs are approximately 3 times slower, 20 times larger, and 12 times less power efficient compared to ASICs
https://pdfs.semanticscholar.org/1ceb/8fedc4269403e682ea89205369bdbf74919d.pdf

seeing how google, fb, amazon, alibaba and tencent are choosing asic over fpga, gpu and cpu for their new data centers.. that shit must have advantages

Bussard Ramjet · Apr 29, 2018

qwerrty said:
all those ai chips are designed to work like fpga

Yes, because those ASICs are meant to be that. FPGAs are flexible, and are not limited to AI usage. They can handle flexible work flows and demands.

j20blackdragon said:
Historically, FPGAs have been slower, less energy efficient and generally achieved less functionality than their fixed ASIC counterparts.
https://en.wikipedia.org/wiki/Field-programmable_gate_array

As a result, FPGAs are approximately 3 times slower, 20 times larger, and 12 times less power efficient compared to ASICs
https://pdfs.semanticscholar.org/1ceb/8fedc4269403e682ea89205369bdbf74919d.pdf

FPGAs, are a different class of products than ASICs. An ASIC is purposely designed to handle a certain type of algorithm or task, and hence gives good performance for the singular task that it is designed to handle.

However FPGAs, are more general purpose, and one can be programmed to one's needs. Hence the very name, field programmable.

Everything has its own uses. Like a CPU is the core of pretty much all computing and is more general purpose. An ASIC may be a thousand times faster than a CPU at a task, but would be incapable of handling another task, while the CPU can.

qwerrty said:
seeing how google, fb, amazon, alibaba and tencent are choosing asic over fpga, gpu and cpu for their new data centers.. that shit must have advantages

All of the above that you mentioned are choosing both ASICs, and FPGAs for their purposes. An ASIC is highly specialized and won't be able to deal with flexibility. They are literally the hardware implementation of an algorithm.

qwerrty said:
seeing how google, fb, amazon, alibaba and tencent are choosing asic over fpga, gpu and cpu for their new data centers.. that shit must have advantages

This is the whole article from where you picked up this slide:

https://anysilicon.com/fpga-vs-asic-choose/

And EVEN THIS ARTICLE, clearly states that both FPGAs and ASICs have their advantages and complement each other.

chengdusudise · Apr 30, 2018

Mr.bussard ramjet is right.actually i think he is A few of few wise indians

qwerrty · May 4, 2018

SmartSens Unveils SmartClarity

PRNewswire: SmartSens launches the 5MP 1/2.7-inch SC5235 BSI sensor. The new sensor is capable of running 5MP (2608H x 1960V) at 25 fps and supports the interline HDR image synthetic algorithm that expands DR up to 100dB. It can be used in security surveillance systems, ip cameras, car digital video recorders, sport cameras and video telephone conference systems.

SmartSens Technology is also launching the NIR enhanced edition-SC5238. It extends the performance advantage of SC5235 based on the optimization of technology to improve QE in 850nm-940nm band. Moreover, SC5238 can run at a speed of 30 fps and supports the image format at 4MP 50 fps for 16:9 video. Both chips are expected to go into mass production in March 2018.

---------------------------------------------------------------------------------

SmartClarity [TM] Series has introduced a new member to the line-up, starlight class 1080P CMOS
SmartSens Technology
5-6 minutes
PR Newswire

SHANGHAI, May 3, 2018 /PRNewswire/ -- Recently, SmartSens Technology introduced a starlight class 1080P CMOS imaging sensor SC2310, making this another product based on the SmartClarity technology after the introduction of the SC5235.

High Sensitivity, High Dynamic Range

SmartClarity™ Series SC2310

SC2310 has an optical format of 1/2.7", and it has an effective array size of 1932Hx1092V. Having benefitted from the advantages of the 3.0um BSI technology, SC2310 can reach a sensitivity of 4800mV/Lux·s, with a maximum SNR of 43dB, and a dynamic range greater than 100dB. Exceptional pixel performance can guarantee the best image quality, satisfying the advanced requirements of the final customers in terms of all color in night vision and high dynamic range.

Near Infrared Sensitivity Enhancement

Also, SC2310 includes near infrared (NIR) sensitivity enhancement, allowing the quantum efficiency (QE) of the wavelength at 850-940nm to almost double. NIR sensitivity improvement can make an IR illuminated scene brighter and clearer; in addition, from a system point of view, fewer LEDs can be used, lowering power consumption and costs.

High Frame Rate

When operating in 1080P, SC2310 can operate up to 60fps. Finally, this sensor can accommodate both MIPI and DVP outputs.

Advanced integrated circuit architecture and BSI process gives this series of product exceptional night vision capability, allowing this series of products to present full color images even under extreme low light conditions with minimum illumination. Relying on these advantages, SC2310 can perfectly accommodate many applications, such as, surveillance monitor systems, IP cameras, car DVs, sport cameras, or teleconferencing systems.

SC2310 uses an advanced CSP package, through extreme quality testing and is suitable to operate from -30ºC-+85ºC, guaranteeing stable operation for extended periods under harsh conditions.

This sensor is targeted to be in mass production by the second quarter of 2018. If more information about this product is needed, please contact SmartSens sales representatives.

About SmartSens Technology

Smartsens Technology is a leading supplier of high-performance CMOS imaging system worldwide. Our global teams are developing and shipping game-changing and market-leading products. Coupled with strong support from strategic partners and ISO-certified supply chain infrastructure, we have delivered innovative, customer-focused, award-winning, and high-quality imaging solutions for security and surveillance, consumer, AR/VR, IoT, automotive, industrial automation, and other mass-market applications. Up to now, chip shipments have exceeded 100 millions units. For more information, please visit the www . smartsenstech . com

-

TaiShang · May 4, 2018

Cambricon reveals China’s first server-based AI chip

CGTN
1068km to Beijing
2018-05-03

Artificial intelligence (AI) chip startup Cambricon Technologies launched a new generation of their core product in Shanghai on Thursday. Dubbed China’s first cloud AI chip, MLU100 can theoretically perform 166.4 trillion times of fixed-point arithmetic per second at maximum, with energy consumption of no more than 110 watts at peak.

The company's business partners Lenovo, Sugon and iFlytek all introduced products that work with the new processor chip at the same time. It will be mainly used in cloud computing, including servers and data centers.

Taking researchers almost two years to develop, MLU100 is able to work under balanced mode or high-performing mode, while the theoretical maximum speed under balanced mode can reach 128 trillion fixed-point arithmetic per second, consuming 80 watts of electricity.

Like other terminal processors in the series offered by Cambricon, the MLU100 is similar to its predecessors in that it supports all kinds of deep learning – neural networks that mimic human learning – and other classic computer learning algorithms. It can also process images, sound and big data at the cloud level under complex scenarios, such as with huge amounts of data and multiple tasks.

Cambricon's CEO Chen Tianshi (L) presents the new board with Chen Guoliang, an academician at the Chinese Academy of Sciences (CAS) at the launch event in Shanghai, May 3, 2018. /VCG Photo

Known for its prominent voice recognition software, iFlytek showed in their test results that Cambricon’s new processor chip helped improve the company’s product, while saving five more times in power consumption than their rival's GPU Cloud. Its strong processing ability also enables smartphones to process more complicated machine learning algorithms, and voice recognition at the local terminal with 9.8 percent more accuracy than those equipped with traditional chips, thus greatly enhancing user experiences.

In 2016, Cambricon released China’s first AI chip, the 1A, claiming it the world’s first commercialized neural network processor chip. It has been used in millions of smart terminals such as Huawei Mate 10 and Honor 10 smartphones.

“There are a large number of deep learning applications today, but all of them are based on traditional generic processors such as Central Processing Unit (CPU) and Graphics Processing Unit (GPU),” said chief executive Chen Tianshi, a former researcher at Chinese Academy of Sciences' Institute of Computing Technology.

“For example, a few years ago Google used more than 10,000 CPUs to train a cat face recognition model, but neither CPUs nor GPUs are suitable for developing neural networks of a size similar to the human brain,” Chen added.

JSCh · May 4, 2018

New chip an intelligent move for Cambricon
By He Wei in Shanghai | China Daily | Updated: 2018-05-04 09:03

Cambricon Technologies Corp Ltd, a successful Chinese startup specializing in artificial intelligence or AI chips, unveiled the MLU 100, a cloud-based AI chip, in Shanghai on Thursday.[Photo/IC]

Cambricon Technologies Corp Ltd, a successful Chinese startup specializing in artificial intelligence or AI chips, unveiled the MLU 100, a cloud-based AI chip, in Shanghai on Thursday.

This is expected to increase the unicorn's lead in a niche segment of the capital-intensive chip sector.

Cambricon's new chip outperforms traditional general purpose processors used in image and speech recognition technologies.

The MLU 100 is designed to make AI-powered applications more concentrated and scalable, said Chen Tianshi, Cambricon's founder and CEO.

It consumes less power, is compatible with various deep learning and classic machine learning algorithms, and meets the needs of technologies like visual, speech, and natural language processing.

Cambricon also unveiled 1M, its latest generation of terminal processors that can effectively speed up a variety of AI functions embedded in smartphones, speakers, security cameras and cars.

The 1M has a peak capacity of 5 trillion synapses per second and boasts higher integration density, making it ideal for many devices.

"Our task is to make intelligent processors more accessible to people and enable the whole world to adopt such chips in a variety of scenarios," Chen said.

At the joint launch ceremony, Lenovo Group, Sugon Information Industry Co Ltd and iFlytek Co Ltd launched systems or products carrying Cambricon's latest products.

"We would like to combine Cambricon's homegrown chips with Lenovo's platforms and channels to aid Chinese companies with their digital transformation," said Tong Fuyao, Lenovo's senior vice-president.

A neural processing unit or NPU handles AI workload, particularly excelling at image recognition tasks with an efficiency that greatly outstrips traditional processors such as the central processing unit or CPU, or graphics processing unit or GPU.

Huawei Technologies Co Ltd's Kirin 970 chip, which was developed to power its latest flagship smartphone Mate 10, has used Cambricon's products.

Cambricon was founded in 2016 by Chen Tianshi and his brother Chen Yunji. It is backed by the Chinese Academy of Sciences and several tech firms, and venture capitals.

AI applications will drive demand for chips capable of processing AI algorithms, according to a recent report by consultancy Gartner Inc.

AI and the cloud-based internet of things are two major areas where China's homegrown chips have a good chance of competing with global players, industry experts said.

"In light of the ongoing intelligence wave, companies that own enough data and run crucial AI-backed applications would have a competitive edge in producing smart chips," said Zhang Jianfeng, chief technology officer of Alibaba Group Holding Ltd, which is among the investors in Cambricon.

qwerrty · May 4, 2018

digitimes
Goodix enters supply chain for Samsung phones

Jessie Shen, DIGITIMES, Taipei
1-2 minutes
Goodix enters supply chain for Samsung phones

Thursday 3 May 2018

Goodix Technology has announced its entry into the supply chain for Samsung's smartphones.

China-based Android phone makers including Vivo, Huawei, ZTE, Lenovo, Oppo and Meize have been Goodix' major clients. Orders from Samsung are regarded as a milestone in Goodix' market presence expansion overseas.

Goodix disclosed its single-chip fingerprint sensor GW32J1 has been commercialized on the new Galaxy J7 Duo to deliver biometric solutions for consumers in India.

Goodix added its fingerprint sensors have already been adopted by vendors including Amazon, LG, Dell, HP, Nokia and Asustek for their terminal devices. Goodix has also tapped into emerging market segments such as IoT and automotive.

Synaptics and Egis Technology (Egistec) are reportedly also fingerprint sensor suppliers for Samsung's Galaxy smartphones.

-----------------------------------------------------------------------------------------------------------------

biometricupdate

Goodix acquires cellular IP company CommSolid
Feb 26, 2018 | Chris Burt
2-3 minutes

Goodix Technology has acquired cellular IoT company CommSolid to accelerate the development of its System-on-Chips (SoCs) for secure, low power wireless solutions. The acquisition was announced at Mobile World Congress (MWC) 2018, and will allow Goodix to target new applications in smart homes, transportation, logistic systems, and industrial segments.

CommSolid’s baseband intellectual property runs on integrated circuits (ICs) to enable smart applications, according to the announcement.

“We will continue to put emphasis on research and development, broadening our product portfolio and bringing diversified innovative solutions to our customers around the world. The CommSolid acquisition will enhance our offering for the IoT market and bring additional capabilities to the benefit of our customers.” said David Zhang, the Goodix CEO.

The companies note that the number of NarrowBand IoT (NB-IoT) nodes could reach 450 million by 2021, according to IHS Markit, and Technavio forecasts the NB-IoT market will grow at a 61 percent CAGR from 2017 to 2021.

“Goodix is very well positioned to gain significant share in the emerging IoT market which provides an exciting opportunity for us. Joining forces with Goodix will enable the creation of highly optimized connected sensor solutions that the market demands. The two companies also share the common vision to bring customer value through constant innovation.” said Matthias Weiss, Managing Director of CommSolid.

Goodix is also demonstrating its second-generation optical under-display fingerprint technology at MWC 2018, and says it will be commercialized on a large scale by top-tier global smartphone brands.

cirr · May 4, 2018

【存储器】国产化SSD真的来了

来源：嘉合劲威 2018-04-28

嘉合劲威集团携手国科微电子推出第一款国产化最高的固态硬盘,光威“弈”系列SSD终于如期而至。光威“弈”系列SSD，从主控芯片、产品品牌到生产制造均由中国企业完成。从产品命名到风格定位再到包装设计也均为中国元素。存储产品高度国产化，它的意义和影响都非常重大。

国产主控

国科微是国家规划布局内的重点集成电路设计企业和首家获得国家集成电路产业投资基金注资的集成电路设计企业，也是中南地区规模最大的集成电路设计企业之一。在2017年下半年国科微发布了拥有自主知识产权的第二代存储产品主控GK2301，迅速进入了产品应用阶段。GK2301主控是国内首款获得国测，国密双重认证的主控。