What's new

Opinionated - China Chipping Away to Semiconductor Dominance

NVIDIA Amps Up AI Cloud Strategy with ODM Partnerships
Michael Feldman | June 2, 2017 01:48 CEST

NVIDIA is hooking up with four of the world’s largest original design manufacturers (ODMs) to help accelerate adoption of its GPUs into hyperscale datacenters. The new partner program would give Foxconn, Inventec, Quanta and Wistron early access to the HGX reference architecture, NVIDIA’s server design for machine learning acceleration.

3X4A6519.gif

Jensen Huang, Founder & CEO of NVIDIA, at GTC 2017

HGX is an attempt to establish an industry-standard GPU box that maximizes computational density for machine learning workloads. It uses NVIDIA’s most advanced GPUs, namely the Tesla P100, and soon, the Tesla V100. It glues eight of these into an NVLink cube mesh, and uses PCIe switching to allow CPUs to dynamically connect to them. Examples of this architecture include Microsoft's Project Olympus HGX-1 chassis, Facebook's Big Basin system, and NVIDIA’s own DGX-1 server.

Facebook’s Big Basin and Microsoft’s HGX-1 systems are GPU-only boxes, which rely on external CPU servers as hosts. Since the processor and co-processor are disaggregated, applications can fiddle with GPU-CPU ratio as needed. In most machine learning situations, you want a rather high ratio of GPUs to CPUs, since most of the processing ends up on the graphics chip. And in hyperscale/cloud datacenters, you also want the flexibility of allocating these resources dynamically as workloads shift around.

The DGX-1 server is a different animal altogether. It’s a stand-alone machine learning appliance and includes two Xeon processors, along with the same eight-GPU NVLink mesh of its hyperscale cousins. As such, it’s not meant for cloud duty, but rather for businesses, research organizations, and software development firms that want an in-house machine learning box. SAP in the most prominent commercial buyer of the DGX-1, at least of those revealed publicly. But NVIDIA never intended to sell boatloads of these systems, especially since a lot of customers would prefer to rent machine learning cycles from cloud providers.

Untitled.png


That’s why the ODM partnership could end up paying big dividends. These manufacturers already have the inside track with hyperscale customers, who have figured out that they can use these companies to get exactly the gear they want, and at sub-OEM pricing. ODMs are also more nimble than traditional server-makers, inasmuch that they can shorten the design-to-production timeline. That makes them better suited to the nearly continuous upgrade cycle of these mega-datacenters.

Given that the HGX-1 is manufactured by Foxconn subsidiary Ingrasys and the Big Basin system is built by Quanta, it’s a logical step for NVIDIA to include the other big ODMs, Inventec and Wistron, into the fold. The goal is to bring a wider range of HGX-type machinery to market and make them available to hyperscale customers other than just Microsoft and Facebook.

The other aspect of this is that NVIDIA would like to solidify its dominance with machine learning customers before Intel brings its AI-optimized silicon to market. Startup companies like Wave Computing and Graphcore also are threatening to challenge NVIDIA with their own custom chips. Establishing an industry-standard architecture before these competing solutions get market traction would help NVIDIA maintain its leadership.

hgx-graphic-790x546.png

Source: NVIDIA

To some extent, NVIDIA is also competing with some its biggest customers, like Google and Microsoft, both of which are building AI clouds based on their own technologies. In the case of Google, it’s their Tensor Processor Unit (TPU), which the search giant has upgraded for an expanded role that threatens NVIDIA directly. Meanwhile, Microsoft is filling out its AI infrastructure with an FPGA-based solution that, likewise, could sideline NVIDIA GPUs in Azure datacenters.

The prospect of using the future V100 Tesla GPUs in HPX platforms actually intensifies the competition, since these upcoming processors are built for both neural net training and inferencing. Although NVIDIA used to build its own inferencing-specific GPUs (the M4 and M40, followed by the P4 and P40), inferencing is also performed by regular CPUs and FPGAs, not to mention Google’s TPUs, running in regular cloud servers.

Inferencing has somewhat different requirements than training, especially with regards to minimizing latency, but with the Volta architecture and the V100, NVIDIA thinks it has designed a solution that is capable of doing both, and doing so competitively. From a hyperscale company’s point of view, there are some obvious advantages in separating inferencing, and certainly training infrastructure from the rest of the server farm – not the least of which is being able to deploy and run machine learning gear in a more flexible manner. And since these upcoming V100 GPUs will be used by hyperscale companies for training, they are also likely to get a shot at some of those same companies’ inference workloads.

Finally, if NVIDIA manages to establish HGX as the standard GPU architecture for AI clouds, it makes its own recently announced GPU cloud platform more attractive. Since NVIDIA’s cloud stack of machine learning libraries and frameworks runs on top of other peoples’ infrastructure, pushing its HGX architecture into the ecosystem would make NVIDIA’s job of supporting the various hardware solutions that much simpler. It would also make it easier for customers to switch cloud providers without having to tweak their own software.

We’ll be able to tell if these ODM relationships pay off when we start seeing additional HGX solutions coming to market and being adopted by various cloud providers. As NVIDIA likes to remind us, its GPUs are used in the world’s top 10 hyperscale businesses today. If all goes as planned, someday it will be able to make the same claim for HGX.


https://www.top500.org/news/nvidia-amps-up-ai-cloud-strategy-with-odm-partnerships/
@TaiShang
 
Japan's Sharp Corp (6753.T) on Friday forecast its first net profit in four years following cost-cutting under new Taiwanese owner Hon Hai Precision Industry Co Ltd (Foxconn) (2317.TW), and said it would resume making active investments.

Looks like they are doing better under a new (and better) management.

At least we do not wish to colonize Japan by force as they did for 50 years on us. :enjoy:

由中投公司、丝路基金、新加坡淡马锡、深圳深业集团、厚朴投资与ARM公司共同发起设立的厚安创新基金在北京正式成立启动,该基金管理人首次亮相。厚安创新基金由半导体知识产权提供商ARM公司及厚朴投资负责管理。基金将结合ARM的全球产业生态系统,专注于投资移动互联、物联网、人工智能等多个关键领域具有潜力的技术公司。

That's a great cooperation which even involves the Silk Road Fund. I love it more when China drops a little bit that toxic "globalization" discourse and dedicates some energy on the region.
 
Looks like they are doing better under a new (and better) management.
Yes finally Sharp is turned around, amazing feat by Foxconn! Really impressed with Taiwan tech companies, not just Foxconn but also Inventec, Quanta, Wistron, Mediateck, FFG and a long list of powerhouses, too strong! Jensen Huang will partner with ODM in next gen Nividia hyerscale datacenter, all four ODM are Taiwanese!
 
国产电子级多晶硅打破国外垄断 年产能2500吨 :enjoy::enjoy:

2017-06-07 10:01:26 来源:中国能源报

多晶硅是微电子行业和光伏产业的“基石”,是信息产业和新能源产业最基础的原材料。一般电子级硅要求含Si>99.9999%,电子级多晶硅则要求达到9个“9”以上。作为战略性原材料,电子级多晶硅的生产技术和市场长期由国外垄断,制约我国集成电路产业、新能源产业、新材料产业的发展。

据《中国能源报》报道,5月24日,国家电投黄河水电新能源分公司正式推出国产高纯电子级多晶硅产品——“黄河水电多晶硅”,实现了电子级多晶硅的“中国制造”,打破国外垄断格局。

20170607080533425.jpg

还原炉硅棒温度检测

缺口巨大

全球多晶硅产能发展始于1980年,德、美、日采用改良西门子法,使得产能达到1000吨—2000吨。到了2000年,韩国加入,德、美、日、韩采用改良西门子法和硅烷法,生产出太阳能级产品,产能突破10000吨。

变化出现在2012年,全球多晶硅产能达到25万吨、产量23.5万吨,我国产量为7.1万吨。到了2017年,全球产能达40万吨、产量38万吨,我国产量则达到21万吨。

我国电子级多晶硅的研制始于二十世纪五十年代,六十年代中期进入批量生产,至七十年代末改良西门子法的生产技术、工艺已臻成熟。那时尽管生产规模小,设备简陋,但因当时国内需求少,生产的产品基本能够满足国内需求。此后由于国内产量不足、质量不稳定等原因,我国开始大量进口电子级多晶硅。

从进口数据分析,2013年至2017年,我国多晶硅进口数量从80653吨逐年攀升至144000吨。2014年,国务院签发了《国家集成电路产业发展推进纲要》,同时成立了国家集成电路产业投资基金,电子级多晶硅迎来重大利好。

从近十年的发展状况来看,虽然多晶硅在我国得到长足发展,但高品质多晶硅仍然需要进口。进口数据中所有的分立器件、集成电路用多晶硅和高效光伏电池用多晶硅全部进口,国内高品质多晶硅缺口仍然巨大。

有专家研判,国内电子级多晶硅的需求是4500吨/年。

“全球电子级硅材料规模迅速扩张,技术要求日趋严格。得益于良好的发展环境,国内多晶硅产业和产品品质快速进步,技术要求向国际水平靠近势在必行,行业内将不遗余力推进国产多晶硅的应用。”有研半导体博士肖清华告诉记者。

填补国内空白

国内首条电子级多晶硅生产线的出现,经历了近10年的时间。

据黄河公司新能源分公司总经理刘刚介绍,“黄河水电多晶硅”的研发工作正式开始于2006年,于2010年全面投产,形成的2500吨高纯电子级多晶硅产能,填补了国内空白,目前国内市场占有率达10%。2015年,黄河水电多晶硅完成国家科技重大专项《极大规模集成电路制造装备及成套工艺》(即“02专项”)全部技术指标要求。

20170607080555427.jpg

多晶硅厂全貌

“黄河水电多晶硅”采用改良西门子法,年产达2500吨电子级多晶硅,实现完全闭路循环,节能低耗、安全环保,其氧含量<0.02ppma,碳含量<0.02ppma,受主杂质<0.03ppba,施主杂质<0.05ppba,基体总金属杂质<2ppbw,经第三方权威检测机构检测,产品质量与德国、日本知名多晶硅质量相当。

“攻克电子级多晶硅的关键要素包括很多方面,对工艺、设备管道选型、仪表选型、洁净室等级都有严格的要求,在施工及检修时,需要按照最高洁净等级规范,在原辅材料、中间产品、辅助气体和介质方面也要进行完善的质量体系控制。”刘刚补充,“每一个细小环节都需要长期的实验积累有效的数据,从中找出制约质量提升的因素。”

来自重庆的某下游企业负责人告诉记者:“设备可以阶段性引入,但材料必须是连续性的,我们曾遇到过某外国厂商的突然断供,造成了很大损失。如果关键材料可以实现国产化,我们可以减少很多后顾之忧。”

步入新阶段

电子级多晶硅国产化,率先在青海取得突破,一部分源于地缘优势。

“十二五”以来,青海省形成了“以硅为主、多元发展、集中布局”的产业格局。青海省副省长王黎明强调:“青海将进一步加快技术创新,努力打造国内重要的集成电路硅材料产业基地,走出一条既符合国家战略、又切合青海实际的绿色发展之路。在保护生态环境的前提下,坚定不移打造千亿元新材料产业链。”

据了解,青海已建成晶硅产能17500吨,2016年该省共生产单晶硅2635吨,多晶硅14228吨。

随着《国家集成电路产业发展推进纲要》、《中国制造2025》、《工业强基工程实施指南(2016-2020年)》等一系列政策的出台,我国电子级多晶硅的发展步入新阶段。在政策环境利好下,黄河公司总经理魏显贵表示,将致力于核心技术研发和质量提升,努力将国内电子级多晶硅市场占有率提升至30%以上。

“黄河水电多晶硅” 打响了民族品牌,在价格上也几乎与进口产品持平。“多晶硅国产化才刚刚起步,我们将进一步提高智能化制造水平和生产控制能力,确保产品稳定性,力拼工艺研发,发展才有持续性。”刘刚说。

电子级多晶硅国产化要协同发力(短评)

电子级多晶硅行业上下游之间的发展关系可以用“皮之不存,毛将焉附”形容。

我国将集成电路作为基础性和先导性产业,附加在全行业中的关键性作用愈加凸显,电子级多晶硅作为国家发展集成电路产业的战略原材料,已成为我国发展集成电路产业的重要支撑。我国半导体行业市场目前位居全球第一、增速也位居全球第一。但是,我国集成电路产业与先进产业国之间的差距较大,大尺寸晶圆的生产研发才刚刚起步,集成电路核心技术研发和关键元器件制造还面临巨大的挑战。

局部的突破不代表整个电子级多晶硅全产业链上的突破。以点带面,求得全面突破才是众望所归。着力培育龙头企业,力求产业链协同创新,加快推广应用,完善产业环境,调动用户企业积极性等显得尤为迫切。

电子级多晶硅实现“中国造”,为行业发展树立了新的里程碑,而为了早日实现其全产业链“中国造”,则需要上下游企业协同发力,在技术、融资、人才、项目合作等多领域,推进全产业链的进步。

http://www.guancha.cn/industry-science/2017_06_07_412022.shtml

@Bussard Ramjet :lol::lol:
 
国产电子级多晶硅打破国外垄断 年产能2500吨 :enjoy::enjoy:

2017-06-07 10:01:26 来源:中国能源报

多晶硅是微电子行业和光伏产业的“基石”,是信息产业和新能源产业最基础的原材料。一般电子级硅要求含Si>99.9999%,电子级多晶硅则要求达到9个“9”以上。作为战略性原材料,电子级多晶硅的生产技术和市场长期由国外垄断,制约我国集成电路产业、新能源产业、新材料产业的发展。

据《中国能源报》报道,5月24日,国家电投黄河水电新能源分公司正式推出国产高纯电子级多晶硅产品——“黄河水电多晶硅”,实现了电子级多晶硅的“中国制造”,打破国外垄断格局。

20170607080533425.jpg

还原炉硅棒温度检测

缺口巨大

全球多晶硅产能发展始于1980年,德、美、日采用改良西门子法,使得产能达到1000吨—2000吨。到了2000年,韩国加入,德、美、日、韩采用改良西门子法和硅烷法,生产出太阳能级产品,产能突破10000吨。

变化出现在2012年,全球多晶硅产能达到25万吨、产量23.5万吨,我国产量为7.1万吨。到了2017年,全球产能达40万吨、产量38万吨,我国产量则达到21万吨。

我国电子级多晶硅的研制始于二十世纪五十年代,六十年代中期进入批量生产,至七十年代末改良西门子法的生产技术、工艺已臻成熟。那时尽管生产规模小,设备简陋,但因当时国内需求少,生产的产品基本能够满足国内需求。此后由于国内产量不足、质量不稳定等原因,我国开始大量进口电子级多晶硅。

从进口数据分析,2013年至2017年,我国多晶硅进口数量从80653吨逐年攀升至144000吨。2014年,国务院签发了《国家集成电路产业发展推进纲要》,同时成立了国家集成电路产业投资基金,电子级多晶硅迎来重大利好。

从近十年的发展状况来看,虽然多晶硅在我国得到长足发展,但高品质多晶硅仍然需要进口。进口数据中所有的分立器件、集成电路用多晶硅和高效光伏电池用多晶硅全部进口,国内高品质多晶硅缺口仍然巨大。

有专家研判,国内电子级多晶硅的需求是4500吨/年。

“全球电子级硅材料规模迅速扩张,技术要求日趋严格。得益于良好的发展环境,国内多晶硅产业和产品品质快速进步,技术要求向国际水平靠近势在必行,行业内将不遗余力推进国产多晶硅的应用。”有研半导体博士肖清华告诉记者。

填补国内空白

国内首条电子级多晶硅生产线的出现,经历了近10年的时间。

据黄河公司新能源分公司总经理刘刚介绍,“黄河水电多晶硅”的研发工作正式开始于2006年,于2010年全面投产,形成的2500吨高纯电子级多晶硅产能,填补了国内空白,目前国内市场占有率达10%。2015年,黄河水电多晶硅完成国家科技重大专项《极大规模集成电路制造装备及成套工艺》(即“02专项”)全部技术指标要求。

20170607080555427.jpg

多晶硅厂全貌

“黄河水电多晶硅”采用改良西门子法,年产达2500吨电子级多晶硅,实现完全闭路循环,节能低耗、安全环保,其氧含量<0.02ppma,碳含量<0.02ppma,受主杂质<0.03ppba,施主杂质<0.05ppba,基体总金属杂质<2ppbw,经第三方权威检测机构检测,产品质量与德国、日本知名多晶硅质量相当。

“攻克电子级多晶硅的关键要素包括很多方面,对工艺、设备管道选型、仪表选型、洁净室等级都有严格的要求,在施工及检修时,需要按照最高洁净等级规范,在原辅材料、中间产品、辅助气体和介质方面也要进行完善的质量体系控制。”刘刚补充,“每一个细小环节都需要长期的实验积累有效的数据,从中找出制约质量提升的因素。”

来自重庆的某下游企业负责人告诉记者:“设备可以阶段性引入,但材料必须是连续性的,我们曾遇到过某外国厂商的突然断供,造成了很大损失。如果关键材料可以实现国产化,我们可以减少很多后顾之忧。”

步入新阶段

电子级多晶硅国产化,率先在青海取得突破,一部分源于地缘优势。

“十二五”以来,青海省形成了“以硅为主、多元发展、集中布局”的产业格局。青海省副省长王黎明强调:“青海将进一步加快技术创新,努力打造国内重要的集成电路硅材料产业基地,走出一条既符合国家战略、又切合青海实际的绿色发展之路。在保护生态环境的前提下,坚定不移打造千亿元新材料产业链。”

据了解,青海已建成晶硅产能17500吨,2016年该省共生产单晶硅2635吨,多晶硅14228吨。

随着《国家集成电路产业发展推进纲要》、《中国制造2025》、《工业强基工程实施指南(2016-2020年)》等一系列政策的出台,我国电子级多晶硅的发展步入新阶段。在政策环境利好下,黄河公司总经理魏显贵表示,将致力于核心技术研发和质量提升,努力将国内电子级多晶硅市场占有率提升至30%以上。

“黄河水电多晶硅” 打响了民族品牌,在价格上也几乎与进口产品持平。“多晶硅国产化才刚刚起步,我们将进一步提高智能化制造水平和生产控制能力,确保产品稳定性,力拼工艺研发,发展才有持续性。”刘刚说。

电子级多晶硅国产化要协同发力(短评)

电子级多晶硅行业上下游之间的发展关系可以用“皮之不存,毛将焉附”形容。

我国将集成电路作为基础性和先导性产业,附加在全行业中的关键性作用愈加凸显,电子级多晶硅作为国家发展集成电路产业的战略原材料,已成为我国发展集成电路产业的重要支撑。我国半导体行业市场目前位居全球第一、增速也位居全球第一。但是,我国集成电路产业与先进产业国之间的差距较大,大尺寸晶圆的生产研发才刚刚起步,集成电路核心技术研发和关键元器件制造还面临巨大的挑战。

局部的突破不代表整个电子级多晶硅全产业链上的突破。以点带面,求得全面突破才是众望所归。着力培育龙头企业,力求产业链协同创新,加快推广应用,完善产业环境,调动用户企业积极性等显得尤为迫切。

电子级多晶硅实现“中国造”,为行业发展树立了新的里程碑,而为了早日实现其全产业链“中国造”,则需要上下游企业协同发力,在技术、融资、人才、项目合作等多领域,推进全产业链的进步。

http://www.guancha.cn/industry-science/2017_06_07_412022.shtml

@Bussard Ramjet :lol::lol:

Great improvement for renewable energy industry of China as well as microelectronics. China dominates wind and solar energy sectors but the share of domestic content is not as high as desirable (which is 100%), so, this is another great step, perhaps more important than the ball-point pen tip feat.
 
Summary of video:
The machine is a 5 nm chip manufacturing equipment (etcher) developed by AMEC and has precision of 0.5 nm.
Most advanced chips on the market is 14 nm with 10 nm chips going into production this year and 7nm chips will be in production soon. While the most advanced chips China can currently produce are 40 nm and 28 nm, creating a 3 generation gap with the most advanced chips in the world. In order for China to compete at the forefront in this competitive market, it must jump generations to leap ahead of competitors. The equipment took 5 years to develop and the company expects the engineers to finalise the machine into a marketable product by the end of 2017. The development of chip manufacturing equipment precedes chip design by 5 years. As one of the engineers said "I estimate that we will only see 5nm chips on the market 5 years from now". Hundreds of engineers from the "core team" had experience from US (Intel, Lam research, etc.) and other first tier semiconductor manufacturing and equipment companies, with most having 20-30 years of experience in designing semiconductor manufacturing equipment.

Video mentions two machines, first one being an etcher and the second one being MOCVD machine. World demand for MOCVD machines is 100-150 annually, the company will supply about 60-70 sets in 2017. The current model (2nd gen) was developed in June 2016. Similar machines on the market typically use 400mm+ 晶圆托盘=wafer tray? while AMEC uses 700mm+ wafer tray, effectively doubling chip production under the same investment and time constraints compared to competitors. R&D took 6-7 years spanning 2 generations with total investment in the several tens of millions of yuan. This is a remarkable breakthrough for china as before the firm's entry, the technology for this key equipment was controlled by the US and German companies.

If successful this will give China the possibility of marketing world class chips within 5 years.

Company website
http://www.amec-inc.com/

Etching
https://en.wikipedia.org/wiki/Etching_(microfabrication)

MOCVD
https://en.wikipedia.org/wiki/Metalorganic_vapour_phase_epitaxy
 
Last edited:
Summary of video:
The machine is a 5 nm chip manufacturing equipment (etcher) developed by AMEC and has precision of 0.5 nm.
Most advanced chips on the market is 14 nm with 10 nm chips going into production this year and 7nm chips will be in production soon. While the most advanced chips China can currently produce are 40 nm and 28 nm, creating a 3 generation gap with the most advanced chips in the world. In order for China to compete at the forefront in this competitive market, it must jump generations to leap ahead of competitors. The equipment took 5 years to develop and the company expects the engineers to finalise the machine into a marketable product by the end of 2017. The development of chip manufacturing equipment precedes chip design by 5 years. As one of the engineers said "I estimate that we will only see 5nm chips on the market 5 years from now". Hundreds of engineers from the "core team" had experience from US (Intel, Lam research, etc.) and other first tier semiconductor manufacturing and equipment companies, with most having 20-30 years of experience in designing semiconductor manufacturing equipment.

Video mentions two machines, first one being an etcher and the second one being MOCVD machine. World demand for MOCVD machines is 100-150 annually, the company will supply about 60-70 sets in 2017. The current model (2nd gen) was developed in June 2016. Similar machines on the market typically use 400mm+ 晶圆托盘=wafer tray? while AMEC uses 700mm+ wafer tray, effectively doubling chip production under the same investment and time constraints compared to competitors. R&D took 6-7 years spanning 2 generations with total investment in the several tens of millions of yuan. This is a remarkable breakthrough for china as before the firm's entry, the technology for this key equipment was controlled by the US and German companies.

If successful this will give China the possibility of marketing world class chips within 5 years.

Company website product page
http://www.amec-inc.com/

Etching
https://en.wikipedia.org/wiki/Etching_(microfabrication)

MOCVD
https://en.wikipedia.org/wiki/Metalorganic_vapour_phase_epitaxy
Bro, I already posted this before, but you explained it better than me. Thanks.
 
this is much cheaper than buying foreign chip companies lol

Korean Broker is Behind the Growth of China’s Semiconductor Industry

SEOUL,KOREA
7 June 2017 - 12:15pm
Cho Jin-young
It has been found that a South Korean broker is assisting in the growth of China’s semiconductor industry by supplying South Korean engineers to Chinese semiconductor companies.

This person, who is a former vice president of both Samsung Electronics and SK Hynix, set up a semiconductor company in Taiwan two years ago. Although this company has the form of a consulting firm, industry insiders say that it is a manpower outsourcing firm supplying the engineers at the request of the Chinese companies. “The company in Taiwan and the Chinese companies have conducted joint projects for at least 18 months,” one of them said, adding, “The latter’s targets include experts experienced in semiconductor manufacturing process management.”

The anonymous person is an expert in the fields of solar power generation and semiconductor manufacturing. He or she joined Samsung Electronics in the mid-1980s and worked as a senior researcher and a senior managing director for 17 years until the early 2000s. Then, this person moved to SK Hynix, worked for about 10 years, and retired as a vice president before serving as a president of STX Solar and Hanwha. The person flew to Taiwan in the latter half of 2015. Regarding this matter, the National Intelligence Service of South Korea is currently looking into the semiconductor company in Taiwan.

These days, Chinese semiconductor companies are trying to employ more and more South Korean engineers in the industry via local headhunters. To this end, those companies are offering an annual salary three to 10 times that in South Korea, free houses and cars, education for their children, etc. The Chinese government is planning to invest up to one trillion yuan until 2025 so that the ratio of semiconductor chips produced by Chinese companies reaches at least 70%.
Code:
http://www.businesskorea.co.kr/english/news/ict/18305-chip-technology-leakage-korean-broker-behind-growth-china%E2%80%99s-semiconductor-industry
 
this is much cheaper than buying foreign chip companies lol
It has been found that a South Korean broker is assisting in the growth of China’s semiconductor industry by supplying South Korean engineers to Chinese semiconductor companies.
I thought Chinese companies are headhunting from Taiwan, now SK engineers too? Good news! Buying companies is not just more expensive but takes time to integrate, hiring talents is far better.

到了2017年,全球产能达40万吨、产量38万吨,我国产量则达到21万吨。
其氧含量<0.02ppma,碳含量<0.02ppma,受主杂质<0.03ppba,施主杂质<0.05ppba,基体总金属杂质<2ppbw,经第三方权威检测机构检测,产品质量与德国、日本知名多晶硅质量相当。
Briefly translate as "World polycrystalline silicon production total 38 tonnes, in which China's share is 21 tonnes, quality already on par with Germany and Japan", good news!

https://en.wikipedia.org/wiki/Polysilicon
 
Next: Homegrown semiconductors of the 5G tech variety

2017-06-12 11:31

China Daily Editor: Feng Shuang

Spreadtrum RDA, the mobile-design subsidiary of China's largest chipmaker Tsinghua Unigroup Ltd, aims to unveil 5G smartphone chips in 2018.

The move is part of its plan to gain an early lead in the next-generation mobile communication technology, and follows the agreement with Britain's Dialog Semiconductor Plc in March to develop smartphone chips and mobile power management technology.

Dialog's chips already power Apple Inc's iPhones and iPads.

Kang Yi, global advance technologies & projects vice president at Spreadtrum, said the company has started research and development of 5G chips and is playing an active part in China's 5G trials.

"We teamed up with Huawei in the first phase of trials to verify key 5G technologies last year. Now we are a part of China's second phase of experiments, which will test key technological solutions," Kang said.

According to him, the company plans to launch standards-based mobile 5G chips as soon as a global 5G standard is rolled out.

The 3rd Generation Partnership Project, or 3GPP, is an industry coalition dedicated to promoting global wireless communication standards. It decided to accelerate some elements in the 5G New Radio, or NR, specification. NR is essential to accelerate the commercial applications of the technology.

According to the expedited schedule, 5G NR specifications are to be completed in December,which is expected to give Spreadtrum RDA's 5G launch plan a shot in the arm.

Spreadtrum RDA came into existence after Tsinghua Unigroup bought Spreadtrum Communications and RDA Microelectronics Inc three years ago and merged them into a new company.

It is locked in fierce competition with MediaTek Inc in the market for lower end smartphone chips, and hopes to join the ranks of global giants such as Qualcomm Inc in the premium segment.

"The design of 5G chips is definitely more sophisticated, for the new technology demands faster speed and quicker response, which complicates the task of balancing power and performance, and integrating resources and lower costs," Kang said.

The company's R&D of 5G chips starts from 12 nanometer technology and plans to leverage a more advanced 7 nanometer technology for mass production. Nanometer technology refers to a technical standard used in chip device fabrication.

Spreadtrum RDA is part of Tsinghua Unigroup's broad efforts to develop homegrown chips so as to reduce China's reliance on foreign semiconductors.

Unigroup signed agreements in March that would grant it up to 150 billion yuan ($21.8 billion) in financing, which would enable it to expand its chip business.

It is working on a $30 billion domestic chip production complex in Nanjing, Jiangsu province. When completed, it would be China's largest such facility.

In 2016, Spreadtrum RDA shipped 600 million smartphone chips, accounting for one-fourth of the world's total shipments.

http://www.ecns.cn/business/2017/06-12/261151.shtml
 
Fishing for chips
By Ma Si | China Daily | 2017-06-12 07:05

Chinese smartphone makers realize processor supplies are key to achieving global supremacy

In the $429 billion global smartphone market, dominance in the domestic and Indian segments may encourage Chinese handset makers to dream of ejecting Apple and Samsung from the pinnacle, but the reality is far removed.

Chinese vendors still have several loose ends to tie up, none starker than their poor control over key sections of the supply chain, particularly chips, or processors, the engines that drive the handset hardware -the brain, if you will.

This painful truth came to torment Huawei Technologies Co Ltd, the current leader of the Chinese smartphone pack, recently.

All hell broke loose when Richard Yu, CEO of its consumer business unit, admitted in his Weibo post last month that shortage of top-end chips forced Huawei to use a mix of relatively less efficient flash memory cards and high-performance universal flash storage or UFS cards in its flagship P10 smartphone.

The UFS chip is faster than a typical flash memory card, and is essential for high-resolution games and movies.

Yu stressed the P10 hardware cocktail was not meant to save money but ensure a steady supply of shipments and timely delivery of orders.

But irate consumers bombarded Huawei with complaints.

Two red flags went up immediately.

One, anything that undermines a chip is a strict no-no.

Two, Huawei's marketing campaign for the P10 high-lighted a chip-related specification as one of the stand-out features, which stood negated by the company's late admission of truth.

"The (Huawei P10) incident is the latest example of Chinese smartphone vendors' overreliance on foreign chips. This factor has subjected their delivery schedule to the influence of supply chain partners," said Xiang Ligang, a telecom expert and CEO of cctime.com, a telecom industry website.

Agreed Jia Mo, an analyst at Canalysys. "The Huawei incident mirrors a broad problem in China's burgeoning electronics sector. Though China is a manufacturing powerhouse in computers, smartphones and other electronic gadgets, the country banks on foreign players for most processors."

Stated differently, Chinese smartphone manufacturers are years away from self-reliance in, and mastery of, chip-making,which is currently the preserve of global majors such as US multinational Qualcomm, and South Korea's Samsung and SK Hynix.

Qualcomm Inc is the undisputed leader of the global mobile processor market -this is borne out by the fact that all the seven Chinese brands among the world's top 10 smartphone labels depend on it for chips, especially in the premium segment.

It's a truth that does not sit well with the fact that China's smartphone giants have posted exponential growth in both home and overseas markets in 2016.

Domestic shipments for the country's top three vendors Huawei, Oppo and Vivo touched 224.2 million units, up more than 80 percent year-on-year, while their overseas shipments reached 91.8 million units, up about 70 percent year-on-year. Together, they accounted for 21.6 percent of the global smartphone shipments in 2016, according to data compiled from IDC.

But that may count for nothing eventually without reliable supplies of top-end chips, the key to successful forays into, and dominance of, the premium smartphone segment.

Dominance in the premium smartphone segment has strategic significance because it not only indicates technological prowess but generates massive profits.

For instance, in the first quarter of this year, Apple's iconic top-end iPhones, which are powered by in house chips, earned $10.1 billion, or 83.4 percent of profits of the global smartphone segment, according to Strategy Analytics, a market research firm.

So, having the ability to make top-end chips will likely also help Chinese companies to better manage other supply chain partners, experts and company executives said.

In this context, Samsung appears to have an edge as it makes, besides handsets, its own chips, which it also supplies to competitors, including those from China.

That's not all. Samsung and SK Hynix also control nearly half the global market for NAND flash memory, a key component in the UFS chip. Both of them warned last month chip supplies will remain tight this year due to production bottlenecks and strong demand from high-end smartphone makers.

Part of that demand can be traced to China.

According to the General Administration of Customs, in 2016, the country imported chips worth $227 billion, nearly double the $116.5 billion it spent on crude oil imports.

In comparison, the homegrown chip industry's 2016 sales were worth just $63.1 billion, including $23.9 billion from sales of chips designed in China, according to China Semiconductor Industry Association data.

Chip design is considered the most sophisticated aspect of the industry. "Most flagship handsets use Qualcomm's processors. But its production capacity in high-end chips is limited. These chips required cutting-edge manufacturing techniques, which need time to mature. The initial investment size is often conservative,"Mo of Canalysys said.

"If the majority of Qualcomm's Snapdragon 835 processors go to Samsung, Chinese players such as Xiaomi would suffer from supply shortage and have no other choice but to delay their product launches."

The situation is unlikely to improve any time soon.

For, Chinese smartphone vendors are itching to compete in the high-end market. So, chip shortages will likely get more acute, experts said.

To be fair, Chinese labels such as Xiaomi and Huawei did anticipate this problem and made efforts to develop their own alternatives. For instance, Xiaomi unveiled its first in-house chip in March.

Way back in 2004, Huawei started its research and development of chips. It can now supply a number of mobile processors for its high-end devices, though it still leans a bit on Qualcomm for the "system on a chip" technology.

Chinese chip maker Tsinghua Unigroup is working on a $30 billion domestic memory chip production complex in Nanjing, Jiangsu province. But it is still several years away before the plant begins to produce chips.

Roger Sheng, a senior chip analyst at research firm Gartner Inc, said: "In-house design and manufacture of chips for specific devices would help deliver differentiated product performance. It can also boost smartphone vendors' bargaining power with suppliers.

"But heavy resources need to be poured into the (chip) segment. Each smartphone maker's annual shipments need to reach 10 million units to cover the cost of developing tailor-made chips.

"It's a risky business. Missteps would add weeks to the product launch schedule. Money alone won't guarantee success. But it's a path that Chinese players have to take if they want to rise to supremacy in the high-end market."


b083fe955aa11aa80be202.jpg

@cirr , @xunzi , @AndrewJin , @Han Patriot
 
TSMC plans trial production of advanced 5nm process in 2019
2017/05/26 14:51:27

DArrr-BXYAI9X1E.jpg


Taipei, May 26 (CNA) Taiwan Semiconductor Manufacturing Co. (TSMC, 台積電), the world's largest contract chip maker, is planning to start production of chips made using the sophisticated 5 nanometer process on a trial basis in 2019.

C.C. Wei (魏哲家), president and co-chief executive officer of TSMC, said at a technology symposium held in Hsinchu on Thursday that the chip maker's efforts to introduce advanced technology processes will allow electronic chips to be used in a wide range of applications under four platforms:
  1. mobile devices
  2. high-speed computing devices
  3. automotive devices, and the
  4. Internet of Things.
However, Wei did not reveal when the 5nm process will start commercial production.

The symposium was watched closely by market analysts who are looking for indications as to TSMC's outlook and technological progress being made by the global semiconductor industry.

TSMC's latest mass production technology is based on the 10nm process which was introduced in the fourth quarter of last year. The chip maker is currently devoting considerable effort to developing 7nm, 5nm and even 3nm processes, in a bid to maintain its lead over peers in the global wafer foundry business.

TSMC currently accounts for more than 50 percent of the global contract chip manufacturing market.

3dchips.jpg

TSMC actively developing three dimensional stacked architecture for 5 nanometer process and investing for 3 and 1 nanometer manufacturing process
https://www.nextbigfuture.com/2016/10/tsmc-actively-developing-three.html

Discussing the 7nm process, Wei said, to date the chip maker has rolled out 12 tap-out designs based on the technology. A tap-out design is the final result of the design process for integrated circuit suppliers.

Wei said the mass production of chips made using the 7nm process is scheduled for 2018.

Although Wei did not elaborate on progress made in developing 3nm processes, local media have widely reported that TSMC is likely to build a new production site for 3nm process production and has speculated that the plant could be located in the United States.

However, TSMC reiterated on Thursday that plans for a 3nm process plant will not be finalized until the first quarter of next year.

Wei said that TSMC is planning to assign US$10 billion to capital expenditure to expanding production capacity and product development in 2017, including US$2.2 billion investment in research and development, to meet demand from clients which currently total about 450.

While TSMC declines to comment on client details, the chip maker is believed to supply processors for iPhone production and its major clients include Qualcomm Inc., NVidia Corp. and MediaTek Inc. (聯發科).

At the technology symposium, Tsai Chih-chun (蔡志群), a section director at TSMC's Asia Pacific operations, said that sales in the global semiconductor business rose 7 percent from 2016 to US$383 billion in 2017.

Tsai said that artificial intelligence, virtual reality, augmented reality and automotive alliances are expected to drive semiconductor business this year.

He also forecast that smartphone shipments worldwide will rise 6 percent from last year to reach 1.552 billion units in 2017, with shipments to the China market expected to grow 10 percent.

(By Jackson Chang and Frances Huang)
Enditem/AW

http://focustaiwan.tw/news/aeco/201705260012.aspx
 
Alibaba Fuels China's CPU Gambit
C-Sky to leverage Alibaba's investment, Yun OS
Junko Yoshida
6/20/2017 03:11 PM EDT


AUSTIN, Texas — C-Sky Microelectronics Co., designer of China’s home-grown 32-bit embedded CPU processing cores, is quite possibly China’s best kept secret.

C-Sky’s core business is designing a 32-bit high-performance and low-power embedded CPU, then licensing the chip architecture. Since 2003, when it rolled out its first CPU core called CK510, C-Sky has been quietly developing a host of embedded cores, SoC platforms, software tools and middleware. The company, now offering seven cores, has picked 70 licensees so far in China.

C-Sky’s embedded CPU cores aren’t for baseband or apps processors for smartphones. “We’ve avoided that market,” said Qi. Instead, C-Sky is promoting its CPU cores for embedded and IoT applications. They include systems for digital audio and video — including AI-based processing and recognition engines (for smart speakers like Amazon’s Echo), information security, network and communications, industrial control and automotive electronics.

C-Sky's ace in the hole might be its unique tie with Alibaba Corp., headed up by Jack Ma — one of three Chinese Internet behemoths often lumped together as “BAT” companies, Baidu, Alibaba and Tencent.

It turns out that C-Sky is the first chip company in the world to receive a sizable investment from Alibaba. Attention from an Internet giant like Alibaba isn’t just unusual. For any chip hardware startup struggling to get funded, this sort of investment is pennies from Heaven.

Although Qi declined to disclose the amount provided by Ma, he revealed that he and his mentor, Prof. Xiaolang Yan, C-Sky’s founder and chairman, have lobbied Alibaba to acknowledge that hardware technology is critical to the future of the e-commerce giant.

full article
http://www.eetimes.com/document.asp?doc_id=1331916&page_number=1
 
Fosun Unit Leads Series A Round In Chinese AI Chip Maker Westwell Lab
Pan Yue June 23, 2017 — 18:02 HKT

part01_01.jpg


A unit of Chinese conglomerate Fosun Group has led a series A financing round in Westwell Lab (上海西井科技), a Shanghai-based artificial intelligence chip maker start-up specialized in Neuromorphic engineering.

Fosun made the investment via its venture capital unit Fosun Tonghao, in the Chinese entertainment and investment conglomerate's first major foray into the AI sector. No financial details of the round were disclosed.

Founded in 2015, Westwell is unique among Chinese AI start-ups. Instead of providing AI capabilities such as voice recognition, image recognition or other tech solutions to enterprise clients, Westwell is focused on developing an AI chip the size of a stamp simulating the activities of human nerve cells.

The company's products include a Neuromorphic chip named DeepSouth, which can teach, train and enhance itself with high level of integration, and a Neuron chip named DeepWell, which is more efficient in the self-learning process than traditional GPUs, the company said.

"AI chip is an important part and the foundation of the entire AI industry. Westwell's brain-like chip can be applied in a number of sectors including precision medicine, smart terminal, financial security and autonomous driving. I believe the company has great potential," said Liu Qikai, founding partner of Fosun Tonghao.​

U.S. chip maker Nvidia Corporation dominates the existing GPU market, with more than 70% market share. GPUs, or graphics processing units, are the most popular chips powering AI currently. Other tech companies including Google have entered the field betting on the market's future growth potential.

While most Chinese AI companies are focused on providing tech solutions, numerous AI chip start-ups have emerged in China lately. Hangzhou-based AI chip maker Canaan, for example, raised a RMB300 million (US$43 million) series A round last month from a number of investors including Chinese hotel operator Jin Jiang International Group Co., Ltd.

Westwell currently has a team of 40 AI engineers from Imperial College London, University of Oxford, and University of California, Berkeley. It plans to use the latest proceeds to invest in research and development.

https://www.chinamoneynetwork.com/2...a-round-in-chinese-ai-chip-maker-westwell-lab
 

Pakistan Defence Latest Posts

Pakistan Affairs Latest Posts

Back
Top Bottom