Opinionated - China Chipping Away to Semiconductor Dominance

[qwerrty]

more incentives from gov to semiconductor and software industry..

MIC 2025 is alive and stronk

国常会定了！促进集成电路和软件产业向更高层次发展
宋笛2019-05-08 21:55
http://www.eeo.com.cn/2019/0508/355412.shtml

 

[Han Patriot]

if everything goes smoothly that would be the same level as intel as they are still struggling with 10nm process.

I was trying to explain to boy genius bussard, apart from TSMC and Samsung, most are still struggling with 7nm especially Intel. When China announces they will achieve something, they will achieve it. We don't like to boast like our southern black neighbours.

 

[j20blackdragon]

Zhaoxin plans to complete verification，tape-out and test work on its 7nm x86 processor KX-7000 this year。

I wish the company good luck。

http://www.zhaoxin.com/

https://twitter.com/x/status/1000740115277406208

 

[cirr]

Xtacking 2.0锁定8月，长江存储或迎跨越式发展



2019-05-16 16:26:49 来源：观察者网

关键字:长江存储Xtacking 2.0

（文/观察者网 陈兴华）

5月15日，在GSA Memory+高峰论坛上，长江存储联席首席技术官汤强表示，为了应对数据量增长对存储器的挑战，长江存储将在今年8月正式推出Xtacking 2.0技术。他还表示，利用Xtacking技术，NAND的I/O速度有望达到行业颠覆性的3.0Gbps。

汤强首先提到，当前正处于一个数据大爆炸的时代，预计到2022年，NAND的容量需求将会接近2015年的10倍，并将持续增长下去。但是实际上，数据和存储能力的增长严重不对等，存储容量的增长远远落后于需求的增长。

但是，他同时表示，目前为了追求更高的存储容量，存储技术正从MLC向TLC向QLC等发展。另外，对于各种应用场景的需求，实际上是需要越来越多的定制化的存储方案来与之对应。而一项新的存储解决方案的开发往往需要12-18个月的周期。

汤强将上述问题总结为行业正面临三大挑战，包括I/O接口速度、容量密度和上市周期。对此，他强调，长江存储去年推出了Xtacking技术，该技术不仅提高了I/O接口速度，而且还保证了3D NAND多层堆叠可达到更高容量，以及减少上市周期。

据介绍，采用Xtacking技术，可在一片晶圆上独立加工负责数据I/O及记忆单元操作的外围电路，使得NAND获取更高的I/O接口速度及更强的操作功能。存储单元将在另一片晶圆上独立加工。当两片晶圆各自完工后，XtackingTM技术只需一个处理步骤就可将二者键合接通电路，且成本增加有限。

长江存储日前公布的数据显示，在传统3D NAND架构中，外围电路约占芯片面积的20-30%，这也使得芯片的存储密度大幅降低。而随着3D NAND技术堆叠到128层甚至更高，外围电路所占据的芯片面积或将达到50%以上。Xtacking技术则可将外围电路置于存储单元之上，从而实现比传统3D NAND更高的存储密度。

据了解，目前NAND闪存主要掌握在三星、东芝、美光、西数等公司中，而长江存储只小批量生产32层堆叠闪存。不过，将于今年底量产的64层Xtacking 3D NAND是长江存储今明两年生产的主力。集邦咨询半导体研究中心预计长江存储到年底至少为60K/m的投片量，而其竞争者达到200K/m以上。

为了尽快缩短与国外厂商的差距，长江存储采取了“跨跃式”发展。在下一代闪存技术上，长江存储将直接跳过三星、美光、东芝、Intel等公司现在力推的96层堆叠闪存，直接进入128层堆叠的研发，这几家公司预计在2020年才会推出128层堆叠的闪存。

据悉，在I/O速度方面，目前NAND闪存主要有两种I/O接口标准，分别是Intel/索尼/SK海力士/群联/西数/美光主推的ONFi，目前ONFi 4.1标准的I/O接口速度最大为1.2Gbps。第二种标准是三星/东芝主推的ToggleDDR，I/O速度最高1.4Gbps。

对此，汤强在本次论坛上表示，目前世界上最快的3DNAND的I/O速度目标值是1.4Gbps，大多数NAND供应商仅能供应1.0Gbps或更低的速度。利用Xtacking技术，有望大幅提升NAND的I/O速度到3.0Gbps，与DRAM DDR4的I/O速度相当。这对NAND行业来讲将是颠覆性的。

另外，他还提到，Xtacking技术还可使得产品开发时间缩短三个月，生产周期可缩短20%，从而大幅缩短3D NAND产品的从开发到上市周期。与此同时，Xtacking依然处于不断进化之中。

目前，Xtacking技术已经成功应用于即将量产的长江存储的第二代3D NAND产品（64层堆叠）。汤强还表示，今年8月推出新一代的Xtacking 2.0技术有望将长江存储的3D NAND提升到一个新的高度。Xtacking将是存储的未来。

 

[kuge]

Xtacking 2.0锁定8月，长江存储或迎跨越式发展



2019-05-16 16:26:49 来源：观察者网

关键字:长江存储Xtacking 2.0

（文/观察者网 陈兴华）

。

is he 汤 with yangtze at the back drop?

 

[Bussard Ramjet]

I was trying to explain to boy genius bussard, apart from TSMC and Samsung, most are still struggling with 7nm especially Intel. When China announces they will achieve something, they will achieve it. We don't like to boast like our southern black neighbours.

Well this isn't true. I remember SMIC announced earlier that it will have 14 nm by 2017 or so, but that was not the case. China doesn't achieve everything that it announces or plans.

Also, as for node technology is concerned, TSMC and Samsung are enough. For such high tech stuff in almost any sphere there are only 2-3 companies that compete.

 

[Han Patriot]

Well this isn't true. I remember SMIC announced earlier that it will have 14 nm by 2017 or so, but that was not the case. China doesn't achieve everything that it announces or plans.

Also, as for node technology is concerned, TSMC and Samsung are enough. For such high tech stuff in almost any sphere there are only 2-3 companies that compete.

Chinese don't usually announce something they are not sure, they might say we aim to achieve this, not we will achieve this. 14nm just matured last year? China doesn't need to be the best, just being in the race is good to maintain industrial base. You think TSMC and samsung makes their own material and equipment? China probably has a more complete chain than both of them.

 

[Bussard Ramjet]

Chinese don't usually announce something they are not sure, they might say we aim to achieve this, not we will achieve this. 14nm just matured last year? China doesn't need to be the best, just being in the race is good to maintain industrial base. You think TSMC and samsung makes their own material and equipment? China probably has a more complete chain than both of them.

Oh no! SMIC has announced multiple times many things that they will do, which ended up being left undone. You can go back in this very thread to find them.

Neither Taiwan nor South Korea are vying for superpower status, are they? They are allies with US and other Western Countries and as such don't face threat from being cut off from western supply chains.

Also, being among the best is actually really required. Look at Chinese smartphone companies. They have major shares in the international market. However today even the medium end phone has a chip made on a 10 nm or 14 nm process. I have multiple mobile including $150 mobiles from Xiaomi that have qualcomm chipsets fabricated on 14 nm. If these chinese companies are denied access to chips from these advanced nodes, then chinese companies will pretty much lose ALL of their market share abroad.

Also, 14 nm in China hasn't matured yet. The mass production hasn't even begun yet. Maturation comes after mass production has begun and the yields and cost has been optimized.

 

[JSCh]

Just want to say that SMIC is listed in both HKSE and NYSE. Listed company would be in really hot water with SEC if you announced something that is not true.

 

[cirr]

OPPO will allegedly announce a smartphone with under display camera in 2019. The full display will have a front camera hidden behind the pixels. When the camera is launched, the pixels over the camera are dislocated and the camera lens will appear. (Android Headlines, Tencent, Sohu, Twitter, GizPrix)



According to TrendForce, Yangtze Memory Technologies (YMTC), which forms the main bulwark of NAND flash development in China, will be mass producing 64-layer Xtacking 3D NAND products this year end as scheduled. YMTC’s capacity expansion in 2020 is projected to impact supply and prices in the global NAND flash market. (CN Beta, TrendForce[cn], TrendForce, press)



Xtacking 2.0 will be launched in August 2019.

 

[JSCh]

'Unstoppable' Chinese NAND fabber YMTC to unleash 64-layer flash flood before skipping ahead to 128 – analyst
You thought the memory glut was almost over? Think again
By Chris Mellor 14 May 2019 at 18:00

Analysis Chinese NAND flinger Yangtze Memory Technologies (YMTC) will be mass-producing 64-layer 3D NAND flash chips by the end of the year and price competition could get ugly in 2020.

That's the opinion of research outfit TrendForce's DRAMeXchange division. It said YMTC had already sent samples of its 64-layer Xtacking chips to potential customers and controller suppliers, with a focus on the domestic Chinese market. YMTC is limiting 32-layer chip production at its new Wuhan plant, ready to switch on 64-layer processes.

It is said to plan output of 60,000 wafers a month in 2020, enough to delay recovery from the current oversupply of NAND, which has caused prices to drop. The analyst said this is expected to certainly have an "impact on NAND flash market prices and prolong the falling trend".

YMTC's 64-layer product will compete with more advanced NAND from its competitors. Suppliers such as Intel/Micron, Samsung, SK Hynix, Toshiba and Western Digital are controlling production volumes while bringing 92 to 96-layer processes on stream, which will lower per-bit costs.

Beyond that, TrendForce said it sees YTMC moving to 128-layer NAND in 2020 to lower costs again and increase bit shipments, as its chart shows:

The analyst envisioned YMTC jumping straight from 64 to 128-layer NAND in the second half of 2020, bypassing the 96-layer process, at which point all the NAND suppliers will be on equal footing.

TrendForce declared: "YMTC's impact on the future market is inevitable and unstoppable."®


'Unstoppable' Chinese NAND fabber YMTC to unleash 64-layer flash flood before skipping ahead to 128 – analyst • The Register

 

[cirr]

Shanghai-based Zhaoxin Semiconductor is launching Kaixian series KX-6000, which features the real SoC design, using 16nm process. The processor is based on x86 standard architecture, compatible with x86 32/64-bit instructions, supports SSE4.2 / AVX extended instruction set, up to 8 CPU cores, clock speed up to 3.0GHz, supports dual channel DDR4 memory, and has a maximum capacity of 64GB. The KX-6000 chip will be mass-produced in Sept 2019. The first customer service is Lenovo, Tsinghua Tongfang, Onda and other enterprises, for all-in-one, desktop machine and server market. (My Drivers, CN Beta, Tencent)



KX-7000 employing 7nm process is planned for taping-out towards the end of this year.

 

[Han Patriot]

Shanghai-based Zhaoxin Semiconductor is launching Kaixian series KX-6000, which features the real SoC design, using 16nm process. The processor is based on x86 standard architecture, compatible with x86 32/64-bit instructions, supports SSE4.2 / AVX extended instruction set, up to 8 CPU cores, clock speed up to 3.0GHz, supports dual channel DDR4 memory, and has a maximum capacity of 64GB. The KX-6000 chip will be mass-produced in Sept 2019. The first customer service is Lenovo, Tsinghua Tongfang, Onda and other enterprises, for all-in-one, desktop machine and server market. (My Drivers, CN Beta, Tencent)



KX-7000 employing 7nm process is planned for taping-out towards the end of this year.

The issue is not making chips comparable to Intel, the issue is with patents and royalty. The x86 patents are all controlled by wintel and amd. If it's wartime we can produce them at home with no worries, but now we can't sell them overseas.

 

[Bussard Ramjet]

Just want to say that SMIC is listed in both HKSE and NYSE. Listed company would be in really hot water with SEC if you announced something that is not true.

That is not correct. Companies are barred from saying stuff that they know is false. You can't say clearly false stuff that you know is false. However, you can make your own projections for your technology which may or may not meet. In fact all companies, the first thing they mention before talking in an investor conference or call is a disclaimer statement which absolves them of any liability from the event where their forward projections are incorrect.

The issue is not making chips comparable to Intel, the issue is with patents and royalty. The x86 patents are all controlled by wintel and amd. If it's wartime we can produce them at home with no worries, but now we can't sell them overseas.

Umm... it's not only the patents. If it were only the patents companies like Zhaoxin who have access to x86 patents would sell as much as intel. But they don't. x86 is just one thing. Intel spends a LOT of money hiring the best engineers to design literally the best performing chips on the market. This is even clear in the server and supercomputer space where Intel doesn't even have competitors like AMD in that space.

So, process node is one thing, access to x86 is another thing, while your CPU design and chip capability is another thing.

OPPO will allegedly announce a smartphone with under display camera in 2019. The full display will have a front camera hidden behind the pixels. When the camera is launched, the pixels over the camera are dislocated and the camera lens will appear. (Android Headlines, Tencent, Sohu, Twitter, GizPrix)



According to TrendForce, Yangtze Memory Technologies (YMTC), which forms the main bulwark of NAND flash development in China, will be mass producing 64-layer Xtacking 3D NAND products this year end as scheduled. YMTC’s capacity expansion in 2020 is projected to impact supply and prices in the global NAND flash market. (CN Beta, TrendForce[cn], TrendForce, press)



Xtacking 2.0 will be launched in August 2019.

I will be really impressed by both of these companies if these projections come true.

 

[j20blackdragon]

Umm... it's not only the patents. If it were only the patents companies like Zhaoxin who have access to x86 patents would sell as much as intel. But they don't. x86 is just one thing. Intel spends a LOT of money hiring the best engineers to design literally the best performing chips on the market. This is even clear in the server and supercomputer space where Intel doesn't even have competitors like AMD in that space.

So, process node is one thing, access to x86 is another thing, while your CPU design and chip capability is another thing.

Process node is still the primary determinant of performance because it determines transistor size. More transistors means more performance. The smaller the transistor, the less power is required. You want transistor density to be as high as possible for a given die size. It's very simple.

But we haven’t seen a transistor shrink from Intel since 2014.

So how does Intel increase performance at the moment?

Intel does it by producing a power-guzzling giant piece of crap like this.

While real companies produce chips with smaller die sizes and higher transistor counts.

Equalize the die size, power consumption, and cooling requirement and see if Intel is still as good as you think it is.

Also you are dead wrong regarding AMD's ability to compete.

One of the big questions will be if Intel can compete with 28 cores on 14nm when AMD is ready to roll out 64 cores on 7nm, and how the performance will differ. One of the clever things Intel has done in this contest is to draw the talk away from just quoting core counts, and help build a platform around its product.
https://www.anandtech.com/show/14146/intel-xeon-scalable-cascade-lake-deep-dive-now-with-optane