SoC design challenges and needs are diverse. There are many diverse IP blocks. It is time consuming to verify. Picking correct IP is critically important.
Speaking on the TSMC Soft IP Alliance program, Dan Kochpatcharin, deputy director, TSMC said that IP sourcing priorities include is it available, Is it from trusted partners, how is the design quality, and what are the specs and cost? Some other points to note are: is an IP verified, has it been silicon proven, what has been tested and how many in production volume already?
TSMC Soft IP Alliance has 5000+ IP titles from 40+ IP vendors. The IP Alliance program has been expanding. It is leveraging on successful IP. More and more customers are concerned about PPA data of soft IP specific technology when doing system design.
The Soft IP Alliance has 16 members. The Soft-IP quality assessment TSMC 9000 is key. New soft-IP handoff kit were rolled out in Nov. 2012. Major partners have now joined to drive soft-IP quality, such as Imagination, Sonics, MIPS, etc.
TSMC Soft-IP 9000 has carried out industry first QA assessment system for RTL based IP. TSMC and IP partners co-optimize RTL/process to deliver PPA optimized IPs.
Mike Gianfagna, Atrenta, spoke on the implementing program with Atrenta IP kit. Atrenta’s SpyGlass is a systematic approach to soft IP quality.
If you look at what’s needed for IP assessment, there are factors such as right abstraction levels must be supported, and soft IP is delivered as generators, RTL or gates; the biggest need is here. The IP must be comprehensive, easy to use, objective and quantifiable, actionable, and dynamic and scalable. Atrenta and TSMC announced SpyGlass IP kit 2.0 in October 2012
What does the IP kit check? Many items that would impact the integration/debug time and chip function were found and fixed. Soft IP qualification can be automated. It results in higher quality deliverables. All soft IP can be improved. Primary beneficiaries are chip integrators.
John Bainbridge, Sonics, spoke on the practical results of program participation. Sonics is a leader in system IP for SoCs. It enables designers to integrate any IP from anywhere, anytime.
Sonics helps leading SoC vendors solve some of the most difficult challenges in SoC design. These can be IP integration, high frequency, memory throughput, security, physical design, power management, development costs, and time-to-market. Sonics is a lead beta partner for TSMC Soft IP 2.0 kit program. It has worked closely with Atrenta and TSMC to ensure a seamless design flow.
Milpitas, USA-based Sonics Inc. participated in TSMC’s Soft IP Alliance 2.0 beta program. Driving high quality soft IP eases customer integration and expedites time-to-market.
Sonic’s role in TSMC beta program
Speaking on the beta program and Sonics’ role, Frank Ferro, director of Product Marketing, Sonics, said: “TSMC’s Soft IP kit 2.0 beta program is part of TSMC’s Open Innovation Platform program that creates a complete ecosystem for customers with the overall goal of shortening design time. This is done by providing a large catalog of partner provided IP that is silicon-verified and production-proven.
For vendors like Sonics, TSMC has extended this ecosystem to include Soft-IP (IP not designed for a specific process, but delivered as RTL). The program allows Soft-IP partners to access and leverage TSMC’s process technologies to optimize power, performance and area for their IP.
IP cores are checked through TSMC’s foundry checklist to ensure the customers have optimized design results with fast IP integration built into their design. This flow also facilitates easy IP reuse for subsequent designs. The soft IP Kit beta 2.0 program is an extension of the current program through implementing additional quality checks, improving results and making the flow easier for customers.
There are several advantages to Sonics as a participant in this program. First, customers of TSMC will have access to Sonics IP through TSMC’s IP library. Given TSMC’s strong market share, this will allow Sonics IP to be visible to a large customer base. In addition, TSMC’s customers will feel securing using Sonics IP because they know that it has been put through a rigorous series of IP checks that meet the highest quality standards. It also allows Sonics early access to TSMC’s process libraries, allowing Sonics to optimize performance and area for each IP product.
So, what can the TSMC’s Soft IP Kit 2.0 do? How does Sonics enhance its capabilities? The Soft IP Kit 2.0 provides a specific RTL design flow methodology and hand-off which includes: lint (RTL coding consistency), clock domain crossings (CDC), power (CPF/UPF), physical design (routing congestion), design for test (DFT), constraints and documentation.
Using this flow enhances Sonics IP quality and reliability because many RTL errors can be caught at an early stage. As mentioned above, this flow ensures lowest power and best performance of the IP for a given process node.
Atrenta SpyGlass improves packaging
There is a role played by Atrenta SpyGlass. According to Ferro, Atrenta SpyGlass is the tool used to run all the tests. The flow was developed to TSMC’s standards and implemented by Atrenta. Given Sonics strong relationship with TSMC and Atrenta, we were invited to be a beta partner using our IP to test the new flow. A number of companies do participate in the program, although only Sonics has announced participation in the beta 2.0 program to date.
This tie up with Atrenta will likely improve IP packaging. As part of the overall flow, the final step, after all basic and advanced IP checks, is IP packaging. This step includes providing the IP with information on the design intent, set-up and analysis reports. Again, this is done using the SpyGlass tool from Atrenta.
This IP packaging was available to customers in the past via the Soft IP 1.0 program. The attraction of this type of IP packaging is a result of the growing number of IP cores being integrated into complex SoCs. As the number of third party IP grew, the need for a better, broader methodology was developed.
TSMC unveiled its schedule for 450mm mass production at the recently held SEMICON Taiwan 2012 450mm Supply Chain Forum. Focusing on lithography as the key, Dr. C.S. Yoo, senior director of the 450mm program at TSMC, noted that IC makers and equipment suppliers should fully leverage the G450C. They need to work and innovate to make the 450mm transition a great success.
TSMC has always been in the relentless pursuit of technology innovation. It has been part of all of the computing waves that have driven the market growth. Right now, mobile computing is the leading market driver. TSMC has been helping the industry produce comprehensive, powerful mobile computing devices.
The future growth drivers and trends include mobile computing, cloud computing and smart devices. However, technical and economic challenges also lie ahead. TSMC has been pushing the lithography roadmap. 28nm is said to be the limit of conventional single-patterning lithography. TSMC has innovations to extend immersion to 20nm. The next-generation lithography (NGL) is being preferred beyond 20nm. Also, EUV and multiple-e-beam concept and feasibility has been proven. The more than 10x throughput gap requires collaborative innovation and funding.
TSMC continues to invest in R&D for transistor architecture trends. There is increasing technology complexity, as reflected by mask layers increase. The technology shrink also leads to design complexity.
There are challenges such as intrinsic wafer cost parity and uncertain technology migration ROI. TSMC’s mission is to be the trusted technology and capacity provider for the global logic IC industry for years to come. TSMC already has capacity leadership. TSMC’s total 12″ cleanroom space will equal more than 32 World Cup football fields by the end of this year..
TSMC customers’ expectations include the offer of leading-edge technology, continue to expand capacity, enable faster time to market, faster technology ramp up, faster manufacturing cycle times, and lower cost /die. To bridge the cost and productivity gap, TSMC no longer maintains cost/transistor trend by 2018 due to the slowing pace of technology shrink, and increasing technology complexity.
IC Insights recently released the 1H-11 top 20 semicon sales leaders. No surprises here, with Intel, Samsung, TSMC, TI and Toshiba as the top five leaders in that order. In all, 10 of the top 20 suppliers outperformed the total global semiconductor industry 1H11/1H10 growth rate of 4 percent.
The fabless companies — notably, Qualcomm, Broadcom, and so on, have registered positive growths. However, if you really look carefully, a lot of the companies thereafter have registered negative growth for the period 1H-11 over 1H-10.
What’s surprising to notice is the fact that at least seven companies — Renesas, Hynix, Micron, AMD, Infineon, Elpida and NXP have registered negative growth! This, during a period when the semiconductor industry was said to be on the rebound? Whatever the reasons, they are all in the red!
Now, we are not spent from discussing an industry turnaround, which is perhaps there! Also, the forecast for 2H-11 isn’t something to go overboard. IC Insights expects the 2H11/1H11 semiconductor market to grow only 6 percent, that is, a full-year 2011 semiconductor industry growth rate of 5 percent.
Closer to home, as usual, there are no Indian firms in the global top 20 list. As things stand, they may not even make it to the list, at least, for quite a while. One hopes that this situation somehow changes. Wonder, how did the India Semiconductor Association (ISA)-Frost & Sullivan study come up with a figure of 28.3 percent growth in 2010! Perhaps, I am mistaken in my calculations somewhere!!
This is a summary by Malcolm Penn, CEO, Future Horizons. For those who wish to know more, please get in touch with me or Future Horizons.
December’s WSTS results were as boring as they were predictable, with no serious data revisions (thankfully) and the results right where we expected. December’s year-on-year IC unit growth was 8.9 percent that, with the 3.5 percent growth (yes GROWTH) in ASPs, yielded a respectable double-digit value growthof 12.8 percent. And this, on the back of a weak Q4 memory market that saw ASPs fall 13.1 percent vs Q3-10!
The yearly growth vs 2009 weighed in at 31.8 percent, hitting $298.3 billion, just shy of the elusive $300 billion threshold. The market is right where we said it would be at our recent 2011 Forecast seminar; we reiterate our position that 2011 will be a good year for the industry. Choppy first-half waters for sure, but watch out for a whopping 2H-11 ricochet.
Connectors are up as well
It is not just semiconductors that are off to a good start. The connector industry is tight as a drum too. Orders in December 2010 were up 13.3 percent versus December 2009, with full year orders up 29.3 percent on 2009, down sequentially 11.1 percent from November 2010. The comparable data for sales was plus 18.7percent, plus 28.4 and minus 13.7 percent.
The December connector book-to-bill ratio was 1.01, unchanged from November. This industry still publishes orders and book-to-bill data by the way, unlike the chip industry which very foolishly stopped publishing this several years ago. All this in the seasonally slow first quarter of the month, yet few people believe there is a supply problem in prospect. Just as this time last year, industry denial is rampant, way beyond reasonable caution and ignoring the underlying trends.
Strong demand for mobile, server and graphics DRAM
We estimate that the worldwide growth rate for PCs in 2011 will be a healthy 10 percent, with 3.9GB the average DRAM content per box. New capacity and die shrinks are putting near-term pressure on over-supply and pricing but there are now move afoot from Elpida and others to start raising prices.
Where they can, to gain a price advantage, DRAM vendors are actively adjusting their supply in favour of mobile from commodity DRAM, given the current strong demand in the smartphone and tablet PC markets, with a 1GB per box average DRAM content.
Server demand continues to be the other star segment, not just in unit demand but in content per box as well, estimated to average around 30GB in 2011. This will drive a 50 to 60 percent increase in server DRAM demand. Finally in graphics demand for specialty DRAM is also very strong, driven by the rapid take off of3D-TV and continuing strong growth in Blue-Ray DVD.
The overall DRAM industry is thus gradually diversifying from manufacturing mainly commodity DRAM to diversified products such as mobile DRAM, serverbasis DRAM, specialty DRAM and graphic memory.DRAM vendors however are faring mixed fortunes, with Elpida and Hynix having the worst net cash positions with barely enough cash to cover their short-term debt.
The Taiwanese vendors find themselves stuck in a technology trap, unable to invest in the immersion technology needed to break through the 5*nm node, meaning that in the absence of a good market uptick to improve cash flow and profits, a shake out in the DRAM supply base seems unavoidable.
Congratulations to TSMC for making it to the top 10 R&D spenders during 2010! If you look at the IC Insights’ table (see here), you’d understand what I am referring to!
First, the table itself. It has no major surprises, barring TSMC, which again is not really a surprise. As IC Insights itself says, “The industry is increasingly dependent on the success (and hopefully not, failure) of the foundries to continue advancing their IC manufacturing capabilities.”
TSMC has jumped up from 19th to the 10th place in the latest R&D spend. IC Insights expects that TSMC’s R&D spending in 2011 will grow another 20 percent, putting its budget over $1.1 billion for the year.
Otherwise, the top 9 in the list consist of Intel, Samsung, ST, Renesas, Broadcom, Toshiba, Qualcomm, TI and AMD — all big, strong players. However, TSMC, is the only foundry in that list, at the 10th spot.
Now, all of this makes great reading! Everyone connected with the global semiconductor industry is very well aware of TSMC’s strengths and capabilities.
The more TSMC grows in stature, the more will its capabilities grow. A lot of firms have already exited the manufacturing industry, leaving that to the likes of TSMC and others, and some more are due soon.
However, I am thoroughly disappointed by some folks who touch upon India missing out on the R&D story. More of that in a bit!
First. where is the R&D strength of India? There are firms, based outside the country, managed by Indians, who seem to look down on the country. I even received an email from a gentleman, which says: ‘Without semiconductors, India cannot gain technological advantage. Semiconductor should be funded by the defense budget.” I am appalled!
Well, we are trying, aren’t we? There are names that come to the mind — Procsys, Ittiam, SoftJin, eInfochips, MindTree, and so on! Yes, I know they aren’t exactly world beaters. But hey, they are all doing their own thing reasonably well!
“During the first three quarters of 2010, foundries were under intense pressure to meet customer demand,” said Len Jelinek, director and chief analyst for semiconductor manufacturing at iSuppli. “The pressure is leading to increased revenue, as consumer spending has come back with a vengeance following a dramatic downturn in the fourth quarter of 2008 and for all of 2009.”
By 2014, total pure-play foundry revenue will reach $45.9 billion, managing a CAGR of 9.4 percent from $26.8 billion in 2008. Pure-play foundries are contract manufacturers whose business consists of producing semiconductors on behalf of other chip companies.
Enhancing foundry forecast
I started by asking Jelinek what were the chief reasons for enhancing the foundry forecast. Jelinek said: “The forecast increase is based on the anticipated strength in demand for products in Q2 and beyond. Additionally, it is also simple math. The foundry market had a good Q2, and last year, Q1 and Q2 were quite challenging. So, by having a good first half of the year, the percentage must increase.”
Also, given that there has been renewed demand for consumer electronics products, what are the specific CE products, besides netbooks, mobile phones, that have been seeing renewed demand, and why?
While speaking on ‘Enabling business growth through effective and collaborative innovation’, at the recently held International Electronics Forum (IEF) 2010 in Dresden, Germany, Dr. Jack Sun, CTO and vice president, R&D, TSMC said that TSMC leads and invests heavily in competitive, energy efficient, and eco-friendly technologies to enable product innovation, such as CMOS platform scaling (40/28/20nm/FinFET, low-R,ELK..), More-than-Moore, and integrated package/3D-IC.
He added that TSMC strives for manufacturing excellence and capacity, and economy of scale, to support customers’ innovation and business growth. The company is also pushing the acceleration of EUV and Multi-Ebeam capabilities for cost-effective density scaling. His clear message was, “We must and can collaborate to innovate and overcome the technical and cost challenges.” That is, a collaborative innovation among the government, the industry and the academia is required to overcome the cost hurdle.
Earlier, he said that the IC industry will continue to grow — with a 22 percent growth likely in 2010, reaching $276 billion. During 2011-2014, he estimated a 4.2 percent CAGR for the IC industry and 7.2 percent CAGR for fabless companies.
Dwelling on the application and technology trend, Dr. Sun pointed out that the trend is SoC and heterogeneous integration at chip, package, and product level with embedded power-efficient processors, hardware accelerators, and special features.
TSMC continues to further expand its offering by including packaging services and silicon foundry services. This will allow the fabless semiconductor companies to achieve ‘More than Moore’ gains in integration by using TSMC as a foundry partner.
Dr. Sun also detailed the how TSMC enables innovation by providing best-in-class technology and design solutions.
TSMC’s 20nm and 28nm leadership
While on TSMC 28nm technology highlights, Dr. Sun said that the 28LP (poly/SiON) yield is approaching mature level on 64Mb SRAM. Also, the 28nm HKMG (28HP/28HPL) development is on track. Here, TSMC developed Gate-Last process with N+/P+ work function and superior performance, yield, manufacturability, variability,and reliability.
Also, it achieved double-digit 64Mb yield, good Vccmin, close-to- targets transistors, and good pre-qual reliability.
Now, on to TSMC’s 20nm highlights. The key technology features include planar transistors with 2nd-generation HKMG and 5th-generation strained Si; low-resistance ultra-shallow junction with M0 and enhanced millisecond anneal and silicide; and enhanced ELK and 2nd-generation Low-R interconnect.
Some other 20nm tehchnology highlights include immersion lithography with innovative patterning and layout solutions to achieve 2x density over 28nm, with the EDA tool likely to be ready by mid 2010. Also, the design rules are compatible for EUV and Multi-Ebeam insertion for selected layers in 2013-2014. Wow, this is really something! Read more…
Next, at the Intel Developer Forum in Beijing, China, it outlined plans for a new Atom processor-based SoC. It is codenamed as the Tunnel Creek SoC for IP phones, printers and in-vehicle-infotainment systems for cars. Excellent!
Then, at its 2010 Technology Symposium, TSMC announced that it will skip the 22nm manufacturing process node and move directly to a 20nm technology. In fact, it proposes to enter 20nm risk production in the second half of 2012. Brilliant!
Now, I have a release from Future Horizons that outlines the state of the global semiconductor industry. On the one hand, Future Horizons indicates that semiconductor sales have continued to be very strong. This looks set to continue for the rest of the year, resulting in a 2010 that is massively improved on 2009.
On the other hand, several companies still remain unjustifiably pessimistic and confused about the state of the market. Companies now have an opportunity to dominate the market. Instead they continue to be cautious, undermining their own prospects of making some serious money. Isn’t that confusing for the industry? Or, is it confusing itself?
I also have a report from Accenture titled ‘Flying blind in the semiconductor industry’, which you can read on the PC Semicon Blog.
According to Scott Grant, managing director with Accenture’s Semiconductor Business, the fallout from the global recession, massive fragmenting of the value chain, the rise of a more diverse world economy, and new sales and distribution models have created tough challenges for semiconductor companies when it comes to understanding and predicting demand for their chips and managing their supply chains. This lack of understanding is a dangerous liability in a world characterized by unprecedented volatility and competition.
Oh my, these are clearly mixed signals all over again! Or, is the global semiconductor industry having problems with its ‘place-and-route’ strategies? Just a fiigure of speech!
For one, is the global semicon industry truly flying blind? Scott Grant has given suggestions as to how the challenges can be tackled. He advises semiconductor companies to focus on three priorities: sales force effectiveness, supply chain integration and optimizing their collaborative planning and fulfillment capabilities.
Look, I’m not an expert! There are several questions that need to be asked, and I hope some knowledgeable folks can answer those.
For one, should the global semicon industry continue to revel in ‘inappropriate pessimism’,’ how will it affect its fortunes in the short and long terms? Or, are those strategies, as advised by Accenture, enough to help the industry? Next, why this need to skip process nodes? What happens to those betting on 22nm? Okay, will all of that have some impact on the semiconductor equipment industry in the long term? What’s really happening with the global semiconductor industry?
There is a need to swing back to optimism, folks, as Malcolm Penn of Future Horizons says in his monthly update.
As I’m about to call it a day (or evening or night), comes the news that EDA industry organizations, Accellera and The SPIRIT Consortium, have completed their merger!
Didn’t I tell you at the very beginning that this is turning out to be quite a week in the global semiconductor industry?
The next week promises to be fun, especially in the Indian semiconductor/VLSI/electronics industry! Well, it has to do with microelectronics! You’ll find out soon. Keep reading this blog, friends.
Recently, IC Insights released the rankings for the world’s top pure-play and IDM foundries. No surprises, as TSMC continues to lead! The surprise entrant is of course GlobalFoundries, which ranked fourth, after having started operations in March 2009.
However, all of the foundries, barring Tower Semiconductor, registered negative growth during 2009. Tower, which acquired Jazz Semiconductor in 2008, was the only foundry to post positive growth during 2009.
IC Insights further stated that if the revenues of Chartered Semiconductors, which was purchased by GlobalFoundries recently, were combined with GlobalFoundries, their combined sales would have amounted to over $2.6 billion in 2009. That’s not very far from UMC, which is ranked no. 2!
Now, I am not an expert to comment on which foundry has done really well, despite the recession, or whether GlobalFoundries can really challenge and overtake TSMC in the future. However, I followed with great interest a discussion on one of my groups on LinkedIIn on this topic.
Daniel Nenni, a critically acclaimed blogger, and an industry colleague, had recently blogged on this subject. There were some interesting comments following that post. I sought the permission of Malcolm Penn, chairman and CEO, Future Horizons, to use some of his remarks for my blog post.
Penn said that there is a reason why TSMC is the no. 1 foundry in the world. However, as competition breeds innovation, hence, the foundry business will be much more interesting to watch with GlobalFoundries challenging TSMC.
He added: “There was also no reason why Intel is #1 in PC MPUs or Microsoft #1 in PC OS except for one key factor. Once a competitor gets to be a certain size, no amount of innovation will help swing the balance of power.” The real question therefore is: “Has TSMC now passed this tipping point?” I fear it already has.”
Fact of the matter is in the foundry business capacity is king and TSMC already outguns all of the other competitors combined. Given the minimum two to three year lead time to build and ramp new capacity, TSMC has incredibly clear visibility to see any competitive threat coming.
Very interesting! Some other observations later!