llmda newsletter ad (2)

Decisive Floorplanning for Faster Design Closure

Decisive Floorplanning for Faster Design Closure
by Pawan Fangaria on 01-18-2016 at 4:00 pm

Semiconductor design automation at system level is gaining its due importance today. It needs an effective, efficient, and seamless flow from system up to silicon. There is lot of effort going on for automating SoC design exploration at system level but that eventually stops at RTL; another level of flow automation takes over from RTL to physical implementation, and then to silicon.

Traditional flows from RTL to physical design are inefficient because of long iterations involved between gate-level physical implementation and RTL; sometimes it becomes time-prohibitive or impossible to correct the design at RTL but live with inefficient work-around at gate-level. This is because RTL synthesis doesn’t have physical information until the gate level implementation. Synthesis tools and Floorplanning which ties together the front-end design at RTL level and back-end design at gate-level rely on inaccurate RC and physical models.

As the design sizes are increasing and moving up to system level, it is utmost important and more reason for RTL-to-Gate level flow to become less iterative and be more decisive with accuracy at RTL level to the extent that design signoff can be done at RTL without waiting until physical implementation. A decisive floorplanning tool at RTL level can very well facilitate design prototyping, SoC full-chip planning, as well as IP signoff at RTL level.

I like Mentor’snext-generation physical RTL synthesis and floorplanning tools including RealTime Designer[SUP]TM[/SUP] and other products. The floorplan is created at the RTL level based on the high-level RTL modules, macros, and design data flow. This enables high-level design optimization and accurate timing and congestion analysis. Incremental changes can be done to the floorplan to adjust PPA (power, performance, and area), congestion and routability.

Physical hierarchies are honored and RTL partitions are correctly assigned within the physical boundaries of the appropriate partitions. Mentor’s patented technology is used to access the detailed netlist of each RTL partition to accurately time the design. The synthesis and floorplanning with real physical and timing information makes the flow reliable and predictable, thus reducing the number of iterations and time to design closure.

The floorplan from RTL can be generated either automatically or incrementally by placing macros and assigning pins according to the constraints such as die size, macros, pin locations etc. provided by the user or supplied in a DEF (Design Exchange Format) file.

The designer can perform fly line and connectivity analysis between regions before placing physical partitions. The regions can be shaped and sized automatically depending on the die size and utilization constraints. All advanced floorplanning features such as fences, blockage, physical guidance, and rectilinear boundaries for space optimization are enabled.

To implement an efficient and testable design, scan insertion can be done during synthesis which enables early debugging of test problems and allows better scan architecture, reduced length of scan nets, and improved routability. Mentor’s tools are very efficient; a design with 10 million instances that might contain 1 million flops can be processed for scan analysis in 10 minutes and scan insertion in 20 minutes.

The advanced physical modeling techniques used during RTL synthesis provides very close correlation to P&R system; timing correlation comes within 5% of Mentor’s P&R system. This provides confidence in synthesis and floorplanning QoR and naturally minimizes the back-end to front-end iterations. Some of the actual experimented results are can be found in a whitepaper at Mentor’s website.


Mentor’s physical RTL synthesis and floorplanning tools are provided with a powerful integrated cockpit where cross-probing can be done easily between RTL and physical databases associated with different design views (logical, physical, or timing). The static and dynamic power map, congestion map, timing map, and hierarchical floorplan view can be debugged easily with different physical views. Any required changes can be done and synthesis re-run quickly to optimize the design.

Mentor’s physical RTL synthesis and floorplanning tools come with high capacity and speed that can enable multiple parallel floorplan explorations with different recipes for design alternatives. The best configuration for implementation can be decided after analyzing results from all runs. The floorplan exploration report contains data such as TNS, WNS, instance-count, area, power, congestion, and wire-length.

The tools and full-chip flat run methodologies as described above in Mentor’s RTL synthesis and floorplanning system provide up to 10x higher performance compared to traditional physical synthesis tools, and also equal or better QoR (area, timing, placement, and production quality floorplan). Designs up to 100 million gates can be easily accommodated.

Arvind Narayanan has described the tools and methodology in detail with some impressive experimental results in a whitepaper at Mentor’s website HERE.

Pawan Kumar Fangaria
Founder & President at www.fangarias.com


Internet of Things 2015 Year End Review (3): IoT Opportunities and Risks Insights from Patents

Internet of Things 2015 Year End Review (3): IoT Opportunities and Risks Insights from Patents
by Alex G. Lee on 01-18-2016 at 12:00 pm

New IoT Product/Service Development
Even though the IoT is getting a huge attention recently the concept of interconnected billions of devices is not new and has been under development for over 10 years. Thus, there are a large number of related patented technologies that can be exploited for developing new products/services, and thus, new business for the emerging IoT market.

The basic principle in the Blue Ocean Patent Strategy is to exploit existing patents to achieve the value innovation, and thus, to serve the customers in fundamentally different ways. For example, a company in the consumer electronics industry that wants to enter the connected car industry can exploit existing patents that cover the factors of the strategy canvas. By deciding which factors (that are covered by the existing patents) are really crucial, and thus, needed to raise and/or create, a new business model that changes the nature of competition away from the typical direction of the medical device industry can be developed. Patents regarding superior UI/UX, compact/portable design, robust wireless connectivity are the good candidates for the BLUE OCEAN FACTORS.

Another way to achieve the value innovation is to integrate into the existing platforms. HBS Professor Iansiti, the author of “The Keystone Advantage,” suggested technological assimilation as a new engine for technological innovations in his article “Creative Construction.” In the technological assimilation frameworks, a core innovation that once provided stand-alone products or services for a specific market can be the building blocks for mass market generating innovations through assimilation to broader platforms that were not existed at the time of innovations. A good example may be the GPS technology providing LBS (location based services) applications for smartphones and automobiles. Another good example may be the DVR technology of TiVo that was integrated into digital set-top box platforms. By contrasting all possible value propositions from an existing platform to the potential value propositions provided by the inventive departures (novelty points) of the existing patent(s), the integration feasibility of the existing patented technologies into the platform can be found, and thus, the new integrated value propositions can be developed.

Patents also can be exploited to identify new IoT (Internet of Things) product/service development opportunities by the scenarios analysis. The scenarios analysis can show potential interactions between the future users and the new IoT product/service via the specific usage of the product/service and behavior of the user under environments provided by the product/service functionality/offering. Thus, the scenarios analysis exploiting patent information can provide the new IoT product/service concept (e.g., specific benefits to the user, product design, product functionality and the technology for the product).

Investment for New IoT Innovations/Startups/M&A/Joint Venture

Patent information can provide insights regarding the state of the art of the IoT innovations. Thus, one can identify the potential further innovation R&D areas (“white space”) that can lead to new products/services development through the patent analysis. For example, in-depth analysis of smart home patents reviles that most of the patent applications are for the incremental innovations of the current market products/services. 13% of the total patent applications are for the innovations of the potential new market products/services: Adaptable Autonomous Smart Home System (The adaptable autonomous smart home system can recognize the contextual or semantic profiling of a person or place or devices (physical environment) based on sensed data by the IoT devices. The adaptable autonomous smart home system determines particular interpretation instructions (define particular IoT device control rules) that are associated with the particular physical environment and dynamically updates the control rules for changing physical environment); Self-Aware Self-Healing Smart Home System; Artificial Intelligence including Deep Learning Applications; Cloud/Big Data/SaaS; Robots; New UX for Smart Home System; Cross-industry Convergence; and Emotion-Aware Smart Home System. Patent information also can provide insights regarding a new statups to become play a leading role in the emerging IoT market (and thus, can be a good investment/M&A target) and joint venture opportunities.

New Patent Monetization Model

A strategically packaged patent portfolio is the collection of patents that the integrated value propositions of each patent of the portfolio target specific value propositions that are provided by emerging new products/services. The strategically packaged patent portfolio in alignment with the specific IoT business interests (e.g., smart home automation) can be developed using the existing patents and/or by strategically developing new patents. Development of the strategically packaged patent portfolio requires a deep understating of the IoT technologies, extensive experiences in patent analysis and development, and insights into the emerging IoT market. The strategically packaged patent portfolio can be used for monetization through patent sale, patent licensing, commercialization, spin-off, patent banking, and patent-backed financing. For detailed information, please visit http://www.slideshare.net/alexglee/best-practices-of-ip-patent-strategy-iot-internet-of-things-case-study.

Potential Patent Disputes

As we have seen in the smartphone market development, it is expected that the super-competition to preoccupy the leadership in the lucrative IoT market can lead to another round of patent wars. The post-smartphone patent wars, however, will be more extensive because of more extensive participation of players across several different industries. The post-smartphone patent wars will also be more complex because of the recent rapid change in legal environment and the learning curve from the smartphone patent wars.

To assess the potential patent disputes risk of the IoT, patent landscapes of three major IoT applications – smart home, connected car, connected health (including healthcare/fitness wearable devices) – and IoT platform connectivity are researched. Followings summarize the potential patent disputes risk for each subfield of the IoT.

Smart Home
Interesting point in the smart home patent land scape is a large number of patents owned by two startups – Allure Energy (9 %) and Ecofactor (5 %) – that are not successful in the IoT smart home market compare to their marker leader competitors such as Nest Labs and Ecobee. A large number of patents owned by commercially unsuccessful companies can be the potential patent disputes risk to the smart home companies.

Connected Car
Interesting point of the connected car patent landscape is the large number of patents owned by two active patent monetizing entities – Omega Patents and American Vehicular Sciences and a suspected entity – AutoConnect Holdings. The large number of patents owned by the patent monetizing entities is a potential patent disputes risk for the connected car industry.

Connected Health

Interesting point of the connected health patent landscape is the large number of patents owned by the active patent monetizing entitiy – Empire IP LLC, which is a potential patent disputes risk for the connected health companies. Interesting point of the personal fitness/health care devices patent landscape is the large number of patents owned by many commercially unsuccessful companies and NPEs. Another interesting point is the small number of patents owned by commercially successful companies (e.g., Under Armour).

IoT Platform Connectivity

Interesting point of the IoT platform connectivity is the large number of patents owned by many active patent monetizing entities – InterDigital, Optis Cellular Technology, Intellectual Ventures, Innovatio IP Ventures, Adaptix (Acacia), Evolved Wireless LLC, Wi-Fi One and WiLAN. The large number of patents owned by many patent monetizing entities is a potential patent disputes risk.

One can develop a strategic forecasting methodology to predict the possible emerging development of the post-smartphone patent wars and a strategic planning methodology to prepare for the emerging post-smartphone patent wars.

For detailed information, please visit http://www.slideshare.net/alexglee/postsmartphone-wearables-iot-devices-patent-wars-strategic-forecasting-methodology and http://www.slideshare.net/alexglee/postsmartphone-wearables-iot-devices-patent-wars-strategy-development.


More articles from Alex…


Synopsys on the Future of Custom Layout!

Synopsys on the Future of Custom Layout!
by Daniel Nenni on 01-18-2016 at 7:00 am

Analog and mixed signal design has received more than their fair share of attention since the mobile revolution and now that FinFETs are in production at the foundries I see that trend continuing. As a result this year there are some interesting things brewing in EDA, especially in the area of Custom Layout.

Innovation in Custom Layout has been nowhere near as rapid as in other areas of EDA. Over the years the custom tools have been tweaked and tuned with a new feature here and there to enable the layout engineers to be as productive as possible while the industry sailed along with the tide of shrinking process nodes. But the waters are getting choppy. Driven by the new challenges that the latest process nodes are bringing, especially FinFET and multi-patterning, it appears that it’s time for Custom Layout to undergo a much needed disruption.

That’s clearly the thinking over at Synopsys where Graham Etchells believes it’s time for Custom Layout to get into the 21st century. Graham is an EDA veteran with over 38 years in ‘the business’ and his involvement with custom layout stretches back to the late 1970’s with the introduction of the Calma GDS 1 systems. He has some interesting insights as to what’s really needed to address this new wave of challenges and has taken to blogging to crowdsource new ideas and approaches.

Graham’s new blog “Custom Layout Insights” starts with a three part series called “We have come a long way” which is a very good read. It has always been my conviction that you must know how you got to where you are today to better decide where to go tomorrow. He starts this series in the late 1970’s with the first Calma GDS systems which was before my time. I didn’t arrive in Silicon Valley until the early 1980’s where the Calma GDS-II system was the defacto IC design standard. In fact, that is where term GDSII came from. That is also where the term tape-out came from since we used to stream the GDSII out to magnetic tapes when the design was finished.

Last week Graham started another series “Hurricane FinFET – part 1.” which is a nice introduction to FinFET custom layout. Hopefully this is a 100 part series because there is a lot of ground to cover. And who better to cover it than the top EDA/IP company with leading edge FinFET IP in production all over the world? And yes all Synopsys IP designers use Synopsys tools including their more than 400 layout people. I know this by experience from the Virage acquisition. All of our engineers were told that they had to switch from Cadence to Synopsys tools in 30 days without schedule delays. Imagine the horror! Not only did the Virage people do it, they love using their own tools and the custom development and support that come with it, absolutely.

About Graham Etchells
I started in EDA before it was termed EDA. It was simply Computer-Aided Design back in 1977. I was working at GEC Traction in Manchester England (yes, I am a Brit) doing control gear for locomotives. It was all heavy duty relays and contactors back in those days. Then came the electronics revolution and with it came the first CAD system. It was a CALMA GDS1 system with green vector refresh displays and huge digitizers for entering the data. It was amazing what you could do with a Data General Eclipse computer and 16K (yes, Kilobytes) of core memory! Pretty soon I was running the CAD system, which at the time was one of the largest in Europe, if not the world. CALMA was expanding and I was recruited as an applications engineer. That was it; I was in EDA and have been ever since. I have held marketing and sales positions at Silvar Lisco and Neolinear and I have been chasing the holy grail of analog/custom layout automation ever since I ran marketing for Virtuoso at Cadence back in 1995. Past experience tells me we may never find the Holy Grail, but there is light at the end of the tunnel. Follow this blog and see how we at Synopsys are progressing.

Also Read: Synopsys Vision on Custom Automation with FinFET


IBM’s OpenPOWER Presence Was Felt Heavily At SuperComputing ’15

IBM’s OpenPOWER Presence Was Felt Heavily At SuperComputing ’15
by Patrick Moorhead on 01-17-2016 at 10:00 pm

IBM is in the process of reinventing themselves as a company, changing how they see themselves, what they do as a company and how they want their partners and customers to view them. This is exemplified best in their mobile alliance with Apple, their Watson cognitive efforts, the sale of their chip fab to GlobalFoundries, the sale of their X86 server and networking division to Lenovo, and the creation of the OpenPOWER Foundation.


IBM’s Brad McCredie, also OpenPOWER President, kicks off analyst meeting at SC15 (Photo credit: Patrick Moorhead)

In the past two years since OpenPOWER’s creation, IBM’s mindshare in big HPC (high performance computing) designs and partnerships across the industry has increased. IBM’s partnerships created through the OpenPOWER Foundation allowed the company to work with companies they never had a chance to in the past, while increasing the overall relevance of IBM’s POWER architecture on a broader, global scale. The evolution of the POWER architecture and IBM’s vision for the future have been heavily shaped by the partnerships that they have created in the OpenPOWER Foundation and now we are starting to see a nice glimpse of exactly what these partnerships have returned.

The creation of the OpenPOWER Foundation was built upon the thesis that the industry simply wasn’t providing enough performance with CPUs alone (including POWER and X86) to fill the compute need and that more competition and “accelerators” were required to shore the gap. The gaps that once existed in IBM’s HPC compute portfolio are starting to get filled with OpenPOWER ingredients. IBM is also enabling more complete HPC solutions to their customers to expand the scope of where HPC can be applied today and into the future. That includes cognitive computing, network data forensics and facial recognition, among the existing HPC applications like financial simulations, genomic analysis, oil and gas imaging and scientific computing. As you would expect, the new challenges and opportunities in HPC have led IBM and their partners in the OpenPOWER Foundation to look to GPUs, FPGAs and fixed function controllers as accelerators for these new, more complex, workloads.

Last week at Supercomputing 2015 (SC15), the world’s premiere supercomputing trade show, IBM’s major announcements were around the acceleration technologies that utilize the OpenPOWER partnerships which allow for even more performance out of POWER-based HPC platforms. These accelerators unsurprisingly come from some of the biggest and earliest partners in the OpenPOWER Foundation, namely NVIDIA, Xilinx and Mellanox Technologies. I got the chance to meet with all three of these companies along with analysts Gina Longoria and John Fruehe and technologist in-residence, Jimmy Pike. We had some very interesting conversations to say the least.

IBM highlighted their acceleration advancements with NVIDIA through NVIDIA’s GPUs and Tesla as well as their upcoming NVLINK interconnect which will be integrated into future POWER processors next year and beyond. However, IBM today has their own CAPI (Coherent Accelerator Processor Interface) which they use with all of their other accelerators in the OpenPOWER Foundation. This interface is great for many reasons, namely that it’s coherent and supports a very large ecosystem of players, far beyond what NVLINK can do and is really what enables OpenPOWER to work at its very core. NVLINK is faster than CAPI, but lacks the coherency of CAPI, which makes it less useful for broader applications and industry adoption.

IBM also announced a strategic collaboration with Xilinx which was squarely focused around adding acceleration to datacenter applications through FPGA-enabled workload acceleration in conjunction with IBM POWER-based systems. As a part of the collaboration, IBM will work with Xilinx to utilize and integrate CAPI to create solutions that improve the overall performance in the data center with a specific focus on OpenStack, Docker and Spark software-driven data center architectures. This announcement also illustrates IBM’s belief that Moore’s Law is no longer meeting the needs of the HPC market and that accelerators and software are going to be needed to fill those gaps. Moore’s Law has moved dates twice, and recently to two and a half years, but I’m not convinced yet it can’t get back to every two years.
While we’ve heard the ‘end of Moore’s Law’ many times, GPUs and other ASICs along with FPGAs have always shown to improve certain types of workload performance and their programmability has only improved, making them better multi-purpose compute platforms. Nevertheless, it is great to see that IBM is pushing the envelope of performance with partners like NVIDIA and Xilinx to deliver the fastest possible solutions to their customers as well as broadening the relevance of the OpenPOWER Foundation. Most of all, I like CAPI’s open, fast, coherent nature.

Hardware without software is useless and IBM went out of their way to show real HPC workloads and applications accelerated by CAPI. OpenPOWER’s HPC offerings continue to grow with solutions that include IBM Watson, Spark, Edico Genome, gpudb, neo4j and Radar. All of these solutions utilize OpenPOWER and IBM’s accelerated platform to deliver accelerated applications in a fast and efficient manner.


OpenPOWER accelerated applications on-display at SC15 (Photo credit: Patrick Moorhead)

IBM has clearly shown that OpenPOWER is starting to flourish and is beginning to pose a pretty serious alternative threat to Intel’s high-end HPC offerings, including their Xeon and Xeon Phi family of processors. Intel has embraced accelerated computing with transcoding solutions, Intel Xeon Phi and their $16.7B Altera acquisition and are keen on filling sockets with their own silicon. They will partner to fill the rest. IBM and their OpenPOWER partners have already won some very big HPC deals and it wouldn’t be much of a surprise to see that momentum to continue forward and to bleed into smaller HPC deployments across different industries.

More from Moor Insights and Strategy


Maybe not the world, but schedules got eaten

Maybe not the world, but schedules got eaten
by Don Dingee on 01-17-2016 at 4:00 pm

It has been almost five years since Marc Andreessen wrote the words, “Software is eating the world.” The premise of his essay in the Wall Street Journal in 2011 was pretty simple: the technology world has seen its intrinsic value shift from hardware to software. New all-software names have appeared on the list of high flying companies, and hardware firms have been forced to transform into a more software-centric mix Continue reading “Maybe not the world, but schedules got eaten”


Intel to Focus on IoT and NOT Mobile?

Intel to Focus on IoT and NOT Mobile?
by Daniel Nenni on 01-17-2016 at 12:00 pm

The Intel Q4 investor call was last week and as Brian Krzanich approaches his 3[SUP]rd[/SUP] year as Intel CEO a new synergistic corporate strategy is emerging:

That strategy is also resulting in the evolution of our business model to focus on three key areas of growth: The Data Center, the Internet of Things, and Memory.

DCG, IoTG and Memory, delivered nearly 40% of Intel’s revenue and more than 60% of Intel’s operating margin in 2015. Additionally, these three adjacent markets delivered $2.2 billion in profitable revenue growth in 2015 alone. As we look ahead to 2016, we will continue to build on that strategy.

You should notice that Mobile, Foundry, and the FPGA businesses are not included here.

Certainly after spending BILLIONS of dollars on mobile Intel is not going to announce that they made a mistake and are leaving the business, but they did, they most certainly are, and it is a great move by Brian. If you look at the top three smartphone providers (Apple, Samsung, Huawei) all are now industry leading semiconductor companies with SoCs as good as or better than what Intel was able to do with Atom. That leaves fabless giants Qualcomm, Mediatek, and a handful of smaller players to fight for the remaining low margin-low growth merchant SoC business.

Bottom line: Intel Mobile will die a slow and silent death inside the Intel Client Computing Group.

Once Intel mobile is dead can Intel Custom Foundry take another shot at the merchant SoC providers? Maybe, but the “Apple using Intel” rumor that is flying around again is not going to happen this year or next. The foundries are now building and ramping their processes for SoC design and I do not see Intel doing that ever. Apple was the driving factor for TSMC 20nm, Samsung 14nm, TSMC 16FFC, TSMC 10nm, and Qualcom is behind Samsung 10nm.

As much as I would like to have another leading edge foundry in the mix I do not see Intel continuing down this bumpy road. Remember, this will be the third time Intel has gone in/out of the foundry business and this time it was started before BK became CEO so he does not own it. As I have said before, Intel should acquire fabless companies that are in emerging markets to better diversify and do what Intel does best, make chips! And this is what BK has done with Altera and will continue to do in IoT, my opinion.

Speaking of Alltera, I have had many conversations with FPGA professionals and the consensus is that Intel will focus on integrating Altera FPGAs into existing Intel business units and not aggressively pursue the remaining mainstream FPGA market segments. If so, some serious cuts will be coming to Altera, absolutely.

One final note, Intel 10nm was also not mentioned in the prepared statement which to me means there are more delays. In regards to process technologies (BEOL), TSMC 10nm can be considered a half node between Intel 14nm and 10nm. Based on conference papers TSMC 7nm BEOL will have a slight process advantage over Intel 7nm.

The TSMC 10nm PDK 1.0 is out now meaning that if all goes well TSMC 10nm wafers will start production in Q4 2016 and be ready for the iPhone 7s refresh in 2017. TSMC 7nm will follow one year later and hit the iPhone 8 in 2018 (TSMC 10nm and 7nm will use the same fabs/equipment so 7nm will ramp much faster than 10nm).

According to what I heard in the hallway last week at the SEMI Industry Strategy Symposium, Intel 10nm will be in full production in early 2018 and 7nm 2-3 years after that. The 2-3 years will depend on EUV which, according to a recent ASML presentation, is closer to 3 than 2 years.

Bottom line: At 10nm TSMC and Samsung will officially take the process lead from Intel and it will continue through 7nm, my opinion.

More articles from Daniel Nenni


Pure-play Foundries to Prevail in Future

Pure-play Foundries to Prevail in Future
by Pawan Fangaria on 01-17-2016 at 7:00 am

In a consolidating semiconductor business environment and innovation in semiconductor fabrication already scaling new heights with existing strong players, where do you think the wafer capacity should concentrate? It’s pure-play foundries or pure-play-like foundries, and those who supply high-volume common components such as memories. By pure-play-like foundries I mean IDMs which are fairly open to manufacture for other fabless companies; Samsung is at the top, very recently they won a deal with Qualcomm to manufacture their Snapdragon 820 chips.

When we talk about memory suppliers, like pure-play foundries they also manufacture memories for the whole world, from B2B up to B2C segments. So that’s natural for them to have large wafer fab capacities for memories.

A chart from IC Insights shows the world wafer fab capacity heavily concentrated, 72% of total with top 10 semiconductor manufacturers.


Samsungbeing at the top, the second wafer producer is the largest pure-play foundry TSMC. From the chart, it’s impressive to see GlobalFoundries having largest yr-yr increase in wafer capacity at 18%; TSMC’s yr-yr increase is next at 14%. Also, GlobalFoundries climbed at rank higher in 2015 compared to 2014.

The top memory suppliers, Samsung, Micron, Toshiba, and SK Hynix occupy highest ranks within top 5. So, that explains very well the memory suppliers’ dominance in wafer fab capacity.

Micron has accumulated substantial memory capacity through acquisitions of Elpida and Rexchip, and also from extra Inotera capacity and stake in IM Flash Technologies fabs. Similarly, Toshiba has substantial addition of flash memory capacity from SanDisk. So, we can see fair amount of consolidation in this space.

If we see overall percentage share region wise, South Korea with Samsung and SK Hynix dominates with 23.6% of total worldwide wafer capacity which is at 16350 thousand wafers per month. North America is next with four companies (Micron, GF, Intel, and TI) there at 22.3%. Taiwan’s TSMC and UMC have 15% share.

Going forward we are going to see more consolidation in fab business; setting up new fab is becoming unbearable by any new or existing weak players. Also we may see IDMs opening up their running fabs for manufacturing chips for other fabless companies, but that depends on how viable their fabs are in terms of technology and cost, and market requirement; IoT and automotive segments can offer varied requirements. At the same time some IDMs may go fab-lite and get their chips manufactured by pure-play foundries. We have to see.

IC Insights report can be found here.

Pawan Kumar Fangaria
Founder & President at www.fangarias.com


ARM on Moore’s Law at 50: Are we planning for retirement?

ARM on Moore’s Law at 50: Are we planning for retirement?
by Scotten Jones on 01-16-2016 at 7:00 am

On Monday morning on December 7, 2016 Greg Yeric of Arm gave an excellent and wide ranging plenary talk at IEDM entitled “Moore’s Law at 50: Are we planning for retirement?”. You can download Greg’s slide deck here.

Continue reading “ARM on Moore’s Law at 50: Are we planning for retirement?”


How Artificial Intelligence Will Kickstart the Internet of Things

How Artificial Intelligence Will Kickstart the Internet of Things
by Ahmed Banafa on 01-15-2016 at 4:00 pm

The possibilities that IoT brings to the table are endless. IoT continues its run as one of the most popular technology buzzwords of the year, and now the new phase of IoT is pushing everyone to ask hard questions about the data collected by all devices and sensors of IoT.
Continue reading “How Artificial Intelligence Will Kickstart the Internet of Things”


AMD’s GPUOpen Initiative Architected To Move The Industry In A Different Direction

AMD’s GPUOpen Initiative Architected To Move The Industry In A Different Direction
by Patrick Moorhead on 01-15-2016 at 12:00 pm

GPUOpen is a new initiative started by Advanced Micro Devices’ Radeon Technologies Group (RTG) as a way for the company to continue to reach out to the needs to developers by giving them a better development environment. The idea behind GPUOpen is born out of the work that Advanced Micro Device’s did with Mantle, their own low-level API which eventually found itself reworked into the industry standard Vulkan API. Thanks to Mantle we also have other low-level APIs that allow developers to get more access to hardware and better performance and experience. GPU Open is the next step in the evolution of AMD’s mission to improve the PC game development environment while also making it more and more open.

The GPUOpen initiative is comprised of a few major components- direct access, open source software and industry standards. The direct access portion of GPUOpen means that Advanced Micro Devices’ RTG will give developers even more access to the functions of the GPU than they have in the past, giving them even more control of the GPU and more down to the metal hardware access than before.

The open source software portion of GPUOpen includes AMD’s own movement of their development tools into open source, enabling developers to see all of the code and improve upon it as they see necessary. In addition to open sourcing their tools, AMD is also putting a renewed focus on their open source Linux drivers, increasing the amount of supported Linux distributions as well as accelerate access to new products and features. However, not all features will be open source as some of them will either remain closed source or start off as closed source and then become open source, like Open CL and Vulkan. In addition to all of the open source tools, there’s also a special set of features, known as the Boltzmann Initiative that AMD has created to allow CUDA apps to be run on AMD GPUs. Another part of the open source software is the implementation of GPUOpen Compute which is AMD RTG’s move to create an open source end-to-end compute infrastructure for usage models that exist today and into the future.

Another aspect of GPUOpen is AMD’s willingness to include others in their standard to try to help make it part of the industry standard. This move is similar to what AMD has done in the past with their technologies like Mantle and FreeSync. While I’m unaware that Intel and NVIDIA have taken up AMD on their offers to make Mantle or FreeSync industry standards, in one way or another AMD’s willingness to be open has turned some form of their technology into an industry standard. With FreeSync, we are seeing companies like Intel stating that they are looking to implement VESA’s adaptive-sync technology which is the fundamental underpinnings of what makes FreeSync possible. Additionally, as mentioned earlier, Mantle eventually became the foundation for Vulkan which is the newest low level industry standard API which already appears to have the support of many hardware vendors and developers. While it isn’t much of a stretch to assume that some portions of GPUOpen could become industry standards, it’s very unlikely we’ll see NVIDIA jumping at the opportunity to participate and mirror AMD’s GPUOpen activities.

Ultimately, what GPUOpen needs to do is make AMD’s GPUs more attractive to both developers and users. AMD believes that opening up their GPUs even more will make them more attractive to developers and hopefully improve performance in games as a side effect, making their GPUs more attractive to consumers and selling more GPUs. AMD managed to squeeze quite a bit of extra gaming performance from the jump to DX12 in Microsoft’s Windows 10 from previous versions and it may be that we could see something similar happen in Linux or in professional environments where AMD needs to regain share the most.


More from Moor Insights and Strategy