Electronics Engineering Herald                 
Home | News | New Products | India Specific | Design Guide | Sourcing database | Student Section | About us | Contact us | What's New


  Date: 02/06/2015

Intel did not let Altera go; the deal is done

Intel being the world's largest semiconductor company and also the world's largest processor chip maker with cutting edge chip fabrication technology tightens (rather we say 'saves itself from loosening') its grip on the future high performance SOC market by agreeing to acquiring Altera at a price of US$16.7 billion. How will this affect the future of no.1 FPGA maker Xilinx? and also the second-biggest PC chip vendor AMD? By some chance if AMD and Xilinx merge that gives a befitting competition to the combination of Intel and Altera. Such a merger is hard to come through, but still Xilinx may have even more strongly branded position in FPGA market, just like Analog Devices able to strengthen its position in analog market even after Texas Instrument acquired National Semiconductor.

When this writer asked Neeraj Varma, director of sales at Xilinx on this acquisition, He said "Xilinx continues to distance itself from the competition with a 3Peat across 28, 20 and 16nm products, ASIC-class tools and software-defined environments… and is building another multi-node scaling advantage by moving directly to 7nm in collaboration with TSMC."
 "Intel’s acquisition of Altera will also give Xilinx the additional significant advantage of being the ONLY vendor with total focus, the best foundry, ARM strategic partnership, committed roadmap, and continuity of operations." He added.

With some interesting mergers and acquisitions happening in 2015 such as Avago acquiring Broadcom and NXP Semiconductors acquiring Freescale Semiconductor, we got to see any more interesting mergers will happen in the year 2015 in semiconductor sector.

Due to some inherent advantages at deeper nodes and also due to the demanding processing for handling multimedia data such as high-definition video and audio and also rich graphics, the role of customised programmable silicon such as field programmable gate array getting embedded inside a SOC chip makes lot of differences to the performance of the processor/SoC chips. FPGA chips are already extensively used for processing video and graphics, But they are more used as stand-alone chips.

Intel gains significantly through this acquisition if not by immediate revenue but on a long-term strategic basis, to give you an example of how a bit of FPGA can enhance the performance of x86 CPU:

In a typical CPU, to process rich-media data or such complex data requires lot of instructions and clock cycles, which causes delay in execution. If the same function can be executed in single instruction using a customised hardware it saves power, time and also generates less heat, even the chip size can be smaller. But the problem with having customised hardware is if you wish to change the instruction specific hardware for some reason, you got to redesign the chip and fabricate fresh chips. Whereas in case of FPGA, the logic can be rewired with a code with in the same chip, so that a new instruction and also related hardware can be added in future without changing the chip. The processors such as x86 can be re-architected with the bits and pieces of FPGA to handle specific instructions and it can bring lot of benefits in real-time processing and also complex function processing such as voice recognition and image recognition algorithms. Enhancing such custom hardware dependent instructions is easier due to FPGA's ability to change the logic whenever is required. FPGA can be used to build a small logic function or a complete custom processor. While FPGA maker such as Xilinx throwing hard-wired ASIC like functional blocks around its larger FPGA fabric to build what it calls programmable SOC, Intel can do the opposite by spraying small to big FPGA blocks on its latest processors for PC and servers.

Intel can leverage programmable silicon technology in designing processor chips as power-efficient or even more power-efficient than the most of the present ARM processor cores. This write-up one aspect of benefit, they are lot more ways Intel can leverage Altera's resources. One thing is sure, Intel is gaining competitive advantage over other SoC silicon vendors more than with the Xilinx.

Here's a reproduction of our previous article on the same subject when there was news of Intel proposing its acquisition plan to Altera:

There are quite a number of news stories reporting the rumour of Intel acquiring field programmable gate array (FPGA) chip maker Altera. Let's look at what makes FPGA such an attractive market. Worldwide there are only six important companies in this market of FPGA chips. They are Xilinx, Altera, Lattice Semiconductor, and Microsemi, and the two new startups Tabula and Achronix. Though QuickLogic makes programmable silicon based chips, they are different compared what is so-called FPGA. But the business model used by QuickLogic helps to understand how programmable silicon can be used for multiple applications. FPGA/programmable silicon helps electronics engineers to redesign the physical-architecture and hardware of digital circuits including processor architecture, so that any change in requirements can be easily implemented without changing the chip or board. Then why not everyone use FPGA? Well, FPGAs are expensive and also consumes more power compared to standard logic chip. But at 14 nm, the power consumption issue of FPGA is not that significant because they do not consume that much high level of power for many number of applications. Depending on the volumes, FPGA cost can be bought down close to many standard logic chips which are also called ASICs. So at deeper nodes such as 16nm/14nm, FPGA is lot more attractive in areas, which earlier never thought off.

What is already proven in the market is, Xilinx and Altera and Microsemi have launched SOC class chips having both FPGA and standard hard logic blocks (ARM Cortex A series processor or a ARM Cortex M series as hard block) and also some analog/interface in one monolithic chip. They are called as "programmable SOC". One of the successful product is Xilinx's Zync SoC IC, which has created lot of new markets.

Another interesting trend to share is, there is continuous change of standards and protocols in multiple number of applications, more so in mobile communications. Such change in standards and protocols demand redesign of electronics hardware with new chips. But with the FPGA, same boards and chips can be reprogrammed to adopt to new standards and protocols. With all-CMOS RF integrated SoC chips under development, having FPGA blocks on SoC make sense for both re-configurable radios and processors. Silicon can change its internals logic and other blocks' physical architecture like chameleon changes its color.

Nearly every ARM Cortex processor and many other popular processor IP architectures can run on FPGA fabric. So with a large silicon real estate of FPGA made using 16nm/14nm/10nm, engineers can design supercomputer to servers running on multiple number of FPGAs. This gives rise to importance of FPGA in server/data center market. And also due to low power consumption of 16/14 nm FPGA chip, they can be used even in personal computers. so that the PC users need not bother about changing hardware when they upgrade or move to new operating system. PC processors can be reprogrammed for any new on-line security threats at silicon level rather than using anti-virus software. FPGA-based chips are also can be used in sensor signal processing in smart phones/mobile devices.

So finally there are high chances of having a block of FPGA in every silicon, to make it more flexible, re-configurable and customisable.

To be more specific, the SOC market is becoming programmable SOC market. Leading semiconductor vendors who want to be in the programmable SOC market, they should have FPGA capabilities. You can't built FPGA business capabilities overnight, because they're heavily IP driven, a lot of patents are already given to leading semiconductor vendors such as Xilinx and Altera. So the best way for leading semiconductor vendors to make the chips more hardware-programmable is to acquire a FPGA company. The two other interesting FPGA vendors in the market other than Xilinx and Altera are Tabula and Achronix. Intel is already offering foundry services to Altera and Achronix. Intel has understood the FPGA market deeper by closely working with Achronix and Altera.

Xilinx is a big fish in the semiconductor sea and has revenues supporting capability to acquire many smaller non-FPGA companies. Its hard to see any bigger fish than Xilinx eyeing to acquire Xilinx in semiconductor industry. From the larger OEM electronics industry, there might be some companies who might acquire Xilinx. There are good number of electronics companies such Apple and Cisco which do significant semiconductor development for their internal use.

If rumours have to be believed, Intel may have to pay anything excess of $10 billion to acquire Altera , making it one of the biggest acquisition by Intel.

Intel stock prices are already rising. We will update you in our next article on this interesting development.
Author: Srinivasa Reddy N
Header ad Author: Srinivasa Reddy N
Header ad

Home | News | New Products | India Specific | Design Guide | Sourcing database | Student Section | About us | Contact us | What's New
©2012 Electronics Engineering Herald