WP_Term Object
    [term_id] => 13
    [name] => ARM
    [slug] => arm
    [term_group] => 0
    [term_taxonomy_id] => 13
    [taxonomy] => category
    [description] => 
    [parent] => 178
    [count] => 363
    [filter] => raw
    [cat_ID] => 13
    [category_count] => 363
    [category_description] => 
    [cat_name] => ARM
    [category_nicename] => arm
    [category_parent] => 178

From ARM7 to such a Large CPU cores Port-Folio

From ARM7 to such a Large CPU cores Port-Folio
by Eric Esteve on 07-08-2014 at 3:13 am

I have heard about ARM processor for the very first time in 1990, when I was interviewed by ES2 Design Center manager before being hired to subcontract an ASIC design for ES2. I don’t know why, but I remember very well that he told me about two of the ES2 partners: ARM as a processor IP core provider and TSMC as a Foundry partner if, by chance, one of the ASIC socket had to run into high production volume (at that time, ES2 was known to support fast prototyping, thanks to an ebeam based processing flow; good for prototypes, not really suited for large production volume). Then, in 1995, ES2 was bought by Atmel and the CEO Georges Perlegos decided to build a brand new fab supporting 350nm to 180nm in Rousset, France, in 1999. I joined Atmel at that time as (Standard Cell) ASIC PMM, reporting to the same person. Our job was not only to win ASIC design in front of the traditional competitors, but also to face not less than three internal competitors: Gate Array (US based group), MHS ASIC (France) and TEMIC ASIC (Germany). Thus, we had to severely brainstorm to highlight our differentiators! The most important was clearly our ability to support ASIC integrating ARM core, our group managing all the ARM-related developments in Rousset. This positioning was good, but the ASIC business decline was inevitable… That’s why my former boss is no more in charge of the ASIC business, but of ARM based Microcontroller family. Within Atmel, ARM based Microcontroller product line has grown from less than $10 million in 2003, to reach several $100 million in 2013…

Another short story, to provide another perspective: in 1998, working as an ASIC PME with TI, I have seen the explosion of the Wireless BU. The WBU success was clearly based on BaseBand IC, or the winning team made of TI DSP and ARM7TDMI IP core, evolving to the OMAP Application Processor of the 2000’s. The selection of the ARM7 core by the Ericsson, Nokia, Alcatel, Motorola (and many more) initially to support the GSM standard in the mid 1990’s, and later all the following standards, including CDMA, 2G, 3G, 4G, has defined ARM IP core as the “de facto” wireless Application Processor CPU. ARM position in such an exploding market has created a huge royalty flow, fueling ARM Ltd R&D and allowing the company to build and develop a complete IP core port-folio as we know it today, covering almost all potential applications, that we have called ubiquity in one of the very first (ARM vs Intel) blog on Semiwiki.

  • Embedded: Automotive Infotainment Embedded Computing, General Purpose MCU, IoT, Smart Card, Smart Meter.
  • Mobile: Computing, Smartphone, Feature Phone, Connectivity and Modem, Mobile Payment
  • Home: Blu-Ray and DVD, Digital Set Top Box, Digital Still Camera, Digital TV, Gaming
  • Enterprise: HDD/SSD, Flash cards and UFD, Home Networking, Network Infrastructure

This long list of application supported by ARM Processor core explains why ARM Holding (ARMH) stock has been x10 within 10 years!

If you take a look at ARM Cortex port-folio, you will face with a dilemma: which Cortex processor to choose, between Cortex A57 or A53, Cortex-A (7 cores), Cortex-R (R7, R5 or R4), Cortex-M (5 cores) or SecureCore families? A family like Cortex-A53 and Cortex-A57 seems to be dedicated to Smartphone. But if you take a look at the above picture (taken during ARM TechCon 2012), you can see a certain evolution, moving from Smartphone dedicated Application Processor counting four Cortex-A53 to Superphone/Tablet, integrating two Cortex-A57 cores on top the four A53. The next step according with ARM is “naturally” the Mobile Computer (four A53 + four A57). If you look at the last application, a MPU integrating 16 Cortex-A57 cores plus a Cache Coherent Network (CCN) and the Level 3 cache, the target is clearly the Server market, delivering today most of Intel margin in $, thanks to high chip ASP. As you can see, a family counting only two cores allows targeting multiple applications! If you consider that ARM CPU IP core portfolio is built around four more families, you will probably need some help to find the optimized solution…

Why not using the specific selector tool available on ARM web site?

We have mentioned the Wireless or Mobile market as the “cow” generating the huge cash flow, thanks to the high production level for ARM-based Application Processor and the related royalty payment. ARM Ltd. has been cleaver enough to get this cash… and even smarter in investing it, first into R&D and port-folio development, and also to build a very strong ecosystem with over 1000 partners delivering silicon, development tools and software. Such a strategy may appear as just common sense or good business practice today in 2014, but we have to remember that ARM has been a real innovator, and was the first company to really understand how important it will be for the future of their business to develop such an ecosystem. Companies like Intel, for example, could have done it 20 years in advance of ARM, but securing and reinforcing their MPU monopoly was their main concern, not building an ecosystem…

Eric Esteve from IPNEST –

More Articles by Eric Esteve…..


0 Replies to “From ARM7 to such a Large CPU cores Port-Folio”

You must register or log in to view/post comments.