Interview - Warren East, CEO of ARM Holdings plc : Flexing ARM for disruptive change
EE Times: Flexing ARM for disruptive change | |
Peter Clarke, Richard Wallace (08/29/2005 9:00 AM EDT) URL: http://www.eetimes.com/showArticle.jhtml?articleID=170100694 | |
Warren East left Texas Instruments Inc. to join core maker ARM Holdings plc in 1994, when the U.K. company was "small enough to worry about whether it could pay the salaries at the end of the quarter." Today, as chief executive officer and director of the market leader in microprocessor intellectual property, East leads a $255 million (2004 revenue) global provider of IP cores at the top end of the semiconductor value chain. EE Times caught up with East at ARM headquarters in Cambridge, England, to get his take on topics ranging from ARM's relationships with its customers to the changing landscape of hardware/software design and the future of the microprocessor.
EE Times: ARM seems to be at the nexus of a disruption in semiconductors. How does ARM view itself in today's world, and what disruptive changes is ARM enabling?
Warren East: Well, we're in the business of enabling the digital world, and we're a semiconductor IP [intellectual-property] company, not a pure semiconductor play. We observe that the digital world is really being fueled by more intelligence appearing in digital products. The picture that I've taken to painting lately is that ARM is at the intersection of the semiconductor IP world and microprocessor world, and it's a unique position.
From the semiconductor IP side, IP is emerging as a subsector of the semiconductor industry out of necessity. At the same time, increasing intelligence is being embedded into digital products, and that is bringing embedded computing to products like cameras and phones.
EET: What does ARM's IP model enable and enhance, from a business and technology perspective?
East: Because we're an IP company we can specialize in designing microprocessors, so essentially we take structural cost out of the industry. We're up to 148 semiconductor licensees. Instead of having all those semiconductor licensees spend millions of dollars on developing and maintaining a microprocessor architecture, they are able to access the open, standard ARM microprocessor architecture at a fraction of that cost, and we're repeating the exercise now with physical IP from [library IP vendor] Artisan [Components Inc., acquired in 2004].
EET: You're offering more of the system components and enabling software.
East: Increasingly, functionality is appearing in software rather than dedicated hardware, because that enables people to have flexibility. So you are going to see that come into ARM's solutions things like data engines and graphics processors. We use the term "compute engine," rather than microprocessor, because within this programmable world a general-purpose RISC microprocessor is not necessarily the ideal processor architecture for all the functions that you want to integrate. We find that people are implementing in hardware gates, alongside our general-purpose microprocessor, acceleration functions. Our Optimode [ARM's configurable VLIW data-engine architecture] provides a solution for people to build those accelerators and have the sort of functionality that they get out of a hardware accelerator, [with] the flexibility and programmability of a general-purpose microprocessor.
EET: You recently did a deal with graphics-IC vendor Nvidia on the ARM core 11 processor. Isn't this moving ARM into the Texas Instruments DSP camp?
East: We're interested in enabling people like TI to provide solutions to those people. We're also interested in enabling people like the person at the other end of the chain. They can either choose a TI solution, with a TI DSP, or they can choose an Nvidia solution with some Nvidia graphics magic. We're trying to make sure that whichever way they do it, there's some ARM technology there. From TI's point of view and from Nvidia's point of view, they need a microprocessor and, rather than develop that microprocessor themselves, they can buy the ARM standard microprocessor and concentrate on providing a really differentiated solution to the end customer.
EET: What about competitive pressures from the other direction? Mitsubishi phones, for instance. We understand Mitsubishi is looking at Renesas Technology as a potential supplier, for access to the DSH architecture. Do you see other solutions pressing in on some of the areas where you have dominance, such as mobile phones?
East: It's a real world out there. There's always competition, and of course, there's a nuts-and-bolts level. There's no way we can avoid competition we don't want to.
EET: Specifically with regard to the Japanese market or Mitsubishi?
East: In-house microprocessors . . . tend to be more prevalent in Japan. The U.S., particularly, has embraced the concept of a standard architecture more readily than the Japanese. It's a region which tends to be a bit more inward-focused and less inclined to adopt the open standard. Having said that, 20 percent of our royalties come from Japanese semiconductor partners, and the 14 Japanese semiconductor partners know they're contributing a substantial portion of ARM's royalty revenue. And they're only doing that because use of the ARM architecture is good for them.
EET: The Artisan acquisition raised eyebrows. It presents ARM with an opportunity to couple its customers with a foundry's via the Artisan libraries. But there's concern that ARM customers who don't use Artisan libraries will be forced to do so.
East: The raised eyebrows were substantially amongst the financial community, not really amongst the engineering community, which I think is well aware of the increasing complexity involved in physical design and the obvious opportunity for all concerned to have more outsourcing of this sort of technology more specialties within companies, like ARM, to develop it and, therefore, removal of structural cost.
As for forcing ARM customers to use Artisan physical IP, that's nonsense. . . . If you look in our tools business, then ARM's tool business probably serves 15 to 20 percent of the total ARM tools market. The important thing is to deliver to the end customers the best solution, so we enable a variety of routes.
There will be people building ARM base systems on Artisan libraries. There have been many ARM base systems built on in-house libraries and even [on libraries from] direct competitors of our Artisan business people like Virage Logic. But if people want to build ARM base systems with Virage physical IP, that's their choice. Our challenge is to make sure that the ARM physical IP is the best solution for them, and that's what we intend to do.
EET: You've acquired a design tool company, Access. Clearly, being able to integrate more of the design tool aspect of building a system seems to be important to ARM's business. Do the EDA vendors see this as a threat?
East: The Access acquisition is all about a different type of design tool; it's about system-level design and the sort of SoCs [systems-on-chip] that people were building based on ARMs. Eight to 10 years ago, you could do all the system design you wanted to on the back of an envelope or, worst case, an Excel spreadsheet. Now the systems are more complex, there's a lot more functionality in the software, the interaction between the hardware and the software is a more difficult problem to solve, it's much more expensive to make mistakes today than it was eight to 10 years ago and, consequently, people have to find a way of addressing this double-whammy program: the increasing complexity of the natural design and the increasing cost of making a mistake. That drives them to do more code design modeling of hardware and software interacting together than they were doing eight to 10 years ago, and it's that sort of problem that the Access tools are addressing.
EET: Apple has made a sudden shift to the Intel microprocessor and, to oversimplify, it basically leaves two major processor architectures dominating the world: Intel and ARM. Is the microprocessor, as we've known it, dead?
East: Gosh, well, from a competitive point of view we still see lots of architectures out there. So, I can't say the microprocessor is dead at all. Today most of the microcontrollers happen to be 8-bit processors, but we believe in the future there's going to be a lot of 32-bit microcontroller designs out there, and we hope that many of them will be ARM-based. It's very interesting that Apple has decided to change its processor, but as far as we're concerned it doesn't really change the future much.
EET: Any potential business for ARM and Apple going forward?
East: ARM is already involved in some of Apple's consumer products. The Apple iPod's a well-known example. We've always steered clear of the desktop, simply because there's more business to be had in the embedded space, which is where we're going. From a common-sense point of view, why waste technical and commercial effort attempting to design perfectly good processors out of desktop PCs when, if you look at a PC from ARM's perspective, there's an opportunity for an ARM in the wireless connection, in the Bluetooth connection, in the printer that's attached to the PC, in the hard-disk drive, in the digital camera that you connect to the PC?
EET: Intel dropped the ARM name and use Xscale as the trade name for its implementation of the ARM architecture, and it's added some new instructions. It seems Intel has wandered off on its own in terms of the license.
East: Intel's one of the very few semiconductor partners that have an architecture license, which explains why there's some additional instructions and that's what the architecture license enables people to do. It's just a different level of differentiation. All of our semiconductor partners market their ARM-based products under family names. Some of them have numerical names and others more descriptive names, like Xscale and [its predecessor, Intel's] StrongARM. That's something we don't particularly get involved in. It's up to the semiconductor partner.
EET: Is ARM concerned about the foundry model going forward, where increasing R&D costs threaten to contract the foundry world down to maybe one or two players?
East: Increasing complexity generally leads to some kind of consolidation, until there's some discontinuity and the whole cycle starts again. So in basic CMOS technology, we're going to see fewer companies owning the fabs going forward. That's pretty clear from where we sit today. However, there will probably be some discontinuities, and there will be the emergence of other players doing other things. We don't particularly see it as a problem.
EET: IP protection in China is still an unsettling issue. Does this raise special concerns for an IP company like ARM?
East: It doesn't raise special challenges. As I said earlier, we defend our IP and take steps to protect it, but I think the opportunity far outweighs the threat in that regard. Also, I think the general climate toward respect for IP in China will change.
Warren East Education:
Current position:
Professional history:
Affiliations:
| |
All material on this site Copyright © 2005 CMP Media LLC. All rights reserved. - - | |
Related News
Breaking News
- Frontgrade Gaisler Unveils GR716B, a New Standard in Space-Grade Microcontrollers
- Blueshift Memory launches BlueFive processor, accelerating computation by up to 50 times and saving up to 65% energy
- Eliyan Ports Industry's Highest Performing PHY to Samsung Foundry SF4X Process Node, Achieving up to 40 Gbps Bandwidth at Unprecedented Power Levels with UCIe-Compliant Chiplet Interconnect Technology
- CXL Fabless Startup Panmnesia Secures Over $60M in Series A Funding, Aiming to Lead the CXL Switch Silicon Chip and CXL IP
- Cadence Unveils Arm-Based System Chiplet
Most Popular
- Cadence Unveils Arm-Based System Chiplet
- CXL Fabless Startup Panmnesia Secures Over $60M in Series A Funding, Aiming to Lead the CXL Switch Silicon Chip and CXL IP
- Esperanto Technologies and NEC Cooperate on Initiative to Advance Next Generation RISC-V Chips and Software Solutions for HPC
- Eliyan Ports Industry's Highest Performing PHY to Samsung Foundry SF4X Process Node, Achieving up to 40 Gbps Bandwidth at Unprecedented Power Levels with UCIe-Compliant Chiplet Interconnect Technology
- Arteris Selected by GigaDevice for Development in Next-Generation Automotive SoC With Enhanced FuSa Standards
E-mail This Article | Printer-Friendly Page |