Starting off as an intellectual property (IP) provider of the ARM embedded Risc microprocessor core, ARM Holdings has expanded into creating and licensing other technology for the silicon and software components that go into devices such as mobile phones. This enterprise may irk some its partners, acknowledged ARM Holdings chairman Robin Saxby.
ARM formally launched the OptimoDE data engine technology for embedded signal processing applications in May this year. In October the company launched its Neon technology, a 64/128-bit single instruction multiple data (SIMD) instruction set that will be implemented in ARM processors, for next-generation media and signal processing applications.
Along the way, the company has also introduced IP for other technology such as multimedia and graphics accelerators.
In the interim, ARM’s business is growing. This year, ARM’s licensees, which the company calls its partners, will ship 1.3 billion chips with ARM microprocessor cores, said the company, which also claimed its cores go into 85% of mobile phone chip shipments.
Saxby spoke about the company’s strategy in extending its IP portfolio, and how its partners such as Texas Instruments (TI) and Intel are likely to react to ARM’s growing portfolio.
Starting with the ARM processor, ARM is now offering IP for a variety of components that connect to the processor, such as the OptimoDE data engine, and multimedia and graphics solutions. What is the strategy behind these moves to offer more and more of the components that go into devices like mobile phones?
We are already the world’s number one IP provider, but clearly we have to grow our value. To grow our value, we provide more of the pieces of the chip, more of the value-added IP that go into these chips.
Another equally important strategy though is to drive the software component and the software business. If you have the latest Sony Ericsson phone, it downloads Java applets and does very neat games from the web, very quickly and very power efficiently. That phone has the Jazelle hardware engine and the Jazelle software engine in it.
The advantage is that when we license the software we get a royalty from the software supplier as well as the royalty from the chip supplier, so in effect we get double royalty per chip.
Doesn’t your strategy of getting into newer IP ruffle your partners, who may have something similar to offer?
The thing about partnerships is that if you don’t upset your partner a little bit, you probably won’t be doing anything. When we started ARM, a lot of our partners did their own tools, but they were all different, and they were all incompatible, and we said we can do tools better. Now every partner we have endorses ARM tools, because when we come out with a new architecture, or a new model we have the tools readily available.
The OptimoDE has very digital signal processor-like features. Do you think that could upset TI, which by the way, hasn’t yet licensed the OptimoDE technology?
I believe that with the roadmap and the choices we have made, our partners over time will license what we are doing, because it will save them money and get them to market faster. That is basically our strategy and our roadmap. It would be stupid of us to go head-to-head with any partner.
Equally, the challenge for our partners is the cost of chip design. The cost of chip design from 0.35 micron to 90 nanometers escalates from $500,000 (£270m) to $24m dollars to do a chip. By licensing ARM technology, effectively we are giving them good value for money, which is helping them save money.
When we started off, every one of our partners did their own microprocessors. Now we have 134 partners, and even Intel is licensing ARM technology. They call it XScale to make it look a bit different because that is good for their marketing.
But they get a better return on their investment by licensing it, because the design costs are ours. If our stuff doesn’t work when you license it, you don’t pay me. If your own internal design team goes over time or over budget, it still costs you money.
Intel has dropped the ARM name with XScale and has also added new instructions. Are you having problems with that relationship?
We have a great relationship with Intel. We developed a product called StrongARM with Digital Equipment Corporation and Intel bought Digital chip business. Intel is a company that believes strongly in branding, so in the next generation of StrongARM they chose to call it XScale, because they wanted to promote the Intel brand.
Every partner has their own names. TI calls it OMAP. It just so happens that Intel probably spends more money on branding than everybody else. But the reality is that the Intel XScale is compatible with all other ARM architectures. They have some extra instructions, but in a system chip, on a TI chip, you got all the extra DSP instructions. So it is no different.
All ARM partners have to be able to have some differentiation in order to make their money.
Where do you see competition to ARM coming from?
Competition comes from within our partners. Everyone of our partners designs their own microprocessors. They don’t have to give us purchase orders. They don’t do it because we are nice people. They do it because we give them more cost-effective solutions than they can do themselves.
Within the partners, there is always the make versus buy decision. We do also have direct competition from players such as MIPS Technologies.
John Ribeiro writes for IDG News Service
This was first published in November 2004