NEW YORK -- A flurry of more than 20 new patents suggests Intel Corp. is expanding its time-tested legal strategy to prevent cloning of its new, flagship IA-64 architecture.
Some experts wonder whether Intel is taking a new tack in corralling its rights to IA-64. That is, rather than submit garden-variety claims to the U.S. Patent and Trademark Office (USPTO), Intel is trying to patent the functions carried out by specific instructions. In doing so, the company appears to be, in effect, trying to patent the IA-64 instruction set itself.
Intel has a long history of fending off cloners with a broad hardware-based patent portfolio. The IA-64 move highlights the growing importance of software in electronics designs and the need to protect those software innovations.
"I don't think that they Intel are going to license these patents," said Rich Belgard, a microprocessor consultant based in Saratoga, Calif. "I think they want to protect the IA-64 from cloning."
Itanium, which is the first chip in the IA-64 family, is already in the hands of Intel OEMs. The second processor, dubbed McKinley, is due next year.
Intel's strategy may be more than an academic exercise. For one, Mips Technologies Inc. has been fighting a year-long, still-ongoing court case against Lexra Inc. of San Jose, claiming that Lexra infringed four instructions protected under its 1989 patent No. 4,814,976, titled "RISC computer with unaligned reference handling and method for the same."
"Our customer base sees this type of litigation as business-as-usual," said Charley Cheng, chief executive officer and president of Lexra.
Among Intel's batch of recent patents are No. 6,115,777, entitled "LOADRS instruction and asynchronous context switch," and No. 6,092,188, "Processor and instruction set with predict instructions."
"It looks very much like a strategy to keep people out of this space by locking up these patents," said Ron Abramson, a partner at the law firm of Hughes, Hubbard and Reed in New York.
Consultant Belgard concurred that Intel appears to be "locking up IA-64 rights under the umbrella of Idea Corp." Idea Corp., which stands for Institute for the Development of Emerging Architectures in Cupertino, Calif., is a joint company set up by Intel and Hewlett-Packard Co., Intel's partner in developing the IA-64 instruction set, as an intellectual-property repository for the IA-64 architecture. Idea Corp., rather than Intel itself, holds the rights to most of the IA-64 patents.
According to experts, one of the pitfalls Intel has faced in the past was a failure to be assiduous in patenting its instruction sets. However, that pattern was reversed once the MMX-equipped Pentiums came along. At that time, the Santa Clara, Calif., company slowly began patenting the functionality of certain instructions.
Shift in the wind
"What's happened is things are becoming more liberalized in the patent area," said Leonard Rubin, an intellectual-property (IP) lawyer at Gordon, Glickson in Chicago. "Patents on business methods are being granted where they never would have been before."
Rubin pointed as one example to a patent Amazon.com received on its one-click shopping method, and noted that, once awarded the patent, Amazon sued competitor Barnes & Noble.
Whereas physical hardware dominated applications in the old days, today, said Rubin, "some companies are seeking patents they never would have before."
Rubin said that this change was likely due to a philosophical loosening up within the USPTO. However, he also noted that many IP attorneys were not happy with the trend.
Intel today seems uncertain as to whether its patent strategy is an evolutionary or a revolutionary development.
"We and HP are much better at the patent process than 20 years ago -- everybody is," said an Intel spokesman.
Many of the newer patents involve prediction, speculation and reformatting data. The latter relates to the branch of processor architectures known as very long-instruction-word (VLIW), which plays a role in IA-64.
The patent claims drill down deeper into certain very specific technical areas.
For example, patent No. 6,119,218 describes a "Method and apparatus for prefetching data in a computer system."
Computer consultant Belgard explained the apparent new twist in the patent by noting that, "Typically, nothing would happen as a result of a prefetch instruction, if an exception occurred -- that is, the instruction would be abandoned. This method of Intel's allows for the instruction to be optimally handled."
Belgard is referring to a "prefetch instruction" -- that is, one that performs a prefetch -- not to any other instruction that may have been "prefetched" during the course of executing a program. Prefetch instructions are invisible to the user, but are inserted by the compiler as needed.
Looking at that same patent from a legal perspective, attorney Abramson said: "Thislooks like a patent on something an instruction does. If nobody ever did an instruction
like that before, why not patent it?"
Indeed, patents often contain some claims that seem obvious to the practicing engineer. For example, dissecting the language of the abstract of patent No. 6,119,218, Abramson noted that it discusses executing a prefetch instruction, a function that's widely performed, so it's not the crux of the matter.
However, the patent appears to hinge on the discussion about handling an exception during a prefetch. If that is new, and cannot be discerned from prior art, then the patent should be rock-solid.
Asked if patenting the functionality of an instruction set was a back-door way to patent the instruction set itself, Abramson answered: "Well, it is tantamount to trying to patent an instruction set, to the extent that the instruction set has new instructions. If the instruction set has 'add with carry,' you can't patent that."