Why Intel's labs celebrate product failures
24th Oct 2012 | 19:04
Ditching expensive failures before they can reach the market
Intel's chief technology officer believes the company's research arm has a key role to ditch expensive failures before they can reach the market.
"Let me suggest that there's another role for an industrial research institution. And that's to prevent poor technologies from ever reaching the market. Product failures – at least in our industry – are extremely expensive," says Justin Rattner. As well as being chief technology officer for the multinational chip giant, Rattner is also Director of Intel Labs.
TechRadar joined Rattner at an Intel Labs get together in Barcelona, where the lab has just celebrated its 10th anniversary in the Catalan city.
"A typical mainstream processor will cost well above $500 million in development and will be built in a fab costing ten times that, so you're talking about an investment in the order of around $5 billion dollars.
How the PC of the future is closer than you think
"If a product flops because of a bad technology decision made somewhere earlier in the technology development, that's really a big problem. And yet, some failures can and do happen. Sometimes they're not public, and sometimes are very public."
Rattner chuckles at this point, perhaps remembering the complete flop of Intel's last-generation Pentium 4 processor. The Prescott chip was power hungry, hot and convinced Intel that it had to change direction and focus on mobile with the Banias Pentium M and subsequent move to the Core microarchitecture.
"I think a 21st century research lab can see a main of its role is to weed out the bad [tech]. We refer to this as 'failing fast' and it's something that we consider as important to the long term success of Intel as all the things we do on the innovations side."
Moreover, Rattner says that researchers should celebrate weeding out these failures and says that other labs aren't doing things right if they're too focused on the future picture rather than sense-checking products. "It should be the responsibility of every research lab to celebrate these kinds of internal failures as well as internal successes There are still a number of companies chasing their vision. We believe these labs are increasingly out of step [with everybody else]."
"The sooner you grasp the cost of failure [you appreciate how important this work is]. It climbs exponentially the closer you get to product delivery. So we'd rather have a million dollar or ten million dollar failure in the lab than a billion dollar or multi-billion dollar failure in the factory.
"Every once in a while everybody's talking about some chip that someone put out where something has gone awry – trust me that costs Intel hundreds of millions of dollars. Some of those failures are almost legendary.
Rattner also says that a research lab can help with situations where a product might fail in the market rather than being a technical failure.
"The market for technology has really changed in the last decade, even the last 4 or 5 years, he says. "Customers are beyond the point where technical novelty sells the product.
"People are looking for solutions that make their lives better or more fun or making things [better] in the corporate space. So at Intel there's another transformation under way to come at future products from that perspective as opposed to 'well, we've got a billion transistors, can we just throw everything including the kitchen sink into it and people will just buy it because it's cool and the spec sheet goes on for pages'. Those days, if not completely over, are numbered.
"In the labs, we take on the question market success by creating capability that really goes directly to consumers and customers to find out if a product will be a success."
Finding a path
Rattner says for the first 15 years of its existence Intel avoided using the term 'research' unless it was directly related to product development. This was because Intel's founders, including Gordon Moore, had problems progressing semiconductor research at their old company. It wasn't until the mid 80s that asmall research team were allowed to form.
Windows 8 review
Windows 8 vs Windows 7: 8 ways it's different
50 Windows 8 tips, tricks and secrets
Windows 8 tablets: what you need to know
Making sense of the Windows 8 versions
All our Windows 8 content
"It was a relatively small organisation if you think of Intel's industry leading semiconductor technology" muses Rattner. "Many of the stunning Intel chip advances were work created by this one research team. What separates them is how they get their inventions out of the lab and into the factory. The key to this is pathfinding."
"The question was 'how do we improve the hit rate of technology coming out of the lab and into products'? It was the fact that the timing difference between what was going on in research and what was going on in product development was out.
"It was difficult to get [technical] developments into the product development part of the lifecycle. The research had potential to kill technologies before they had gone anywhere; it was like a 'valley of death' they had to pass through. This includes technologies such as virtualisation. We doubled down and managed to get virtualisation on the next generation processor feature set.
Rattner says there's no actual pathfinding department at Intel. "Really the pathfinding department is a senior technologist and an admin. It's formed on the fly of researchers and developers who come together for a period of, say, 18 months with the purpose of finding a technology for a future development processor technology.
"They've hit every Moore's Law deadline, often with months to spare. They have a 100 per cent success rate. The challenge was how to scale that pathfinding process to cover the dozens of different technologies that go into a new Intel chip. Despite our fears, extending this to various areas of research has been a huge success for our company." Intel now has over 50 pathfinding teams at any one time.