Computer chip giant is gearing up for the next front in its battle with AMD, its chief technology officer, Justin Rattner, tells John Collins
It's easy to be generous to your opponents when you think a battle royal is swinging in your favour. That certainly seemed to be the stance adopted by Intel's chief technology officer Justin Rattner when he visited Ireland last week for the opening of the Intel co-funded Innovation Value Institute at NUI Maynooth.
Rattner conceded that it was Intel's fault that AMD had been able to get a number of new technologies, such as the more powerful 64-bit processors, to market before its larger rival. From being seen as an also-ran in the microprocessor market a few years ago, AMD is now a significant rival to Intel and now has about quarter of the market for servers and desktops that run on "x86" processors. "X86" refers to the chip architecture developed by Intel which now runs the vast majority of Windows computers.
"I think we got ourselves into a position with the Pentium 4 micro-architecture where we just tried to stretch it beyond what was a reasonable point," said Rattner. "In some sense, we didn't heed our own warnings."
But as you might expect from a man who has worked with the chip giant since 1973, Rattner is quietly confident of Intel's position as new generations of technology are introduced. Press and analysts have criticised some of the choices Intel has made about how to implement multi-core processors - packages which contain a number of chips or "brains" rather than the traditional single processors. This is the next front in the battle with AMD, but Rattner believes Intel has made the right decisions which will allow it to shortly introduce quad-core processors ahead of the competition.
"The bus architecture has given us one very unique advantage and that is the ability to introduce first dual and then quad core, very early," says Rattner. "It's extremely straightforward to package two dual core die and we get really excellent performance out of that dual die approach. The reason the competition isn't doing that is they can't, it's technically impossible. They have to wait until they can put four processors on a die, so Intel is going to be in the market probably six or nine months before we see it from the competition."
Despite the fact that Intel has a lead in what Rattner calls the "core wars", he believes the importance of multiple cores can be overplayed.
"I think a number of us are concerned that it becomes cores for the sake of having cores rather than cores for user benefit," explains Rattner.
The two areas that he expects to benefit immediately from having four cores in a single processor package are widely contrasting - financial services and computer gaming. Rattner says competitive advantage in financial services is increasingly delivered by information technology and so firms in this sector readily embrace new technology. On the games front, he says the realism of games can be enhanced significantly by dedicating cores to tasks such as running the physics engine (the "rules" of the games virtual world) or providing artificial intelligence that makes computer opponents act more like real ones.
Finding applications that will encourage other sectors to invest further in IT generally and Intel-based products will be part of the remit of the Innovation Value Institute (IVI) in Maynooth. One of several such centres that Intel sponsors worldwide, it looks at how businesses and the public sector can get increased value from their IT investments. The Irish centre will look initially at further developing and encouraging the uptake of frameworks which measure the maturity of a company's IT infrastructure.
Rattner is particularly keen on the ability of the IVI centres to demonstrate technologies around artificial intelligence, or machine learning as it is more commonly known now. In fact, Intel makes use of this new generation of software technologies in its Fab 24-2 facility in Leixlip, Co Kildare, which was officially opened in June.
While the underlying mathematics that the software uses may be beyond the comprehension of most people the application is simple - it enables Intel to service multimillion euro chip manufacturing machinery at the optimum time.
While such examples should encourage large enterprises to look at how they can benefit from advanced software, the reality is often more prosaic. A report from industry analysts Gartner released in the same week that Rattner visited Ireland showed that 91 per cent of technology spend is on maintenance of existing technology rather than investment in new systems.
It's a problem that Rattner acknowledges, but again he believes centres like IVI can help by bringing together a number of industry players working on open standards.
"I think there is a structural fault in the profession in how governments and industry get value from IT," says Rattner. "Intel's role is to act as a catalyst. We've been struggling with this ourselves. We've developed some frameworks and practices and we've been nurturing a community of like-minded organisations from public sector groups to end users and fellow players. Having people like SAP and Microsoft on board is really important."
It's not a problem that is likely to go away any time soon as companies like Intel continue to produce more powerful processors and software companies find new applications to harness the additional power. In fact, Rattner says he is confident that Moore's Law, named after Intel co-founder Gordon Moore, will continue to hold true for at least the next 10 years. Moore's Law states that the number of transistors on a computer chip, and hence its processing power, doubles every 24 months while costs stay the same. Currently, Fab 24-2 in Leixlip manufactures chips using 65 nanometer (nm) process technology and Rattner says he is confident the law will hold for the next generation 45nm and 32nm processes.
"The challenge of taking the next step is increasing in a non-linear fashion," admits Rattner. "It's getting harder and harder and in fact we are on the verge with 45nm technology of changing the basic transistor architecture, going to a new device stack, and bringing new materials into the Fab. It's not just a simple scaling exercise."
