FOR THIS SECOND PART of our interview with Intel's senior vice president of the Digital Enterprise Group, Pat Gelsinger, we kicked about some of the big ongoing changes in computing models. Unsurprisingly perhaps, the Intel man is not big on client devices that don't have powerful local processing, but we started out with a grouse about laptop batteries that run out when you are halfway through your artic...

Mobile computing has obviously been a huge change in the way we consume MIPS but, I moan, how come laptops still don't last all day.

"What’s the goal?" asks Gelisinger. "All-day working, which I define as 10 hours with Wi-Fi on all day. We’ve done well on the CPU and the next problems are at the component and OS because when the system is doing nothing it’s still checking [for activity] every 100 milliseconds. Changing some of the attributes is hard but we’ll get to all-day computing whether it’s [helped by] ripping out the laptop hard drive or something else."

OK, but when? Gelsinger plays a straight bat, saying that the solution is " several years out".

So what about server-based computing? Intel has a vested interest in perpetuating the client/server status quo but aren't all signs pointing to the server in the cloud, or at least the datacentre, becoming the boss of next-generation computing?

Gelsinger wants to take that one in parts.

"You can have thin terminals which is back to 3270 terminals. We’re not proponents of that model. We think that’s like going from future to the past because server MIPS are more expensive and there’s the myth that bandwidth is free and there’s no networklatency. There are other examples like VDI and that’s not a whole lot more efficient. The model we believe in is streaming where you put state on the server and store data locally. That’s the best of both worlds where you have the datacentre control the IT guy wants, the [centralised] storage for things like Sarbanes-Oxley, and the traffic is very low. The client can be a diskless client or a client [with disk] but it depends on which things you want to do. Dell offers diskless and you have no state. You can put the disk in and use it as a cache, or still have the local OS. What we’re trying to do with vPro is make the client identical [to enable all scenarios]."

What about virtualisation? Like Victor Kiam, Intel liked it so much it bought (a chunk of) the company when it paid $218.5m for 2.5 per cent of VMware, earlier this year. It's a sobering thought for would-be competitors that Intel has already made more out of that investment than many well-known companies will turn over this year, but, with one eye on lost server sale opportunities maybe, Gelsinger isn't getting carried away with the hypervisor hype.

"We believe it’s still in its infancy and has a long way to go. Less than 10 per cent of servers use it in any form and we haven’t scratched the surface in client."

Further out, load balancing, failover, redundancy, even mission-critical computing stand to gain and the resulting clusters could be “more robust than a mainframe”, he says, but it's early days and the competitive landscape is not overly, well, competitive.

"VMware has been the only game in town," he notes. "Before we would have said it was a three-horse race with XenSource and Microsoft, but [they] haven’t really showed up for the race."

Is VMware going to stay the 800-pound monster though? Maybe not, thinks Gelsinger, pointing, without prompting to Parallels and Virtual Iron as well as renewed efforts from Citrix-XenSource and Microsoft's Viridian.

Still to come this week: Gelsinger on the Classmate PC, OLPC, the Tablet PC, how Intel thought about buying a GPU company, accelerators, the Wintel relationship, how AMD taught Intel a lesson, his Mac habit shame, and the truth about kicking Mike Magee. µ