Intel surprised many observers when the company hired outsider Venkata Renduchintala to lead the company's PC, Internet of Things, and Systems Architecture groups.
With more than a year under his belt, he's spearheading a cultural change inside the company, getting employees to think beyond PCs and talk about technologies like 5G and IoT.
There's been a lot of chatter about changes in the company's chip development strategy, with the recent announcement of the 8th Generation Core processors, an unprecedented fourth chip architecture on the 14-nanometer process. The chip industry veteran sat down with us at Mobile World Congress in Barcelona to talk about what spurred the move and also his thoughts on 5G.
How is 5G development going?
Renduchintala: I'm noticing there are a hell of a lot more people interested in 5G. The interest in anything that happened before -- whether you're an industrial conglomerate, a car company, a transportation company -- you are now talking about much more interest in 5G than ever before. It's the whole age of everything coming down to data. There are so many colliding identities that get aggregated under the banner of 5G. For example, IoT is now becoming inextricably linked with the passage of 5G.
5G will be across a wide range of devices. What can users expect?
Renduchintala: You'll have network operators becoming more like a cable operator where you can basically get your phone, internet, and TV all in one contract. An autonomous car, all of your cell phones, your tablets, your media consumption, and even the automation and control of your house via an IoT network, is all going to be in one bill. It's going to be different types of networks all generating different types of traffic but ultimately concatenating into one of all integrated services. A network operator is going to basically say: I deliver media, I deliver services, I deliver mobile broadband communications, and it may be delivered over two or three different network architectures that all work together.
What's happening to the 5G network?
Renduchintala: What's happening in the network, although it may be more esoteric to explain, is actually just as profound. A network that used to be custom pieces of silicon -- that used to be hand-stitched together to give you performance optimization and load balancing -- is now collapsing down to a reconfigurable entity that used to live in the data center [and is] now being shoved into the network. It's general purpose computing, general purpose data distribution, reconfigurable memory and programming logic, all of which can be reconstituted by the upgrade of a binary in terms of the function.
Do the cloud providers understand the implications of 5G?
Renduchintala: It's difficult to actually see. It's only when you really stand back and look at the totality of what's changing you'll actually understand how profound it is. They're no longer going to say, I build five or six mega data centers and stuff. Let's say, for example, Google wants to build out a network that is supporting self-driving cars. I don't think its network architecture is going to look anything that it looks like today, where its main service is essentially providing things like Google Drive and offsite data storage and retrieval. It could be a completely different architecture.
You are developing modems a lot quicker. Why is that?
Renduchintala: The cadence is not set by Intel. The cadence is set by the market. You're going to have to have an upgrade of a modem every year. There's a big transition coming along probably in around the late 2018/2019 timeframe where instead of incrementally evolving LTE, you are going to flip to your first multimode LTE/5G. And then you get to this big disruption and then get a big stride in incremental improvement thereon. And then over time, it becomes less and less discernable to the end user. To me, this is like the classic period of a modem upgrade followed by a rapid improvement on the very first generation device and then the ability to get more and more improvement becomes diminishing returns over time.
With 5G supposedly around the corner, how valuable is gigabit LTE?
Renduchintala: There's nothing innovative in gigabit LTE -- all you are doing is, instead of aggregating two carriers, I'm aggregating four carriers over a wider bandwidth. That means bigger memory, bigger chipsets, more power -- there's nothing technically mind-blowing. It's just arithmetic extension -- if I want to get a gigabit per second, I have five carrier aggregation across 100 megabits of contiguous bandwidth and therefore I get more data. With LTE, you're really getting to a point where we have to basically go towards both spectrum efficiency gains as well as better data resiliency and service security. That's going to be truly profound.
At Intel's recent analyst conference, you mentioned 8th Generation Core chips coming out on the 14-nm process, while Cannonlake is around the corner. Why the change in chip development plans?
Renduchintala: What we're moving towards is a model where people can expect from Intel a yearly cadence of platform upgrade that actually gives meaningful performance improvements in the generation that preceded it. For example, Skylake to Kaby Lake was about 15 percent in overall system performance improvement. What we said is we will deliver at least 15 percent improvement from 7th Generation to 8th Generation.
We're focusing on making sure we get meaningful performance upgrade on an annual cadence and to some degree, the process technology that lives underneath that is going to be less conspicuous. There will be a much more of a seamless transition from node to node. We can translate that into a more predictable cadence of product which delivers meaningful performance to stimulate PC upgrades.
Will Cannonlake be 8th generation or is it a different lineup?
Renduchintala: I don't know yet. I think we're going to be more focused on generation by the amount of performance increment it will give us. I don't think that generations will be tagged just to node transitions. It will be tagged towards the fact that it needs to have meaningful performance [gains]. Whatever we launch in September of 2018 has to have at least as much improvement year over year as what happened between Kaby Lake and the 8th Generation. If Cannonlake comes out at the end of the year, it could be interesting what we actually market it as we haven't actually decided yet.