AT IDF THIS morning, Paul Otellini gave a broad overview about the future of Intel, and how it is becoming a full stack provider. Everything was about what you would expect until about half way through when the caffeine induced hallucinations kicked in.
The vision for the future of Intel has three pillars, transformation, essential compute developments, and making everything smart. The transformation one is probably the biggest, with Intel going from a chip maker to a full compute stack provider. This means hardware, software, and programs, all from Intel.
The new Intel stack
If you recall, Intel bought Wind River a while ago, and that is part of this transformation. Adding security through McAfee is another. The idea is to have Intel adding value at every part of the chain, or at least have an option for that.
The compute essentials piece involves three sub-parts, energy efficient performance, security and internet connectivity. Energy efficient performance is driven by transistor technology, and this time it was focused on 22nm. Intel said they had 22nm CPUs going through the fabs now, with shipments in the second half of 2011. Not bad.
Security was the reason for the McAfee buy, bloat, false positives, and painfully obtuse UIs were not mentioned. It is a laudable goal that could have been much more easily achieved by dumping Microsoft, but that move would not generate revenue for Intel. I wonder if the lessons about why the vPro antivirus filter didn’t work out so well sunk in. White lists were also mentioned, but did Intel really need to spend $7B for that?
Last of the three is connectivity, which is why Intel bought Infineon’s wireless business, and TI’s cable modem chip business. If you have to think about how this would tie in to Intel’s strategy, go get another cup of coffee, we will wait till you get back.
The last of the initial three points; making everything smart. This means Google TV, apps that are tied to a single device, and graphics. This is where things got weird, something I will blame on the lack of caffeinated coffee in the executive green room.
Ummm…. Riiiight. Caffeine NOW!
What is the problem? Intel claimed that they would increase GPU performance by 10x this year over an unspecified earlier point. So far so good. The problem happened when Paul claimed that it wasn’t a 10x jump, it was a 25x jump. No, not with Sandy Bridge, but with the current stuff. Someone get this man some espresso, no make it a double, we don’t want to lose him.
From there, there were some demos, the most interesting was a four way fully encrypted video conference running on a Sandy Bridge. Can you say AES-NI funneled to the IGP? I knew you could. In any case, once bandwidth catches up, this will be a very nice business tool to have, especially when you are on the road.
The last bit was the Romley platform, basically Sandy Bridge EP. This part is known by basically everyone. What is new is that they said that the two socket machine has 16 cores and 32 threads. With a little math, we can confirm what everyone strongly suspected, Sandy Bridge EP has 8 core variants.S|A
Latest posts by Charlie Demerjian (see all)
- More on Intel’s 10nm process problems - Sep 17, 2018
- Intel puts out another 14nm 2020 server platform - Sep 11, 2018
- Why Can’t Intel Supply Enough 14nm Xeons? - Sep 10, 2018
- Intel can’t supply 14nm Xeons, HPE directly recommends AMD Epyc - Sep 7, 2018
- AMD reintroduces the Athlon name with two CPUs - Sep 6, 2018