This post is a follow-on to my previous report on the 21 April 2009 Xeon 5500 kickoff event hosted by Intel corporation (see: Intel’s New Chip ). My previous post talked about the new “Intelligent Processors” being fielded by Intel. With this post I want to provide a bit more information on a key presenter there, Justin Rattner.
It turns out Justin works for a pretty cool company. While talking with Justin and Nigel Ballard (Intel’s Federal Marketing Manager) I mentioned that I run a blog focused on technologists and they invited me to attend a small session with members of the press (this is another great reason to run a blog, you can elbow your way into all sorts of places!).
Some of what I learned from this session with Justin follow:
The new Nehalem chip may very likely change the security game. We got hints of that at the presentation earlier in the day when he talked about the enhanced ability to support encryption on the chip. When we were able to ask follow-on questions he underscored something that many of us in the profession have come to realize—encryption is just one of many capabilities that must be brought to bear against the threat. The best approach is a defense in depth approach.
As part of a defense in depth approach, the new management capabilities of Nehalem provides an ability to watch what is going on in the processor. Little chips on the processor watch the big chips on the processor. That enables review of the code that is executing.
When fully taken advantage of, this capability will significantly mitigate the ability of bad guys or malicious code to install and run root kits, for example. No anti-virus capability can detect a well masked rootkit. None of the OS guys can do this. It must be done on the chipset. The management module on the new Nehalem chip can provide that ability. It can scan memory and watch processes and compare against known valid code and take action. In Justin’s words: “It has become painfully obvious that none of the current defenses are adequate.” A stand alone management system that can command other processor components is a fundamental solution that will change the game.
I asked Justin one of my favorite questions about Moore’s law. We all know no mere mortal should ever try to predict its death, but it is time to predict its loss of relevance, I believe. Afterall, the performance statistics provided by Justin in his opening session showed dramatic, exponential performance enhancements that are far greater than the Moore’s law driven increase in number of transistors on a chip over time. So I asked Justin to comment. I asked, if Intel is blowing past the performance predictions of Moore’s law, isn’t Moore’s law less relevant? Shouldn’t some sort of “Core’s Law” be cited instead, where the number of cores double every two years? Isn’t that a better predictor of performance?
Justin’s comment: Intel has now realized that Moore’s law is no longer about performance. Moore’s law is remarkable in terms of predicting the number of transistors, but the number of transistors it takes to increase performance count has grown geometrically. In general, a 2x increase in performance might require a 4x increase in transistors. Smarter design is required. Smarter design, like multi-core and multi-thread gives dramatic increase in performance and many other benefits as well, including an ability to manage thermal and power budget on the chip, or adding virtualization support on the chip. Doing things like this requires transistors, but it also enhances overall performance.
I asked Justin about the internal tensions between people who are both consumers of tech and high tech workers. As consumers they are receiving more and more capabilities, from handhelds to desktops to cloud services. Then they go to work and as enterprise workers they are served with sub standard IT. He knows this tension well and knows it is occurring everywhere, but sees big gains being made in servers available to enterprises and believes now is the moment to upgrade to provide increase enterprise capability to workers. He believes in establishing private clouds, but also realistically knows public clouds will be with us forever.
I asked Justin about the handheld market. There are over 3.3 billion active cell phone subscriber accounts, for example, and only 1.2 billion PCs. Justin was clear in this: Intel recognizes this market and is coming. Expect their continued pursuit of a wide range of consumer products, building on success of their Atom chip and netbooks. Look for Intel moving faster into embedded computing. Opportunities are expected to “explode” in digital signage, automobiles, healthcare, and personal ecommerce.
I asked about Twitter. It is in use by many folks at Intel, and there are internal collaborative tools and social media inside the corporation. It seemed clear that they are not designing chips by 140 character Twitter micro-blog bursts, but they are enhancing internal coordination and cooperation. And that is good for all of us IT users.
Overall I found Justin to be a world class tech leader and I felt honored to be in the same room with a person who is changing the world in such a huge way. Thanks Justin for the time.
- An OODAcast Conversation with Dr. David Bray of the Atlantic Council Geotech Center (Part One) - April 3, 2020
- OODAcast– A Conversation with Dan Gerstein - April 1, 2020
- Update on The End Coronavirus Project and Need for Volunteers - March 28, 2020