Rattner details Intel's multi-core future
Intel CTO Justin Rattner chose to focus on the future of multi- and many-core computing during the closing keynote of IDF.
The CTO stood up to those who posited more cores did not always result in better performance by giving an hour's worth of examples in which Intel's developers had managed to push past programming roadblocks.
"We're seeing excellent scalability. This has been very encouraging and has really given us a lot of confidence that people will put this to work," said Rattner, whose researchers are consistently charting the course to higher and higher levels of parallelism.
For example, Intel's engineers are currently working with scientists at CERN to build enough compute power to find the tiniest of sub atomic particles - the Higgs Boson.
CERN's Hadron Collider - technically the largest machine ever built by man - boasts a whopping 250,000 Intel cores across its distributed structure. Intel's Many Integrated Core (MIC) architecture programming has also eliminated the need for multiple programming models, said a representative from CERN, making it far more scalable and easier to use.
"Intel has delivered a very good machine that works very well with very dense code," explained CERN staff researcher Andrzej Nowak adding, "we will take as many cores as you can throw at us."
Rattner also did his best to dispel the myth that it takes programming geniuses to make use of many core computers.
"You think you'd have to be some kind of freak to program for multi core, but you don't have to be a ninja programmer," he said noting, "our goal at Intel is to banish ninja programmers forever."
Multi core should be able to go mainstream, said Rattner, saying it wasn't just for CERN scientists. As an example, Rattner cited web apps, which today account for a plethora of databases and content repositories, all struggling to serve up content concurrently to millions and millions of users.
A way around this, said Rattner, was to move the database into memory, putting all the heavy lifting on RAM, which is fast and offers a huge performance boost.
Indeed, an onstage demo showed a system performing 800,000 transactions a second across 48 cores, with Rattner predicting "one million queries per second is on the horizon."
"What we have demonstrated today only scratches the surface of what will be possible with many-core and extreme scale computing systems in the future," he added.