In a high-tech game of leapfrog, makers of the world’s fastest supercomputers (and their countries of origin) compete to assemble the machine that will sit at the pinnacle – a determination that happens every six months and is published on the website Top500.org. Taking the number one spot is seen as a validation of a supercomputer company’s technical prowess, and has become a proxy for the innovativeness of countries at large. This clash of the countries started in 2002 when Japanese company NEC reversed years of American leadership by developing a machine that debuted at 5 times faster than the previous number one machine, an IBM system. Although the NEC machine, dubbed Earth Simulator, stayed at number one for two and a half years (a record for longevity at the time), its fundamentally limited architecture prevented it from growing in a financially viable way. In late 2004 IBM took back the mantle with a system twice as fast as Earth Simulator yet one that consumed a fraction of electricity and floor space.
At the end of the month in Frankfurt, Germany, the Top500 group will formally announce that another IBM system nicknamed Summit will grab the top spot – the first time an American company has taken the crown since China achieved number 1 five years ago – and has continually maintained the position since then. (Sidebar: Summit is about 3 million times faster than the IBM system that retook number one less than 15 years ago.)
This remarkable improvement in performance is partially attributable to the ongoing increase in the speed and improvement in the design of microprocessors – just as it is with laptops and smart phones. But today, thanks much to software innovations supercomputers are able to grow in scale by “simply” lashing together more and more computers that work interdependently, breaking up problems into smaller pieces that are solved in parallel.
In reality, the only practical limit to making the world’s fastest supercomputer is money. That’s why the cache of holding down the top spot on the Top500 list is as much about a country’s economics as it is about the computer maker’s technical chops.
Researchers in universities and national labs and less-smart members of Congress all soiled themselves when Japan swooped by native IBM and every other U.S. based vendor. (Some dubbed the arrival of Earth Simulator “Compute-nik” in a nod to American apprehension that accompanied the Soviet launch of Sputnik.) And that’s why the half-decade Top500 leadership by Chinese manufacturers seemed to herald a depressing future where China would come to dominate the making of things more strategic than tee shirts, plastic toys and bootleg DVDs.
If China can fund what it takes to make the world’s fastest supercomputers over a long run, what will stop them from dominating any strategically important sector? No doubt they are conjuring up the resources right now to make Summit a one-termer.
The funny thing about the Top500 list is that the rankings are determined by running a computer workload that is not very representative of the kinds of problems actual companies encounter. It may have been OK 20 years ago, but today there is a deep disconnect between the problem that supercomputers run to get on the list, and what modern industry needs to conduct artificial intelligence – yet the old workload persists so that equivalent comparisons can be made year after year.
As an illustration of how the Top500 has become more of a marketing tool than an honest assessment of the state of computer power, consider how companies respond to the publication of the list every June and November. A search of past company press releases will show that when manufacturers like IBM or Cray achieve the number one spot (or some other milestone like the most entries on the list of 500), they pour out gallons of ink touting their innovativeness and commitment to the scientific community. When those same companies underperform, they make pronouncements that the Top500 is not representative of reality and being on the list is actually meaningless – until they capture the top spot again when magically the list becomes relevant once again.
In a telling article in Market Watch a few years back it was reported that the Chinese were primarily using supercomputers as a PR statement more so than to conduct research. According to the author, “it seems a lot of these massive machines, usually made with large government investment, lie idle after they are made, or are even abandoned midway, due to fundamental defects in China’s traditional bureaucratic management system.”
Perhaps one day the Top500 fever will break and companies and governments will focus their attention on investing in technologies that benefit a broad swath of humanity, and not just on the guys who are trying to perfect a nuclear weapon that can fit in your hat.
IHOP Flip Flops
Presumably sensing greater opportunity in the business of lunch over that of breakfast, the International House of Pancakes – better known as IHOP – is revamping its menu to major on burgers over sugar-soaked stacks of carbs. So dramatic is the move that IHOP is actually changing their name (and all their branding which could costs millions) to IHOB.
Kicking it off is this moronic TV ad campaign featuring Manager I-Blob atop the famous blue tiled roof of his restaurant flipping the switch on the new sign – and then, of course, goofing it up and falling off the roof like every half-wit white guy must do in a TV commercial today.
Who knows if IHOP can make such a radical change in focus and go into the very crowded segment against McDonalds, Burger King, Wendys and the many upstarts like Five Guys. In fact, can we be certain that IHOP won’t go in another direction in the months or years to come?
Rumors abound that in the era of Trump nativism the company wants to concentrate solely on the American market. Also, CEO Julia Stewart is known to favor Mexican food.
Might we then expect IHOB to change again, this time to American House of Lamb Enchiladas?
Will that doughy white guy in the TV commercial get back on the roof to flip the sign to … drumroll, please … A-HOLE?