Only 500 leopard wild cats survive in the almost urban jungle of Taiwan and are on the verge of extinction. In order to protect the endangered species, DT42, a Taiwan-based deep learning startup is using AI (in collaboration with Taiwan’s Directorate General of Highways) to “detect leopard cats when they near roads and keep them out of harm’s way, reducing roadkills.” Just a reminder that while we can fear AI, it can come in handy.
Each Tesla computer has two AI chips, a redundant design for better safety, Venkataramanan said. There’s redundancy in the chips’ power supplies and data input feeds, too. Even the car’s cameras are on two separate power supplies to guard against failures. ….Each Tesla AI chip runs at 2GHz and performs 36 trillion operations per second. That performance is possible because Tesla optimized the chips for self-driving cars and dropped anything more general purpose…..For example, the chip handles data recorded as 8-bit integers instead of the 16-bit floating-point numbers more common in AI tasks but that require more power to process. For another, it’s got an extremely limited set of instructions it can process. And it’s got a gargantuan 32 megabytes of high-speed SRAM memory on the chip, which means it doesn’t have to wait around while fetching data from much slower conventional DRAM memory.
Apple has taught Silicon Valley the importance of owning its own chip destiny and now pretty much everyone willing to push the technical edge is building their own chips, for more vertical integration in their designs. Tesla gave a glimpse into how it can keep ahead of its deep pocketed rivals. At the Hot Chips conference, Tesla showed details on its newest innovation. The chip took 14 months to design and Samsung is going to make the processor. It is in newer Tesla cars. It is 21-times faster than the Nvidia chip they were using and about 80 percent of the cost.
Also: live blog from HotChips conference that has more details.
As usual, this has been a real “Dr. Jekyll and Mr. Hyde” kind of week for Tesla. It had a disastrous earnings report card, and Elon keeps creating all the wrong sorts of headlines. But in the middle of this maelstrom, the company announced a new chip that is going to eventually become the brain for their electric car. This chip is not just any chip — it will be able to make sense of a growing number of sensors that allow the car to become better and better at assisted (if not fully automated) driving. Continue reading “What Did Tesla Do Now?”
It is a damn shame to see AMD fall on tough times. They were always a great story. Every new chip breakthrough invigorated interest in the company. For a while their chips powered my home-assembled, constantly upgraded PCs that I used to put together in my NYC East Village apartment. I still remember this beautiful 15-inch gaming laptop — Ferrari logo and seriously fast and powered by AMD. It had great graphics, terrible battery and would warm up a cold apartment. The fans were like two turbochargers. I played Age of Empires on many a few cold winter nights. The computer almost always got me in the end, much like Intel got AMD.
From the very day I started covering the semiconductor industry, I have thought of Advanced Micro Devices as an interesting, slightly crazy and yet fun upstart. I can’t remember when it shed its “perennial also ran” status. Its founder and chief executive Jerry Sanders was a colorful man, known for his luxe lifestyle and one-liners that even Aaron Levie can’t match. He was a consummate salesman, a great marketer and started the company with seven colleagues from Fairchild in 1969. And he lacked fiscal discipline and lived larger than his means — which ultimately became the company’s culture and its undoing. It started out strong and had about a dozen great years but the rise of the PC turned it into the bridesmaid, but never the bride. And ever so often the company would come up with a new chip to compete with Chipzilla, a moniker The Register founder Mike Magee had for Intel Corp.
For most of its life, the company was an underdog and relished in that role. It was essentially a second-source supplier to companies that were using Intel’s chip. It has a technology agreement with the giant chip maker and it always wanted to figure out its own path. Intel, especially under the leadership of Andy Grove was so good and dominant and rich that even when AMD seemed to have good products, Intel would muscle its way ahead through sheer financial power. Still, AMD did a lot of interesting things — and in fact pushed the envelope for a few years.
In 1995, AMD acquired a company called NexGen and that sowed the seeds of what would become AMD’s rise to glory. In 1997, the company released K6 chip the slow resurgence started. Then came K7, dubbed Athlon. From 1999, when it released its Athlon line-up of chips to 2005, the company produced great chips and the market responded accordingly. AMD even filed a law-suit against Intel for anticompetitive business practices — the good old days when being a reporter was really a lot of fun and you could read legal documents at leisure.
One has to tip one’s hat to AMD’s engineers. (Many have made their way to Apple by the way.) AMD was first to put memory controllers on the CPU and lead the way on putting more capabilities on the CPU itself. It came up with 64-bit extensions and even Intel licensed their x86-64 technology. It went after the server market with its Opteron processors and carved up nearly 25 percent of the server processor market. It pushed multicore processing technologies hard. However, all these were for naught as the company was spending money on fabs and facilities. Sanders is rumored to have once quipped, “Real men have fabs.”
Sanders’ bombast and a whole lot of other executive suite challenges have kept the company from realizing its true potential. It has been run by many chief executives and none of them could stall the slow-motion tailspin. It was always buying a company (NexGen, ATI Technologies, Alchemy, Geode, SeaMicro) or selling/spinning off its divisions (to Spansion, GlobalFoundries, Qualcomm, Broadcom and others), restructuring or doing something — anything really to find a way ahead. AMD perfected the art of activity without achievement, and allowed Intel to stay dominant.
AMD can blame Intel and the markets for its eventual downfall, but the troubles are its own making. AMD’s long standing problem has been mismanagement and big-spending ways. It was apparent even in the halcyon days — and you could pick up that attitude from the executive. Even as late as 2008, when I simply backed away from writing about chips, the company was in denial about its future. It was becoming a flash memory company with some microprocessor and graphics chip business. It was one big tangled mess, according to Ars Technica which ran a two part series on the company’s troubles — it’s worth a read.
AMD, in many ways is a perfect example of a company that had the right ideas, could find the right technical talent and yet couldn’t execute on its own plans, brought down by hubris. I have seen this many times in Silicon Valley and elsewhere, the most recent being Nokia, which too was a great company with smart people and smarter ideas, but too beholden to an image and carriers.