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.