Tesla's New AI Chip Is Part of a Bigger Plan
Earlier this calendar month, Tesla CEO Elon Musk appear that his company had adult a new electric car AI chip that'south 10 times faster than the current ones from Nvidia.
At Nvidia'southward spring developers conference, the company said its AI Xavier processor was the most powerful processor for AI bachelor. Then when Musk fabricated his declaration, I assumed the Tesla flake was 10 times faster than the Xavier, which would accept been quite a feat. Only in a recent call, Danny Shapiro, Nvidia's Director of Automative Engineering science, pushed back.
"The performance claims are against what they have in the vehicle today, which are iii years old," Shapiro told me. "Nvidia's latest silicon is at least 10 times faster than that, which would put information technology on a par with Tesla'south chip."
Without Nvidia, Tesla cars would not exist on the market today. What I disliked about Musk's proclamation almost his own processor is that he did not clarify that he was making a comparing to an older chip, and in essence, threw Nvidia under the bus. That is no fashion to treat a partner that bent over backwards to work with Tesla and help it achieve the success it has today.
That said, Tesla is moving toward vertical integration past creating its own AI chips and, where possible, creating components for its cars internally. In fact, information technology makes its own car seats already, and I doubtable that over time it volition move toward creating as many components for its vehicles as it can.
Google, Facebook, and others are doing something similar. The No. 1 thing they are trying to take control of is the core CPU and GPU needed for any ancillary hardware products they make to extend their products and services. Google is creating chips for smartphones, laptops, and AR and VR headsets, while Facebook is creating AI chips for its Oculus VR headsets and who knows what other hardware.
A company taking command of its ecosystem is not new; only look at Apple tree. Early, Apple tree obtained a core license to the ARM architecture and has congenital a custom semiconductor on top of an ARM cadre, assuasive it to command its own destiny at the chip level. To date, Apple tree nonetheless uses Intel chips in its Mac desktop and laptop products, but I would non exist surprised if in the next two to iii years, information technology migrates macOS to piece of work with ARM-based processors, too.
At the heart of these moves toward vertical integration is the desire to command a company's ain destiny. Apple has been criticized for the slow updates it delivers to Macs, but that is by and large due to its reliance on Intel's processors.
It is also at present much easier for a company to create its own processors by doing semi-custom design; in more extreme cases, they might build the fleck from scratch. In the past, for them to build these chips, near had to create their own fabs to brand them. Today, thank you to fabs like Global Foundries, TSMC, and Samsung, they tin can accept their design to these semiconductor manufacturers and let them build it for them.
In a recent post for Creative Strategies' Techpinions web log, my son Ben, who is much more semiconductor literate than I am, wrote about what is at pale when one's does custom silicon.
"By going downwardly the proprietary silicon route, companies are betting their differentiation on this strategy. I say it this way because of ane common theme with companies efforts in custom silicon is how focused it is on a specific surface area of differentiation. If one did an in-depth analysis competitively, y'all would likely discover where these companies believe they are differentiated past where they try to build custom silicon to heighten or deepen that differentiation.
"The run a risk, however, in this scenario is their efforts in custom silicon are not able to sustain their differentiation every bit companies like Nvidia, Qualcomm, Intel, AMD, etc., design better, more than powerful solutions that enable said companies competitors to accept an even better offering."
While I hold with Ben about the risks, I suspect that these companies will plow alee with their own custom silicon solutions anyway. Their desire to control their own destiny even at the chip level, and use it to differentiate, seems besides of import to their future for them to trust whatever given supplier to run across their futuristic vision with off-the-shelf processors.
Near Tim Bajarin
Source: https://sea.pcmag.com/opinion/28932/teslas-new-ai-chip-is-part-of-a-bigger-plan
Posted by: hermanrompairs.blogspot.com
0 Response to "Tesla's New AI Chip Is Part of a Bigger Plan"
Post a Comment