December 3, 2022

Even if Nvidia didn’t point out AMD, it sounds as if to be clarifying that it’s NIO’s spouse in high-performance computing chips, now not its competitor.


US chipmaking large Nvidia (NASDAQ: NVDA) highlighted its partnership with NIO (NYSE: NIO, HKG: 9866, SGX: NIO) in what seems to be an oblique reaction to a drama remaining week about its rival AMD (NASDAQ: AMD) operating with the electrical car (EV) maker.

NIO is the usage of the NVIDIA HGX A100 to construct a complete information heart infrastructure on which to expand AI-powered, software-defined automobiles, Nvidia China stated in a put up on its legit WeChat account Thursday.

With high-performance computing, NIO can iterate on advanced deep studying fashions to construct robust independent riding algorithms in a closed-loop atmosphere, the put up stated.

Nvidia’s high-performance computing answers have change into NIO’s gasoline pedal at the highway to independent riding, stated Bai Yuli, head of NIO’s AI platform, as quoted within the put up.

NIO has introduced a number of cars in response to the infrastructure, together with its flagship sedan, the ET7, and the ET5, which can move on sale in September, Nvidia stated.

Each fashions are in response to high-performance information facilities and have the Adam supercomputing platform in response to 4 NVIDIA DRIVE Orin system-on-chips, the put up stated.

Their self-driving and clever cockpit functions may also be iterated and stepped forward in the course of the information heart to ship a disruptive buyer enjoy, Nvidia stated.

Whilst the Nvidia article does now not comprise any point out of its rival AMD, it sounds as if to elucidate that it’s NIO’s spouse in high-performance computing chips, now not AMD.

See also  China mulls extending EV subsidies because of expire at year-end, file says

On June 6, a video posted by means of AMD China on Weibo stated that NIO will use AMD’s EPYC circle of relatives of processors in its HPC (high-performance computing) platform, which is predicted to hurry up AI deep studying coaching and shorten product building cycles.

AMD’s video therefore generated standard consideration in addition to numerous media protection in China.

Alternatively, on June 8, Ma Lin, senior director of company communications at NIO, stated whilst sharing AMD’s video that NIO had no cooperation with AMD and had now not licensed AMD to hold out this advertising marketing campaign.

Within the feedback phase of his Weibo put up, Ma implied that NIO had handiest bought servers that used AMD chips.

Alternatively, Ma then briefly softened his phrases and stated his earlier remark used to be too sturdy, announcing AMD is an excessively first rate corporate.

At the night time of June 8, AMD China issued a rationalization, announcing that NIO had bought servers for HPC building that used AMD EPYC processors in response to the Zen 3 structure. AMD appears ahead to cooperation between the 2 events sooner or later, it stated.

Again at Nvidia, it stated NIO’s scalable AI infrastructure is powered by means of NVIDIA HGX, with a complete of 8 A100 Tensor Core GPUs and NVIDIA ConnectX-6 InfiniBand NICs.

This scalable supercomputer cluster incorporates a collection of NVME SSD servers interconnected by means of a high-speed NVIDIA Quantum InfiniBand community platform, consistent with Nvidia.

This robust infrastructure has an ultra-high switch price of 200 Gbps to switch huge quantities of deep studying coaching information to supercomputer reminiscence or NVIDIA A100 graphics reminiscence, it stated.

See also  How Are Conventional Chinese language Carmakers Like Geely Outperforming NIO, XPeng and Li Auto?

The NVIDIA HGX A100 is a high-performance server platform designed for AI eventualities and incorporates many huge information units and sophisticated fashions, akin to the ones used to force self-driving automobiles, consistent with the Nvidia put up.

With the HGX A100, the NIO independent riding building platform (NADP) allows versatile building and deployment of scalable AI techniques, changing more than one siloed infrastructures with a one-stop platform to regulate huge numbers of advanced AI programs, consistent with Nvidia.

This allows NIO to beef up type building potency by means of 20 instances, which in flip shortens the time-to-market cycle for self-driving automobiles and develops more moderen, quicker architectures, consistent with Nvidia.

NIO has introduced plans to double the capability of its Hefei facility to 240,000 cars in step with yr, with a most general manufacturing capability of 300,000 cars, consistent with the Nvidia put up.

As NIO’s capability expands and it continues to amplify in world markets, NVIDIA HGX may even amplify to lend a hand NIO deploy the automobile trade’s maximum complicated AI platform, Nvidia stated.


This text used to be first revealed by means of Phate Zhang on CnEVPost, a site specializing in new power car information from China.