NVIDIA Thor chip may experience further delay! Is Xiaopeng Motors considering shelving its installation? New forces are accelerating the development of self-developed chips
嫦娥的情人矩
发表于 昨天 19:21
1139
0
0
Nvidia Thor chip mass production may be delayed again.
Recently, according to 36Kr, the Nvidia Thor chip was originally planned for mass production in mid-2024, but has now been significantly postponed to "expected to be launched in mid-2024, and only for the entry-level version". In addition, the report cited sources close to Nvidia stating that Thor may have encountered issues with its chip architecture. Nvidia has not publicly responded to this.
It is understood that Thor chips have been upgraded to the latest Blackwell GPU architecture, specifically designed for workloads such as Transformer, LLM, and AIGC. However, since the beginning of this year, the mass production of Blackwell chips has encountered many difficulties, which may lead to a significant delay in Blackwell's shipment plan and further affect the production progress of Thor chips. Some argue that Thor's continuous delays are increasing the risk of losing its core customers.
The planned mass production time has been delayed multiple times
According to public information, the Thor chip was unveiled by Nvidia at the 2022 GTC conference. The chip has a computing power of 2000 TFLOPS, which is twice that of Altan and eight times that of Orin. The original plan was to achieve mass production in 2024 and get on board in 2025, and Jike was one of the first customers to use Thor. In addition to Extreme Krypton, Ideal Auto, BYD, and Xiaopeng Motors are also customers of Nvidia Thor.
It is worth noting that after the launch of the Xiaopeng P7+this year, the product managers of the Xiaopeng Motors P and G series stated in an interview that the Xiaopeng P7+was originally planned to directly use the Nvidia Thor chip during development, but due to the continuous delay of this chip, it ultimately adopted a dual Orin-X configuration.
In fact, Thor chips have been rumored to have been delayed multiple times before. It is reported that Thor chips were originally planned for mass production in mid-2024, but were soon revised to be mass-produced and delivered in April May 2025. However, after a new round of changes, the latest news indicates that Thor is expected to be mass-produced and delivered by mid-2025.
The product managers of Xiaopeng Motors' P and G series said on social media in November, "Thor is still delayed and there is no definite SOP (Standard Operating Procedure) time yet. It seems that meeting in 2026 would be good
In addition, the supply of Nvidia chips is also affected by other factors. According to CCTV News on December 2nd, the United States has issued new semiconductor export control measures to China, further tightening export controls on semiconductor manufacturing equipment, storage chips, and other items to China, and adding 136 Chinese entities to the list of export control entities.
On December 3rd, the China Association of Automobile Manufacturers (CAAM) issued a statement stating that "the arbitrary modification of regulatory rules by the US government has seriously affected the stable supply of American chip products. The trust and confidence of the Chinese automotive industry in purchasing American chip products are being shaken, and American automotive chip products are no longer reliable and safe. To ensure the security and stability of the automotive industry chain and supply chain, the association recommends that Chinese automotive companies purchase American chips with caution.
In a related research report, CITIC Securities stated that in the long run, China needs to strengthen independent research and development, which may accelerate the localization process of the semiconductor industry.
Weixiaoli accelerates self-developed chips
Against the backdrop of the delay in Thor chips, the pace of new product launches by many car companies has also been affected. The above report states that Xiaopeng Motors is considering shelving plans to use Thor chips for its new car next year. Meanwhile, the report cited informed sources as saying that Xiaopeng Motors is accelerating the installation of its self-developed intelligent driving chip "Turing".
The reporter from the Daily Economic News learned that the Turing chip of Xiaopeng Motors was successfully taped out in August this year. In October of this year, the Turing chip has successfully implemented the latest version of intelligent driving functions, completing up to 2791 functional verifications in 40 days. As for when to install the new product in the new car, the reporter asked Xiaopeng Motors for confirmation, and they said that they have not yet announced the corresponding plan.
Similar to Xiaopeng Motors, NIO also completed the chip fabrication of its self-developed chips this year. In July of this year, NIO Chairman Li Bin announced at the "2024 NIO Innovation and Technology Day" that the NIO "Shenji NX9031" car grade 5-nanometer intelligent driving chip has been successfully taped out and will be installed on NIO's flagship model ET9.
It is worth mentioning that when Nvidia unveiled the Orin, NIO became one of the early car companies to use the chip. As for the recent Thor chip, there is currently no news of cooperation between Nvidia and NIO. The above report also stated that NIO did not pre order Nvidia's next-generation chip Thor next year, and that the new car next year will not be equipped with Thor chips.
In addition to Xiaopeng Motors and NIO, Ideal Motors is also developing its own intelligent driving chips. The reporter learned that the intelligent driving chip code of Ideal Automobile is "Schumacher", with a design process of 5nm, manufactured by TSMC, and is expected to complete the chip fabrication work by the end of 2024. Previously, there were reports that Ideal Auto plans to establish a chip research and development office in Hong Kong and is recruiting AI chip architects at the 18-19 level.
From the layout of the above-mentioned enterprises, self-developed intelligent driving chips have become the choice of many companies. Whether it is the delay in the supply of smart driving chips or the China Association of Automobile Manufacturers' suggestion to purchase American chips cautiously, both have accelerated the pace of enterprises developing their own smart driving chips.
Self developed chips need to balance input-output ratio
It is worth noting that there have been incidents such as stock shortages and sanctions in the chip industry before, which can have a significant impact on the business development of enterprises. Host manufacturers with intelligent driving as their core have realized the strategic significance of self-developed chips and have begun to promote their own research plans. A research and development personnel from a car company told reporters that one of the reasons for car companies to independently develop chips is due to supply chain considerations.
If the supply of intelligent driving chips is tight, it will affect the research and production plans of car manufacturers, "said Jiang Han, a senior researcher at Pangu Think Tank, in an interview with reporters. This situation will accelerate the reshuffle of the automotive intelligent driving chip industry and promote its development towards a more autonomous and controllable direction.
In addition, cost reduction has become another important reason for companies to choose self-developed chips. Li Bin believes that in the long run, self-developed chips can achieve technological cost reduction. Previously, Li Bin stated in an interview with reporters that he purchased a large number of Nvidia chips last year, which cost the company a lot of funds. Therefore, the company turned to self-developed chips, and the function of one chip can replace four, which can reduce costs.
In terms of system research and development, car companies need to pay attention to how the cost of the platform continues to decrease. In terms of the hardware cost of intelligent driving (including sensors and computing platforms such as radar and cameras), Tesla is $1500, and Ideal is $4000. "Li Xiang, chairman of Ideal Motors, once said," If Ideal makes its own inference chip, it can achieve the same cost as Tesla because the algorithm is in its own hands, including the entire training platform and chip behind it
However, it should be noted that car companies invest heavily in self-developed chips. The input-output ratio will be a heavy topic that these car companies will have to face. According to relevant research reports, taking the high-performance SoC with 7nm process and 100+TOPS as an example, its research and development cost is over 100 million US dollars. From the perspective of economic considerations for car companies, we believe that if the shipment volume of self-developed chips is less than 1 million, it may be difficult to achieve a balanced input-output ratio, "said Liu Yudong, Executive General Manager of Chentao Capital.
Xinmou Research Consultant Wang Xiaolong believes that domestic chips have made rapid progress in technical indicators, but there is still a gap in reliability and production consistency between some chip products and foreign manufacturers, which requires experience accumulation and time polishing.
CandyLake.com 系信息发布平台,仅提供信息存储空间服务。
声明:该文观点仅代表作者本人,本文不代表CandyLake.com立场,且不构成建议,请谨慎对待。
声明:该文观点仅代表作者本人,本文不代表CandyLake.com立场,且不构成建议,请谨慎对待。
猜你喜欢
- Nvidia's new generation AI chip GB200 orders explode, H100 chip hits cold
- ParTec files patent infringement lawsuit against Nvidia in Munich
- Nvidia's intraday market value surpasses Apple again, and the battle for the top spot in the US stock market is becoming increasingly fierce
- How will 'Trump 2.0' affect Nvidia? Wall Street consensus: More is good!
- Huang Renxun makes a major announcement! Nvidia and SoftBank collaborate, SoftBank accelerates AI layout
- Wall Street raises Nvidia target price one after another: Blackwell craze is coming!
- The rise and fall of new car making forces in the US stock market were mixed before the market opened
- Germany heavily subsidizes chip industry, Amazon challenges Nvidia
- Mingchuang Youpin CMO Liu Xiaobin: How to build a "super brand" that moves towards the world | Directly targeting the Consumer New Forces Conference
- Nvidia Tmall flagship store clears and removes all products? Insiders: Tmall flagship store has always been used only for displaying products and does not directly sell them. There is no such thing as "delisting"
-
生成式人工知能(AI)が巻き起こす技術の波の中で、電力会社は意外にも資本市場の寵児になった。 今年のスタンダード500割株の上昇幅ランキングでは、Vistraなどの従来の電力会社が注目を集め、株価が2倍になってリ ...
- xifangczy
- 昨天 12:14
- 支持
- 反对
- 回复
- 收藏
-
量子計算会社は年内に狂った。 現地時間12月17日、米株3大指数は下落した。ダウ平均は9営業日連続で下落し、1978年以来の最長連続下落を記録した。 人気のある株では、テスラとアップルの株価が再び高値を更新した ...
- SOHU
- 3 天前
- 支持
- 反对
- 回复
- 收藏
-
隔夜株式市場 世界の主要指数は金曜日に多くが下落し、最新のインフレデータが減速の兆しを示したおかげで、米株3大指数は大幅に回復し、いずれも1%超上昇した。 金曜日に発表されたデータによると、米国の11月のPC ...
- SNT
- 26 分钟前
- 支持
- 反对
- 回复
- 收藏
-
長年にわたって、昔の消金大手の捷信消金の再編がようやく地に着いた。 天津銀行の発表によると、同行は京東傘下の2社、対外貿易信託などと捷信消金再編に参加する。再編が完了すると、京東の持ち株比率は65%に達し ...
- SNT
- 1 小时前
- 支持
- 反对
- 回复
- 收藏