Industry News

240000 yuan each! NVIDIA's strongest GPU has been put into full production, and more than 50 server models have been launched

Update time : 2022-09-24 03:17:53

The latest news is that Nvidia released its H100 computing card in March, which finally started shipping, and it was also launched in October. From this point of view, NVIDIA announced that the NVIDIA H100 Tensor Core GPU was fully put into production at GTC 2022, which is also its position on the rumor that TSMC has placed a "super urgent" order.

On September 22, it was reported that at the GTC 2022 conference last night (once in the first and second half of the year respectively), NVIDIA not only released the RTX4080/4090 series graphics cards, but also announced one thing, that is, the H100 computing card released in March finally began to ship, and it was also launched in October.
In other words, Nvidia H100 Tensor Core GPU has been put into full production. NVIDIA's global technology partners plan to launch the first batch of products and services based on the groundbreaking NVIDIA Hopper architecture in October.
In fact, it is just half a year since the H100 accelerator card was released at the GTC Conference in March this year. It uses Hopper architecture, GH100 core, TSMC 4nm manufacturing process, CoWoS 2.5D packaging technology, and integrates 80 billion transistors with a core area of 814 mm2.
It is reported that it has 18432 CUDA cores, 576 Tensor cores, 60MB L2 cache, supports six HBM3/HBM2e with 6144 bit width, supports PCIe 5.0, and supports the fourth generation NVLink bus.
In addition, the H100 computing card has two styles: SXM and PCIe 5.0. In the SXM version, there are 15872 CUDA cores and 528 Tensor cores. In the PCIe 5.0 version, there are 14952 CUDA cores and 456 Tensor cores. The power consumption is up to 700W.
The key is that the H100 enables enterprises to reduce AI deployment costs. Compared with the previous generation, it can improve energy efficiency by 3.5 times, reduce the total cost of ownership to 1/3, and reduce the number of server nodes used to 1/5 when providing the same AI performance.
Fortunately, NVIDIA DGX H100 system has now begun to accept customer reservations. The system includes 8 H100 GPUs, and the peak performance of FP8 accuracy reaches 32PFlops. Each DGX system contains NVIDIA Base Command and NVIDIA AI Enterprise software, which can realize cluster deployment from a single node to NVIDIA DGX SuperPOD, and provide support for advanced AI development of large language models and other large-scale workloads.
As for the most concerned problem of the demander, there is no official information about the price of H100. However, there has been a pre-sale in the Japanese market before. The price of PCIe version is more than 4.75 million yen, and the price of SXM version is more expensive.
However, the latest news shows that the H100 accelerator card was launched in October, among which Amazon, Google and Microsoft, the three major cloud service providers, took the lead in adopting it, as well as scientific research institutions and universities, Los Alamos National Laboratory, Swiss National Supercomputing Center and Tsukuba University of Japan will also purchase it.
On the other hand, the systems equipped with H100 provided by the world's leading computer manufacturers are expected to be delivered in the next few weeks. By the end of this year, more than 50 server models will be available, and dozens of models will be available in the first half of 2023. Partners that have been building systems include Atos, Cisco, Dell Technology, Fujitsu, Gigabyte Technology, Huiyou, Lenovo and AMD.
A few days ago (September 19), Taiwan, China Economic Daily reported that NVIDIA, the global leader in GPU, recently placed a "super hot runs" order with TSMC to advance the production of some products originally planned to be shipped next year. It is rumored that this batch of "super urgent items" involves 5000 wafers, and the delivery time of related products will be significantly shortened from the original estimated 5-6 months to 2-3 months. TSMC will start delivering to Nvidia at the end of October to the beginning of November as soon as possible.
From this point of view, NVIDIA's announcement on GTC 2022 that the NVIDIA H100 Sensor Core GPU was fully put into production is also a "statement" of the above rumors.
Why are you in such a hurry? At the end of August, the US ordered NVIDIA and AMD to stop selling some high-performance GPUs to Chinese Mainland, Hong Kong and Russia, including NVIDIA's A100 and H100.

In this regard, NVIDIA has actively mediated with the relevant departments in the United States. Soon on September 1, they announced that they had obtained the approval of the United States government and could continue to provide A100 products to American customers (to China) before March next year, and could continue to fulfill orders for A100 and H100 before September next year.
In general, considering that China is one of the most important markets for NVIDIA and AMD, the US side decided to grant a grace period of up to one year for the ban. In order to cope with the uncertainty of the market and export control policies, NVIDIA also wants to protect customers' long-term demand as much as possible in this year. Therefore, it places orders for "super urgent" goods to TSMC, and produces the quantity of goods shipped next year early to meet the market customers' demand for "hoarding".

Previous : The Total Valuation of Qualcomm's Automotive Business Orders Grew to $30 billion

Next : Nvidia released Thor, an intelligent car chip, which meet six demands such as automatic driving

How many chips does a car need?

Sep .19.2024

Automotive chips can be divided into four types according to their functions: control (MCU and AI chips), power, sensors, and others (such as memory). The market is monopolized by international giants. The automotive chips people often talk about refer to

Position and Function of Main Automotive Sensors

Sep .18.2024

The function of the air flow sensor is to convert the amount of air inhaled into the engine into an electrical signal and provide it to the electronic control unit (ECU). It is the main basis for determining the basic fuel injection volume. Vane type: The

Chip: The increasingly intelligent electronic brain

Sep .14.2024

In this era of rapid technological development, we often marvel at how mobile phones can run various application software smoothly, how online classes can be free of lag and achieve zero latency, and how the functions of electronic devices are becoming mo

LDA100 Optocoupler: Outstanding Performance, Wide Applications

Sep .13.2024

In terms of characteristics, LDA100 is outstanding. It offers AC and DC input versions for optional selection, enabling it to work stably in different power supply environments. The small 6-pin DIP package not only saves space but also facilitates install

About Us

About Us What we do Quality People

Products

CHIP/IC Active devices Passive device Components OTHER

China electronic components supplier----Wah Luen Shing Electronic.----www.wlschip.com

Leave a message