NVIDIA GTC 2024 Keynote Speech: After three years, it returns offline, scheduled for March 18, 2024

NVIDIA announced that the new "Leather Jacket" is ready for their annual GTC Developer Conference from March 18th to 21st next year at the San Jose Convention Center in California, USA. The content collection has now ended

Taking into account the time difference and NVIDIA's practice, Beijing time is likely to be 23:00 on the evening of March 18th. It is worth mentioning that NVIDIA has been holding online digital conferences since 2020, so GTC 2024 will represent NVIDIA's return to San Jose.

According to reports, this live conference will bring together thousands of innovators , researchers, creators, thought leaders and policymakers, who will come together to share ideas, discuss innovation strategies, and explore how accelerated computing and AI are changing the way we work and live.

NVIDIA will release its latest generation graphics technology at GTC as usual, and will also launch new data center products

Considering the latest revelations, everyone at this GTC conference Attention will be focused on the highly anticipated Blackwell series of products. At the same time, this series of products will also be expanded to the gaming field, although the underlying architecture of the gaming chip may be different

Notes from this site: Blackwell is David Blackwell, an academician of the National Academy of Sciences, who is the first person of the National Academy of Sciences A black academician and the first black tenured professor at the University of California, Berkeley, died of illness on July 8, 2010 at the age of 91.

Image source Pexels

It is worth mentioning that Kopite recently shared information about Nvidia’s next-generation GeForce and data center GPUs Related information

Codenamed Blackwell product line, GB20X is mainly for consumer-grade GeForce RTX graphics cards, while GB10X is mainly for data centers.

##GB100

Source Kopite stated that GB100 has a total of 8 Each group of GPCs contains 10 TPCs. It is estimated that each TPC contains 2 SMs, which is equivalent to 160 SMs.

There are 20 SMs in each GPC. Assuming that each SM can accommodate 128 CUDA cores, then GB100 may contain 20480 CUDA/FP32 cores

GB202

Message Source Kopite indicates that GB202 has a total of 12 groups of GPCs. Each group of GPCs contains 8 TPCs, and each TPC contains 2 SMs. There are a total of 192 SMs, which is equivalent to 24576 CUDA cores.

Kopite reports that GB100 and GB202 will include 8192-bit and 512-bit memory buses respectively. This site attaches the following information:

• GB100 (data center): 160 SMs | 20,480 CUDA cores | 8,192-bit memory bus What needs to be rewritten is: GB100 (data center): 160 SMs | 20480 CUDA cores | 8192-bit memory bus

• GB202 (game): 192 SMs | 24576 CUDA Core | 48 GB G7 Memory

Since the above chip is not yet in mass production, it is difficult to determine its performance. The initial leaked news that the RTX 5090 is 1.7 times faster than the RTX 4090 has not been verified

Related recommendations:

"Next-generation NVIDIA flagship graphics card RTX 5090 exposed: 3nm process, codename Blackwell"

"It is reported that NVIDIA has tested Samsung's 3nm GAA process, and mass production will be as soon as 2025"

"It is reported that NVIDIA RTX 5090 graphics card has 50% more CUDA cores than the previous generation, and its overall performance has improved by 70% %》

Advertising statement: The external jump links contained in the article (including but not limited to hyperlinks, QR codes, passwords, etc.) are used to convey more information and save selection time. The results are for reference only. , all articles on this site contain this statement.

The above is the detailed content of NVIDIA GTC 2024 Keynote Speech: After three years, it returns offline, scheduled for March 18, 2024. For more information, please follow other related articles on the PHP Chinese website!