Home >Technology peripherals >AI >The next big thing in AI: Peak performance of NVIDIA B100 chip and OpenAI GPT-5 model
After the debut of the NVIDIA H200, known as the world's most powerful AI chip, the industry began to look forward to NVIDIA's more powerful B100 chip. At the same time, OpenAI, the most popular start-up AI company this year, has begun to develop a more powerful and complex GPT-5 model. .
Guotai Junan pointed out in the latest research report that B100 and GPT5 with boundless performance are expected to be released in 2024, which will be a major upgrade or release unprecedented productivity. The agency stated that it is optimistic that AI will enter a period of rapid development and its visibility will continue until 2024.
Compared with previous generations of products, how powerful are B100 and GPT-5? Nvidia and OpenAI have already given a preview: B100 may be more than 4 times faster than H100, and GPT-5 may achieve super intelligence.
In the latest special presentation, NVIDIA conducted a performance comparison of B100 and A100, H100 and H200 running GPT-3 175B, and came to some surprising conclusions
The inference speed of H200 is 18 times that of A100 released in 2020, and the B100 based on the next-generation Blackwell architecture. The chart provided by NVIDIA does not show specific numbers, but the technology website Tweaktown will compare the columns of B100 and H100 Figure, found that the speed of B100 is at least 3 times faster than H100. Translate this content into Chinese: The H200 has 18 times the inference speed of the A100 released in 2020, and the B100 is based on the next-generation Blackwell architecture. Although Nvidia did not provide specific numbers, the technology website Tweaktown found by comparing the histograms of the B100 and H100 , B100 is at least 3 times faster than H100
Regarding configuration, according to Tweaktow’s article, it is reported that B100 will have 178 billion transistors embedded inside, which is more than twice the number of current GPU transistors, posing a challenge to the limits of chip integration
B100 will use Micron’s latest HBM3e memory, which is the same as the just-launched H200. In addition, the H200 supports up to 141GB of HBM3e memory with a memory bandwidth of up to 4.8TB/second, so the outside world can expect a similar memory configuration for the B100.
According to reports, the B100 will use an in-house chiplet (small chip) design, which is Nvidia’s first attempt to compete with AMD’s upcoming Instinct MI300
Is GPT not far away from super intelligence?
Attention turns to OpenAI, a high-profile start-up artificial intelligence company that is launching a more powerful and complex GPT-5 model
OpenAI CEO Sam Altman confirmed this in a recent media interview, and claimed that it can achieve superintelligence.
However, Altman has not yet determined a release schedule for GPT-5, nor its possible features, as there is no way to know until it is complete
GPT-4’s capabilities are impressive enough, but it is only close to humans in terms of knowledge and understanding. The next generation of artificial intelligence models will not only surpass humans in knowledge reserves, but also in reasoning and processing complex ideas. Comparable to humans in ability.
This is called general artificial intelligence (AGI) in the industry. It not only mechanically imitates what is given, but also has the ability to create and express novel content
Then the question is, how much can GPT-5 do? Can it help greatly shorten the distance between humans and AGI?
Altman also said that the construction of GPT-5 will require further investment from Microsoft to become a reality.
Microsoft has invested more than $10 billion in OpenAI. Altman expressed the hope that more funds will follow up, because the road to AGI is still long and requires the construction of a large amount of computing power
The above is the detailed content of The next big thing in AI: Peak performance of NVIDIA B100 chip and OpenAI GPT-5 model. For more information, please follow other related articles on the PHP Chinese website!