Similar artificial intelligence systems, especially ChatGPT, require a large processing power. While a large amount of GPU power is needed for these models, Nvidia, one of the biggest players in the artificial intelligence market, according to Trendforce, comes to the fore to meet the demand.Previous estimates put the number of GPUs powering the GPT (Generative Pre-Trained Transformer) model around 10-20k, but with the new model OpenAI is expected to use much more than that. According to the report, the expected number of GPUs for the final model to power ChatGPT is 30,000, but it looks like we could see even more demand.
The boom in demand for ChatGPT is a boon for Nvidia
According to TrendForce, the number of parameters used in training GPT models in 2018 was 120 million. This number has reached a huge level of 180 billion in 2020. In the report shared by TrendForce, “The number of GPUs required by the GPT model to process training data reached 20,000 in 2020. In the future, the number of GPUs needed for commercialization of the GPT model (or ChatGPT) is estimated to exceed 30,000.” says.
According to the statements, at the moment, Nvidia’s A100 GPU and DGX A100 platform, which can provide performance capacity at 5 petaFLOPS level, are mostly used for artificial intelligence. Therefore, if a strong competitor for Nvidia does not come by AMD or any other company, the green team will continue to be the biggest player in the artificial intelligence industry.On the other hand, Nvidia CEO Jensen Huang has praised OpenAI’s ChatGPT tool many times, both during interviews and during meetings for the company. The company has rolled out RTX Video Super Resolution, which uses the artificial intelligence of consumer GPUs to improve video and streaming quality across multiple applications. As a result, the demand explosion for ChatGPT in this period when video card sales are experiencing serious declines seems to be a great boon for Nvidia.
34662
