ChatGPT Causing Shortage of GPU!

Feb 16

GPU graphics products are now at risk of shortage price hikes due to AI, especially the fire-generative conversation robot ChatGPT.

According to media reports, major DRAM manufacturers like Samsung and SK Hynix have increased their offers for graphics memory chip products, advanced HBM3, by as much as 5 times, knowing that the average price of HBM would have been 3 times that of ordinary DRAM because the production process and technology are more complex.

Video memory prices quadrupled! GPU shortage again: graphics cards restart
HBM3 is utilised in NVIDIA's H100 compute card and Intel's Xeon Max x86 processor.

Video memory prices quadrupled! GPU shortages: graphics cards rise.
Another underlying concern is that the industry's AI follow-through, including machine learning training, may cause another GPU chip scarcity. Developers employ 10,000 Nvidia GPUs to train the massive ChatGPT AI analysis data database.

OpenAI has employed around 25,000 NVIDIA GPUs to handle server demand, with more expected as demand rises. ChatGPT requires 13.5 EPLOPS of computing power, which is supported by 69,000 NVIDIA DGX A100 80G servers, and the existing global computer capacity can only handle 100 million average daily online users.

Some experts estimate that Google Search will need 512820 A100 HGX servers and over 4.1 million A100 GPUs to completely incorporate Bard platform AI technologies.

Source: Techradar

ChatGPT is helping a lot of industries and students but also a lot problem in terms of education and GPU shortage as well. What do you think about ChatGPT? Is it doing more good than harm?

All comments (1)
No contentNothing here, please try again later.