Following the scarcity of graphics processing units (GPUs) that plagued the industry until late 2022, the introduction of AI applications has raised new issues as a result of the shifting market and AI boom, resulting in an unanticipated rise in demand for gaming cards.
Data center cards were once thought to be the key target for AI systems, however, a lack of AI GPUs has forced some to turn to gaming cards instead. George Hotz, a software developer, reported a large bulk purchase of AMD Radeon cards for AI training. This news drew notice on Twitter, where Hotz posted a snapshot of a Radeon RX 7900 XTX amidst a slew of unopened GPU boxes, signifying 7.38 PFLOPS of processing power with plans to grow up to the exaFLOP level.
Analysts estimated the computing capacity of these devices, which cost roughly $60,000 in total. Hotz’s ideas might necessitate a substantial expenditure ranging from $20 million to $60 million, depending on whether he seeks 2 exaFLOPs or almost 8 exaFLOPs of compute.
This mass acquisition not only highlights AMD’s bargaining tactic of selling straight to developers, but it also threatens Nvidia’s AI dominance. Companies’ willingness to invest in AMD’s products, even outside of their Pro cards, indicates a substantial shift in the GPU industry.
The sources for this piece include an article in ExtremeTech.