Nvidia, already the market chief in offering high-end processors for generative AI use, will launch an much more highly effective chip because the demand to run giant AI fashions continues.
The company announced the provision of the GH200 tremendous chip, which Nvidia mentioned can deal with “essentially the most complicated generative AI workloads, spanning giant language fashions, recommender programs and vector databases.”
The GH200 can have the identical GPU because the H100, at the moment Nvidia’s strongest and widespread AI providing, however triple the reminiscence capability. The corporate mentioned programs operating on GH200 will begin within the second quarter of 2024.
Nvidia didn’t reveal the worth for the GH200; the H100 line at the moment sells for roughly $40,000.
Advanced AI fashions require highly effective GPUs so the system could make the computations essential to generate textual content or a photograph of a horse within the type of Banksy. Working these fashions requires a ton of processing energy, and even with Nvidia’s H100 chips, some must “break up” the fashions amongst different GPUs simply to run.
Nvidia has a close to monopoly in generative AI-capable GPUs. Cloud providers like AWS, Azure, and Google all use Nvidia’s H100 Tensor Core GPUs and tack on providers to assist shoppers get initiatives utilizing giant language fashions up and operating to distinguish themselves.
