Gaudi 3, an AI processing chip designed by Intel, outpaces Nvidia's H100 GPU, offering faster acceleration for enterprise-level AI workloads.
Gaudi 3 showcases enhanced efficiency compared to its predecessor, Gaudi 2, by leveraging TSMC's 5nm process, 128GB of HBM2e, 3.7 TB/s bandwidth, and a 900W TDP.
Intel demonstrated the superior performance of Gaudi 3 through benchmarks against Nvidia's H100 GPU, with impressive speed boosts for large language models like LLAMA2-7B, LLAMA2-13B, and GPT 3-175B.
Furthermore, Intel claims Gaudi 3 is significantly more power-efficient than Nvidia's H100, offering up to 2.3 times better efficiency for AI workloads.
[
add
]
[
|
|
...
]