Tuesday, May 23, 2023 at 2:00 p.m. 00 p.m. Indochina time.
“As the demand for high-performance computing and AI continues to increase, All the hardware that drives it requires more power.”
via tweaktown.com In a roadmap released today, Giga Computing, the server and enterprise division of GIGABYTE, outlines what the next generation of data center technologies from Intel, AMD and NVIDIA will look like in terms of power draw. And when it comes to cooling data center hardware such as CPUs and GPUs, this is likely to become a very tense issue by 2025.
This slide shows a significant increase in CPU power consumption from Intel and AMD, with 500W and 600W CPUs being released between 2024 and 2025 compared to mainstream CPU hardware in 2022.
PCIe-based GPUs will also see an increase in power consumption, with NVIDIA GPUs set at 500W and AMD GPUs set at 400W. The slide mentions the NVIDIA Grace CPU Superchip and Grace Hopper Superchips capable of delivering up to 1000W in SKUs.
Of course, we have a desire to demand high performance computing (HPC) that can handle complex AI workloads. which is set to drive demand for higher-performance chips and hardware. regardless of the energy costs Of course, the performance is still high. The listing is for every hardware manufacturer. It looks like the architecture is moving towards more overall power consumption. NVIDIA’s H100 Tensor Core GPU can be configured up to 700W, so this GIGABYTE slide and presentation might be accurate about what’s coming next.
Intel and AMD server-based Xeon and EPYC CPUs for data centers expected next year are also reported to increase to 600W, so this slide serves as a reminder that In addition to the raw processing power The need for enhanced high-performance computing And the rise of AI will see the same increase in power consumption. Hope all this hardware can be cooled efficiently.
Thank you news source