Search
Clients like NVIDIA and AMD Urge HBM Suppliers to Bolster Thermal Control
2026-06-08 /
Read about 0 minute
Author：小编
NVIDIA's newest AI platform, Vera Rubin, has commenced mass production. The rivalry among SK Hynix, Samsung, and Micron has pivoted from the quantity of stacked layers to technological advancements, with the thermal management of internal chips emerging as a pivotal factor in the HBM5 era. As AI hardware undergoes rapid iteration, the power consumption per chip of NVIDIA's and AMD's latest-generation AI server GPUs is nearing 1000W. HBM4 boasts 12 to 16 stacked layers, whereas HBM5 will boast up to 20 layers.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
