AI inference accelerators are moving to 3D DRAM stacked directly on compute for higher bandwidth and lower energy. Will this become the standard architecture?
#DRAM #3DDRAM #HBM $NVDA $QCOM
chippub.substack.com/p/why-is-eve...