DFSX (Dongfang Suanxin) just demonstrated a TY64 SuperNode with 960TB/s memory bandwidth - that's 2x NVIDIA's GB200 NVL72 system.
And they're doing it on a mature 14nm process using 3D near-memory compute architecture.
The key insight: GPUs sit idle most of the time waiting for data from memory.
DFSX stacks memory directly on top of compute layers, eliminating the bottleneck.
At BHK Cloud, we provide both sides of the AI equation: - RTX 3090 GPU compute at $0.15/hr - Cloud storage at $2.49/TB - Buy Now Pay Later - zero upfront, pay after 1 month https://ai.bhkcloud.com/?utm_source=reddit&utm_medium=social&utm_campaign=bhk_social_2026w32&utm_content=morning
Spin up an RTX 3090 in 60 seconds. Storage at $2.49/TB. Zero egress between GPU and storage.
Try BHK Cloud free
Originally posted on reddit