Oriental Computing’s DF1000 is pitched as a 14nm chip with Hopper-like utility
teortaxesTex · x · 2026-07-23
- A post and quoted article about Oriental Computing argues its DF1000 chip may offer Hopper-class utility on 14nm despite running hot.
- The image claims 6.4 TB/s bandwidth today and speculates about 20 TB/s by 2027, comparing its bandwidth-per-watt favorably with NVIDIA parts.
- The write-up says the chip uses wafer-level DRAM-Logic hybrid bonding, cutting interconnect distance by over 90% versus traditional 2.5D+HBM approaches.
- It also highlights a heterogeneous rack-scale system shown with MetaX, plus an open-sourced programming stack called CAAP built around a three-level dataflow paradigm.
- The image quotes company CEO Wei Shaojun and notes his long involvement in China’s domestic chip industry and national chip programs.
Related event: Domestic DF1000 Chip Claimed to Rival Hopper Performance(3 posts)→
More from Companies & People
- A long-term AGI roadmap puts agents first, then continual learning, then embodied AI — teortaxesTex · 2026-07-23
- DeepSeek says open source won’t hurt revenue if models recoup costs in 10 months — teortaxesTex · 2026-07-23
- ByteDance-linked discussion says AI ARR could hit hundreds of millions, maybe $1B — teortaxesTex · 2026-07-23
- A model router stops being neutral once incentives push specific models — amasad · 2026-07-23
- Frank Hutter to appear on *The Information Bottleneck* after Prior Labs’ €1B SAP deal — ziv_ravid · 2026-07-23
- Google’s Anthropic bet is a strange way to fund its future rival — bindureddy · 2026-07-23