Not available in English yet
微软推出 5995 美元 AI 开发机 — 大厂正式下场做本地大模型工作站
Related Reading
More on #NVIDIA
RTX Spark Rumored to Launch Next Week: A $20K Workstation for Local 70B LLMs
RTX Spark rumored Oct 7: GB10 chip, 24-128GB VRAM, $1,800-$2,900. NVIDIA is bringing 70B-model compute from data centers to ~$20K workstations.
NVIDIA Robots Assemble GB300 — Even Screw-Tightening Enters the AI Spotlight
NVIDIA reveals robots now assemble GB300 test trays—once done by skilled workers. Even AI firms must solve hardware factory automation.
NVIDIA's cuPhoton Brings GPU to the Observatory — Tightening the CUDA Moat
NVIDIA launches cuPhoton, moving imaging for telescopes, lasers, X-ray sources to GPU and extending CUDA into research infrastructure.
NVIDIA's One Nemotron Hits IOI, IMO Gold — Olympiad Tuning Is LLMs' New Arena
NVIDIA tuned one Nemotron base to gold at IOI and IMO. One stack, two golds: methodology now beats parameter-counting in reasoning races.
Used Mining GPU Beats DGX Spark 5x — Open Source Breaks NVIDIA's AI Pricing
Used mining GPU with open-source tool unlocks 1.5TB/s bandwidth — 5x DGX Spark at half the price. NVIDIA's AI compute pricing narrative just cracked.
NVIDIA Adds Direct GPU Networking — AI's Bottleneck Shifts to Communication
NVIDIA's DOCA GPUNetIO lets GPUs launch network ops directly, skipping the CPU. AI's next bottleneck: GPU-to-GPU communication, not compute.