Unverified50% confidenceFactExact time
M系列芯片在20-40瓦范围内即可完成相当水平的推理任务，而NVIDIA GPU动辄数百瓦功耗
1
Sources
50%
Confidence
Long-term
Relevance
8/27/2026
First Seen
Sources
Related Entities
Related Claims
Unverified英伟达H100 GPU单卡功耗高达700瓦，一个配备5万块H100的超算集群总功耗可达35兆瓦（MW）77% similarVerifiedConsumer-grade GPUs such as the RTX 4090 have 24 GB of VRAM and can run quantized models with billions of parameters.76% similarUnverifiedDeploying an open-source model with 70 billion parameters running at FP16 precision requires at least four A100/H100 GPUs with 80 GB of VRAM each, with hardware procurement costs alone exceeding one million yuan.73% similarUnverifiedA tester ran a 20.7B dense model on a dual-4090 (48 GB VRAM) workstation and achieved 27.6 tokens per second, only slightly faster than Dwarf Star, but with over 10 times fewer parameters.69% similarUnverifiedTo host GPT-5.6 Sol with 2 trillion parameters, only 14 B300 chips are needed with NVIDIA, whereas 90 WSE3 chips are required with Cerebras.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/810128API
curl https://kongchang.com/api/v1/knowledge/claims/810128MCP
get_claim(id=810128)