Expired50% confidenceFactExact time
An 8-card H20 configuration has a total video memory of 1128GB, enabling complete loading of ultra-large-scale models such as DeepSeek-V4-Flash without video memory quantization.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/16/2026
First Seen
Valid until: 10/14/2026(expired)
Sources
Related Claims
UnverifiedDeepSeek-R1 32B模型约需20GB以上显存，通常需要专业级显卡或多卡并行75% similarUnverifiedRunning a million-token context used to require 10 H20-level GPUs, whereas DeepSeek V4 now only needs 1–2.74% similarUnverifiedDeepSeek-R1 14B模型约需10-12GB显存，需要中高端独立显卡如RTX 3080/407073% similarUnverifiedUsing the V100 (16GB VRAM) as a reference, the Ornith 9B model at Q6_K requires approximately 7–8GB of VRAM, while the 35B MoE requires approximately 20–22GB and needs dual V100s.72% similarUnverifiedThe Hanguang M890 chip features 144 GB of large memory, an inter-chip interconnect bandwidth of up to 800 GB/s, and supports a 128-card super-node design.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/530681API
curl https://kongchang.com/api/v1/knowledge/claims/530681MCP
get_claim(id=530681)