Unverified50% confidenceFactExact time
AntAngelMed's BF16 full-precision inference requires 8 Ascend 910B (64GB) cards or 4 Kunlun P800/PPU 810 (96GB) cards.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/14/2026
First Seen
Valid until: 1/12/2027
Sources
Related Entities
Related Claims
UnverifiedMeta Llama 3的405B参数版本以BF16精度加载需约810GB内存，量化至INT4精度需约200GB72% similarUnverifiedUD-Q4_K_XL量化版本体积155GB，Top-1精度为96%69% similarVerifiedAt a sequence length of 4096, the N×N attention matrix alone consumes approximately 64 MB of memory in FP16 precision.68% similarUnverifiedIn benchmark testing on an AMD RX 7900 GPU, Ornith 35B achieved an overall accuracy of 85% (18 out of 21 passed), a generation speed of 74 tokens/s, a prompt processing speed of 850 tokens/s, and a VRAM usage of 20 GB.67% similarVerifiedTaking a 70B parameter model with FP16 precision and 32 attention head layers as an example, the KV Cache for a single 200K-length request requires approximately 80–120 GB of VRAM, far exceeding the 80 GB VRAM limit of a single A100 GPU.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1032160API
curl https://kongchang.com/api/v1/knowledge/claims/1032160MCP
get_claim(id=1032160)