Unverified50% confidenceFactExact time
On the RTX 4070 Ti SUPER, NVFP4 runs via the weight-only Marlin kernel, with weights kept in low-precision storage but the actual computation still performed at 16 bits.
1
Sources
50%
Confidence
Long-term
Relevance
10/18/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOn a single RTX PRO 6000, peak VRAM for Qwen3.8-27B-NVFP4 dropped from 72.9 GB to 40.2 GB, and per-step time was reduced by approximately 11%73% similarUnverifiedThe RTX 4070 Ti SUPER uses the Ada architecture, has 16GB of VRAM, and has no FP4 compute units.71% similarUnverifiedThe RTX 3090 cannot run the NVFP4 quantization format because its Ampere architecture Tensor Cores do not support NVFP4 instructions, leaving only the Q4 format available, with speeds of 40–48 tokens/second.68% similarUnverified7900XT搭载Navi 31核心，拥有96个计算单元、20GB GDDR6显存，理论FP16算力超过80 TFLOPS67% similarVerifiedRTX 4090的理论FP16算力为82.6 TFLOPS，显存带宽为1008 GB/s67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1057655API
curl https://kongchang.com/api/v1/knowledge/claims/1057655MCP
get_claim(id=1057655)