Verified65% confidenceFactExact time
该工作流在RTX 3060 6GB显存加16GB内存配置下即可运行
3
Sources
65%
Confidence
Medium-term (~90 days)
Relevance
7/11/2026
First Seen
Valid until: 10/9/2026
Sources
低显存也能跑！ID LoRA人脸转视频工作流实测解析
redditr/comfyui7/9/2026
Related Claims
Verified24GB显存的RTX 3090/4090可以使用Unsloth微调7B-13B规模的模型80% similarUnverified在RTX 5060Ti 16GB显存环境下，LFM2.5 BF16原版17GB无法装入，使用FP8量化后可正常运行，启动后占用约14GB显存79% similarUnverifiedRTX 3060的12GB版本区别于仅有8GB显存的3060 Ti，其12GB显存能运行大多数主流AI生成模型而不触发显存溢出79% similarUnverifiedRTX 5090上Qwen3.6 27B模型预填充速度从174提升到253 token/s，显存占用约18GB79% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/485739API
curl https://kongchang.com/api/v1/knowledge/claims/485739MCP
get_claim(id=485739)