趋近智
The AMD RX 6600 XT (8GB) provides 8 GB of memory with 256 GB/s bandwidth, supporting local inference via ROCm and Vulkan acceleration.
总显存
8 GB
可用显存上限
8 GB
显存带宽
256 GB/s
最大推理模型级别
8B Models (Q4)
热设计功耗
160W
随上下文增长模拟的显存需求。当曲线超过参考线时将导致显存不足(OOM)。
量化精度:
不同量化精度下可运行的最大模型参数级别。
上下文长度:
4位量化
8B Models
8位精度
< 3B Models
16位全精度
< 1B Models
本地模型训练与适配器微调支持能力。
微调上下文:
QLoRA
8B Models
LoRA
3B Models
全参数微调
< 1B Models
Guidance for optimal precision and operational ceilings on AMD RX 6600 XT (8GB).
Inference Sweet Spot
Recommended for lightweight models (1B to 3B) at FP16/Q8, and 7B/8B models at aggressive 4-bit quantizations.
Bandwidth & Speed Profile
With 256 GB/s aggregate bandwidth, batch size 1 inference operates in a memory-bandwidth bound regime. Generates approximately 57 tok/s on an 8B Q4 model and 7 tok/s on a 70B Q4 model.
APX AI
在线