Head to head

RTX 3090 vs Radeon RX 7900 XTX

Both hold 24 GB, so capacity is a wash. The Radeon RX 7900 XTX has more bandwidth at 960 GB/s, which decides how fast tokens come out once a model fits.

Model (Q4_K_M, 8K)RTX 3090tok/sRadeon RX 7900 XTXtok/s
Llama 3.1 8B Runs comfortably 112 Runs comfortably 105
Llama 3.1 70B Won't fit 1.67 Won't fit 1.52
Llama 3.2 3B Runs comfortably 246 Runs comfortably 230
Llama 3.2 1B Runs comfortably 650 Runs comfortably 609
Llama 4 Scout 109B-A17B Won't fit 5.39 Won't fit 4.91
Qwen3 8B Runs comfortably 108 Runs comfortably 101
Qwen3 14B Runs comfortably 63.5 Runs comfortably 59.3
Qwen3 32B Fits, but tight 29.5 Fits, but tight 27.6
Qwen3 4B Runs comfortably 196 Runs comfortably 183
Qwen3 30B-A3B Runs comfortably 117 Runs comfortably 112
Qwen3 235B-A22B Won't fit 3.36 Won't fit 3.07
Qwen2.5-Coder 32B Fits, but tight 29.5 Fits, but tight 27.6
Qwen2.5 7B Runs comfortably 125 Runs comfortably 116
Qwen2.5 72B Won't fit 1.59 Won't fit 1.44
Gemma 3 4B Runs comfortably 209 Runs comfortably 196
Gemma 3 12B Runs comfortably 76.1 Runs comfortably 71.1

Specifications

RTX 3090Radeon RX 7900 XTX
Memory24 GB24 GB
Bandwidth936 GB/s960 GB/s
FP16 compute71 TFLOPS123 TFLOPS
Launch price$1,499$999
ArchitectureAmpereRDNA3