Head to head

RTX 4070 Ti Super vs NVIDIA DGX Spark (GB10)

The NVIDIA DGX Spark (GB10) holds more — 128 GB against 16 GB — which decides what you can load at all. The RTX 4070 Ti Super has more bandwidth at 672 GB/s, which decides how fast tokens come out once a model fits.

Model (Q4_K_M, 8K)RTX 4070 Ti Supertok/sNVIDIA DGX Spark (GB10)tok/s
Llama 3.1 8B Runs comfortably 77.6 Runs comfortably 35.1
Llama 3.1 70B Won't fit 1.22 Runs comfortably 4.37
Llama 3.2 3B Runs comfortably 172 Runs comfortably 78.1
Llama 3.2 1B Runs comfortably 456 Runs comfortably 209
Llama 4 Scout 109B-A17B Won't fit 4.42 Runs comfortably 14.8
Qwen3 8B Runs comfortably 75.1 Runs comfortably 33.9
Qwen3 14B Runs comfortably 43.9 Runs comfortably 19.8
Qwen3 32B Won't fit 4.34 Runs comfortably 9.17
Qwen3 4B Runs comfortably 137 Runs comfortably 62.2
Qwen3 30B-A3B Won't fit 40.2 Runs comfortably 53.5
Qwen3 235B-A22B Won't fit 3.05 Won't fit 6.20
Qwen2.5-Coder 32B Won't fit 4.34 Runs comfortably 9.17
Qwen2.5 7B Runs comfortably 86.3 Runs comfortably 38.9
Qwen2.5 72B Won't fit 1.18 Runs comfortably 4.24
Gemma 3 4B Runs comfortably 146 Runs comfortably 66.4
Gemma 3 12B Runs comfortably 52.7 Runs comfortably 23.8

Specifications

RTX 4070 Ti SuperNVIDIA DGX Spark (GB10)
Memory16 GB128 GB
Bandwidth672 GB/s273 GB/s
FP16 compute88 TFLOPS125 TFLOPS
Launch price$799$3,999
ArchitectureAdaBlackwell