Head to head

RTX 5090 vs NVIDIA DGX Spark (GB10)

The NVIDIA DGX Spark (GB10) holds more — 128 GB against 32 GB — which decides what you can load at all. The RTX 5090 has more bandwidth at 1792 GB/s, which decides how fast tokens come out once a model fits.

Model (Q4_K_M, 8K)RTX 5090tok/sNVIDIA DGX Spark (GB10)tok/s
Llama 3.1 8B Runs comfortably 222 Runs comfortably 35.1
Llama 3.1 70B Won't fit 2.65 Runs comfortably 4.37
Llama 3.2 3B Runs comfortably 478 Runs comfortably 78.1
Llama 3.2 1B Runs comfortably 1235 Runs comfortably 209
Llama 4 Scout 109B-A17B Won't fit 6.75 Runs comfortably 14.8
Qwen3 8B Runs comfortably 214 Runs comfortably 33.9
Qwen3 14B Runs comfortably 126 Runs comfortably 19.8
Qwen3 32B Runs comfortably 59.0 Runs comfortably 9.17
Qwen3 4B Runs comfortably 380 Runs comfortably 62.2
Qwen3 30B-A3B Runs comfortably 159 Runs comfortably 53.5
Qwen3 235B-A22B Won't fit 3.83 Won't fit 6.20
Qwen2.5-Coder 32B Runs comfortably 59.0 Runs comfortably 9.17
Qwen2.5 7B Runs comfortably 247 Runs comfortably 38.9
Qwen2.5 72B Won't fit 2.50 Runs comfortably 4.24
Gemma 3 4B Runs comfortably 405 Runs comfortably 66.4
Gemma 3 12B Runs comfortably 151 Runs comfortably 23.8

Specifications

RTX 5090NVIDIA DGX Spark (GB10)
Memory32 GB128 GB
Bandwidth1792 GB/s273 GB/s
FP16 compute209 TFLOPS125 TFLOPS
Launch price$1,999$3,999
ArchitectureBlackwellBlackwell