DGX Spark x2 vs RTX 3090
Best sourced output tok/s per model (single-stream generation) — engine and version, quantization, context and a source link on every row. Hover a value for the quant.
DGX Spark x233 sourced runs · best 474.0 tok/s
RTX 309018 verified runs · best 709.5 tok/s
Specs and evidence
| DGX Spark x2 | RTX 3090 | |
|---|---|---|
| VRAM | 128 GB | 24 GB |
| Sourced runs on the board | 33 | 18 |
| Models measured | 7 | 11 |
| Best output tok/s | 474.0 tok/s | 709.5 tok/s |
| Highest measured peak VRAM | — | 22.0 GB |
| Street price (latest snapshot) | — | $1,720 |
Head-to-head
On the 3 models measured on both cards, DGX Spark x2 leads on 3, RTX 3090 on 0. The widest gap is on DeepSeek-V4.1-Flash: DGX Spark x2 154.0 tok/s vs RTX 3090 26.0 tok/s (DGX Spark x2 492% faster).
The measured matrix
| Model | DGX Spark x2 | RTX 3090 |
|---|---|---|
| LFM2.5-1.2B-Instruct | — | 709.5 |
| Qwen3.5-0.8B-MTP | — | 476.1 |
| Qwen3.8-Flash-Next-hibrid48 | 474.0 | — |
| NVIDIA-Nemotron-3.5-Lightning-30B-A3B | — | 353.7 |
| LFM2.5-2.6B | — | 305.6 |
| Qwen3.5-4B-MTP | — | 240.4 |
| Qwen3.8-27B | — | 164.6 |
| Qwen3.5-9B-MTP | — | 164.3 |
| DeepSeek-V4.1-Flash | 154.0 | 26.0 |
| Qwen3.8-Flash-Next | 93.0 | 38.6 |
| DeepSeek-V4.1-Flash-1017 | 92.8 | — |
| Beam-1.0 | 58.9 | — |
| GLM-5.3-Flash | 46.3 | 28.1 |
| MiniMax M3.1 Flash | 39.4 | — |
| Ornstein3.6-27B-MTP-NSC-ACE-SABER | — | 32.9 |