NVIDIA DGX Spark vs NVIDIA RTX 5090 for local LLMs
71 reports on the NVIDIA DGX Spark and 126 on the NVIDIA RTX 5090
Which is faster for local LLMs?
On Gemma 4 26B · 4B active at 4-bit, the NVIDIA DGX Spark runs at 59.1 tokens per second (2 runs) and the NVIDIA RTX 5090 at 243.9 (one run). Those are community reports whose engines and settings differ, so the gap is a rough sign, not a controlled measure of the two cards. Memory bandwidth, which caps how fast a card can write, is 6.6x higher on the NVIDIA RTX 5090 (273 GB/s on the NVIDIA DGX Spark, 1,792 on the NVIDIA RTX 5090). Across the 2 models both have been measured on, the NVIDIA DGX Spark's median is higher on 0 and the NVIDIA RTX 5090's on 2.
- Memory
- 128GB unified
- Memory bandwidth
- 273 GB/s
- FP16 compute
- 125 TFLOPS
- Reports
- 71
- Memory
- 32GB
- Memory bandwidth
- 1,792 GB/s
- FP16 compute
- 419 TFLOPS
- Reports
- 126
Measured on both
Median tokens per second from plain runs (one device, one request, no speculative decoding, the whole model in the device's memory) of the same model size at the same bit class. Each side shows the engines and reported context behind its median. Context can mean a configured window or prompt depth, so these are community comparisons with differing settings. How to read these
| Model | NVIDIA DGX Spark | NVIDIA RTX 5090 |
|---|---|---|
| Gemma 4 26B · 4B active4-bit | 59.12 runsllama.cpp, vLLM256K context, 1 not stated | 243.9one runllama.cppcontext not stated |
| Qwen3.6 35B · 3B active4-bit | 78.3one runvLLM128K context | 215.1one runllama.cppcontext not stated |
Estimated by the calculator
The quant the calculator recommends for each card and its estimated speed at a 32,768-token context with 32 GB of system RAM, from memory bandwidth and the model's shape. Where a model was also measured above, the measurement wins.
| Model | NVIDIA DGX Spark | NVIDIA RTX 5090 |
|---|---|---|
| Gemma 4 26B · 4B active | 10.8 t/sFP16 | 114.5 t/sQ6_K |
| Qwen3.6 35B · 3B active | 14.5 t/sFP16 | 163.0 t/sQ5_K_M |
| Qwen3.8 27B | 4.0 t/sFP16 | 44.9 t/sQ6_K |
| Qwen3.8 125B · 6B active | 14.4 t/sQ6_K | 20.3 t/sQ2_K, experts in RAM |
| DeepSeek V4 Flash 284B · 13B active | 38.3 t/sQ2_K | doesn't fit at these settings |
| Qwen3.6 27B | 4.0 t/sFP16 | 44.9 t/sQ6_K |
| GLM-5.3 320B · 18B active | 12.7 t/sQ2_K | doesn't fit at these settings |
| DeepSeek V4.1 Flash 552B · 16B active | doesn't fit at these settings | doesn't fit at these settings |
Change the context, RAM or model in the calculator: NVIDIA DGX Spark · NVIDIA RTX 5090
Every report on each card, with its source: NVIDIA DGX Spark · NVIDIA RTX 5090