NVIDIA DGX Spark vs Intel Arc Pro B70 for local LLMs
99 reports on the NVIDIA DGX Spark and 26 on the Intel Arc Pro B70
Which is faster for local LLMs?
On Qwen3.6 35B · 3B active at 4-bit, the NVIDIA DGX Spark has 78.3 generation tokens/s (one run); prompt processing 3556.7 tokens/s (one run); the Intel Arc Pro B70 has 63.7 generation tokens/s (5 runs); prompt processing 977.4 tokens/s (4 runs). 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 2.2x higher on the Intel Arc Pro B70 (273 GB/s on the NVIDIA DGX Spark, 608 on the Intel Arc Pro B70). Across 1 shared models with generation measurements on both cards, the NVIDIA DGX Spark's median is higher on 1 and the Intel Arc Pro B70's on 0.
- Memory
- 128GB unified
- Memory bandwidth
- 273 GB/s
- FP16 compute
- 125 TFLOPS
- Reports
- 99
- Memory
- 32GB
- Memory bandwidth
- 608 GB/s
- FP16 compute
- 183.5 TFLOPS
- Reports
- 26
Measured on both
Generation and prompt processing speeds 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 | Intel Arc Pro B70 |
|---|---|---|
| Qwen3.6 35B · 3B active4-bit | 78.3one runPP 3556.7 (one run)llama.cpp, vLLM128K context, 1 not stated | 63.75 runsPP 977.4 (4 runs)llama.cpp127K to 256K context, 2 not stated |
| Qwen3.8 27B4-bit | -0 runsPP 899.9 (one run)llama.cpp16K context | 25.6one runPP 512.1 (one run)llama.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 | Intel Arc Pro B70 |
|---|---|---|
| Qwen3.6 35B · 3B active | 14.5 t/sFP16 | 55.3 t/sQ5_K_M |
| Qwen3.8 27B | 4.0 t/sFP16 | 15.2 t/sQ6_K |
| Qwen3.8 125B · 6B active | 14.4 t/sQ6_K | 18.6 t/sQ2_K, experts in RAM |
| DeepSeek V4 Flash 284B · 13B active | 38.3 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 |
| Qwen3.6 27B | 4.0 t/sFP16 | 15.2 t/sQ6_K |
| Gemma 4 26B · 4B active | 10.8 t/sFP16 | 38.8 t/sQ6_K |
| GLM-5.3 320B · 18B active | 12.7 t/sQ2_K | doesn't fit at these settings |
Change the context, RAM or model in the calculator: NVIDIA DGX Spark · Intel Arc Pro B70
Every report on each card, with its source: NVIDIA DGX Spark · Intel Arc Pro B70