NVIDIA DGX Spark

What hackers pick instead

• Apple Mac miniA small, fast, power-efficient Mac whose unified memory suits local AI models and always-on macOS work, though storage upgrades are costly and it struggles with big LLMs.390recommend
• AMD Strix HaloA normal x86 PC with up to 128GB of unified memory for running fairly large local models, though memory bandwidth limits speed, ROCm is fiddly and prices have climbed.130recommend
• Framework DesktopA compact, low-power mini PC with up to 128GB of shared memory for running large local AI models on Linux, though critics say memory bandwidth limits LLM speed.93recommend
• Intel NUCSmall, quiet, cheap mini PC people use for home servers and basic desktop work, with replaceable RAM and storage. Critics say some models overheat.42recommend
• Beelink mini PCSmall mini PCs that run Linux or Windows well, stay quiet and cool, and last for years. People also praise the support and the value for the money.24recommend

Every quote

The 40 most detailed of 448 comments. Each links to the original.

1. Now running a health-AI node for all my longitudinal health data - insights are generated on the box so that raw data never leaks.

   Recommends · Jul 2026 · in “Unified Memory, Explained: Why Mini PCs Can Run 70B Models a Big GPU Can't”

2. the prompt processing speeds are a killer improvement of the Spark which even 2 Strix Halo boxes would not match

   Recommends · Jan 2026 · in “Dell's version of the DGX Spark fixes pain points”

3. I've found it very useful for running big models, but it's not a screaming powerhouse in terms of raw compute.

   Mixed · Jun 2026 · in “Nvidia is proposing a beast of a CPU system for Windows PCs”

4. the Spark's prefill is so much faster it can often answer the same query before the mac's prefill is finished

   Recommends · Jun 2026 · in “Nvidia is proposing a beast of a CPU system for Windows PCs”

5. works darn well for our purposes, and we control the compute and data flowing through it, so totally worth it

   Recommends · May 2026 · in “Cut Off: why access to frontier AI will soon be scarce and selective”

6. The spark can fine tune models in 1/4 the time and excels at other compute tasks in ways that Mac never can.

   Mixed · Jun 2026 · in “Unsloth GLM-5.2 – How to Run Locally”

7. they are a bit of a mixed bag as they don't expose some Blackwell-specific features that make a difference

   Mixed · Jan 2026 · in “Apple Is Fighting for TSMC Capacity as Nvidia Takes Center Stage”

8. I can get something like 55 tokens a sec running a larger model like gpt-oss-120b-Q8_0 on the DGX Spark.

   Recommends · Mar 2026 · in “Tinybox- offline AI device 120B parameters”

9. The NVIDIA will slightly win on all generation speed, and be _much_ faster on first token response time.

   Recommends · Nov 2025 · in “Asus Ascent GX10”

10. The prefill and NVFP4 is a significant advantage. Where as DGX Spark you're at the mercy of NVIDIA BSP.

    Mixed · Aug 2026 · in “Qwen3.8-Max: A New Bar for Coding and Cowork”

11. It will pay for itself within three months doing stuff on prem and us not sending data to someone else.

    Recommends · Aug 2026 · in “Ten advances in mathematics and theoretical computer science”

12. mine are silent and can fit 2 qwen/gemma at 8bit or give you room for a bunch of other, smaller models

    Recommends · Jun 2026 · in “Qwen 3.6 27B is the sweet spot for local development”

13. memory bandwidth is abysmal compared to Apple chips - 273 GB/s vs 614 GB/s on M5 Max for similar price

    Mixed · Mar 2026 · in “Apple discontinues the Mac Pro with no plans for future hardware”

14. You can just about reach the lower end of the latter category with a 128GB machine like a DGX Spark

    Mixed · Jun 2026 · in “Ask HN: Has anyone replaced Claude/GPT with a local model for daily coding?”

15. shrug the GB10 has far more compute than it has memory bandwidth, so prefill isn't it's bottleneck.

    Recommends · Jun 2026 · in “Running local models is good now”

16. These devices are for AI R&D. If you need to build models or fine tune them locally they're great.

    Mixed · Jan 2026 · in “Dell's version of the DGX Spark fixes pain points”

17. Very good for training, okay for inferencing if you are running batch jobs and don't mind waiting.

    Mixed · Apr 2026 · in “Intel Arc Pro B70 Review”

18. Also limited to 128GB memory and comparatively low bandwidth, but higher compute than Strix Halo.

    Mixed · Aug 2026 · in “Apple introduces M6 and M5 Ultra”

19. will have better throughput and access to bigger models (more vram when running tensor parallel)

    Recommends · Jun 2026 · in “Qwen 3.6 27B is the sweet spot for local development”

20. I would rather get a Spark and have the RAM to experiment with larger LLMs, even if it was slow.

    Recommends · Sep 2026 · in “M5 Ultra Mac Studio Review”

21. I think the spark is actually useful, I don't think pure inference is a good use case for them.

    Mixed · Jan 2026 · in “Dell's version of the DGX Spark fixes pain points”

22. Purchase a Mac Studio or DGX Spark and use the latest local models to securely audit your code.

    Recommends · Sep 2026 · in “Don't be the out of touch Kung Fu master”

23. you basically have to use nvidia's linux (from what I've read) and it doesn't appear the nvidia is interested in upstreaming their kernel changes

    Criticizes · Jul 2026 · in “Nvidia DGX Spark as a daily driver”

24. It would run so slow it would be functionally unusable or at such a low quant that it wouldn't be useful for serious work.

    Criticizes · Jun 2026 · in “Semgrep: GLM 5.2 beats Claude in our Cyber Benchmarks”

25. It does seem really shady that they'd claim it to be 5th gen tensor cores and then not support the full feature set

    Criticizes · Jan 2026 · in “Dell's version of the DGX Spark fixes pain points”

26. Its main issue is low memory bandwidth, i.e. it has quite fast token prefill but awful token generation.

    Criticizes · Apr 2026 · in “Intel Arc Pro B70 Review”

27. Those things aren’t spectacular at inference… it’s not really why you drop that kind of money on them.

    Criticizes · Sep 2026 · in “Grok outage”

28. It has to work on day one whatever the Apple Mac M5 Ultra or Mac M6 Ultra are they will work well on day one, the cost of the laptop Spark probably in the thousands of dollars has to work from the start or its dead.

    Mixed · Jun 2026 · in “Nvidia announces new AI chip for personal computers”

29. The DGX Spark hits a sweet spot for me where it can simultaneously work for general computing and run local LLM inference fast enough for some hobbyist dabbling.

    Recommends · Jul 2026 · in “Nvidia DGX Spark as a daily driver”

30. You can run multiple instances of these models in parallel on the DGX Spark which somewhat mitigates the difference if your task is parallelizable.

    Mixed · Jun 2026 · in “Running local models is good now”

31. DGX Spark, Strix Halo, lower-RAM machines with SSD offload) where you have more compute than you could use for single-stream decode.

    Mixed · May 2026 · in “A Few Words on DS4”

32. The Spark IS really nice if you want to test out ConnectX or if you really need something small and compact and quiet.

    Mixed · Aug 2026 · in “Meta Muse Glimmer – open weights 30B local coding model”

33. once you start scoping out the cost of a M3 Ultra with 128GB or 256GB, the DGX Spark doesn’t look bad after all

    Recommends · Jun 2026 · in “Unsloth GLM-5.2 – How to Run Locally”

34. truly, utterly, mind-blowingly great platform to do cross platform high performance application development

    Recommends · Oct 2026 · in “Apple's git still trusts the compromised eMudhra certificates”

35. Super-powerful (if rather pricy) Linux desktop that happens to play games while doing everything else

    Recommends · Jul 2026 · in “Nvidia DGX Spark as a daily driver”

36. Getting an OEM (DGX) Spark is good if you want complete control or an excellent learning environment

    Recommends · Oct 2026 · in “Ask HN: How to get the most intelligence out of your agents?”

37. Zero issues so far. Extremely reliable and gives me straight path to production for VPS deployment.

    Recommends · Jul 2026 · in “Nvidia DGX Spark as a daily driver”

38. For a dedicated headless setup, I'd probably use something like NVIDIA DGX Spark rather than a Mac

    Recommends · Sep 2026 · in “Apple caught off guard by AI demand for Mac Mini and Mac Studio”

39. one of the few that I can actually buy with Linux pre-installed with a fully OEM supported version

    Recommends · May 2026 · in “Mini PC for local LLMs in 2026”

40. NVIDIA has only the advantage (in its more expensive variants) of a GPU equivalent with RTX 5070.

    Mixed · Jun 2026 · in “Nvidia is proposing a beast of a CPU system for Windows PCs”