--- title: RTX Spark laptops are fast, until Linux and Arm get involved description: Nvidia's RTX Spark laptops run fast on Windows but hit major snags on Linux and Windows on Arm. On an MSI Prestige, plugging into wall power on Linux locks the... canonical: https://daily.dev/posts/rtx-spark-laptops-are-fast-until-linux-and-arm-get-involved-fuzpeirbw twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: RTX Spark laptops are fast, until Linux and Arm get involved | daily.dev og:description: Nvidia's RTX Spark laptops run fast on Windows but hit major snags on Linux and Windows on Arm. On an MSI Prestige, plugging into wall power on Linux locks the... og:url: https://daily.dev/posts/rtx-spark-laptops-are-fast-until-linux-and-arm-get-involved-fuzpeirbw og:image: https://api.daily.dev/og/posts/fuZPeirbW.png og:image:alt: RTX Spark laptops are fast, until Linux and Arm get involved og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # RTX Spark laptops are fast, until Linux and Arm get involved **[Trends](https://daily.dev/sources/trends)** · 3 min read · 0 upvotes · 0 comments ## Summary Nvidia's RTX Spark laptops run fast on Windows but hit major snags on Linux and Windows on Arm. On an MSI Prestige, plugging into wall power on Linux locks the GPU clock to about 780 MHz, versus 2 GHz-plus on Windows. Running GPT-OSS-120B, Windows and WSL hit roughly 65 tokens per second while native Linux managed only 13.9. The cause is an encrypted, Windows-only firmware power table Linux can't decrypt. A teardown found the RTX Spark shares its SoC with the DGX Spark's GB10 but gets faster memory (9,400 MT/s vs 8,533) and an unlocked NPU. On Windows on Arm, an RTX 4080 in a Thunderbolt enclosure falls back to a basic display adapter since Nvidia hasn't shipped an Arm64 driver for it, and Torch, OpenCV and cryptography lack Windows on Arm wheels on PyPI. ## Content The sharpest advice on RTX Spark laptops right now is oddly specific: don't run Linux natively. @russano puts it bluntly: run local models in WSL, because the same machine and the same model gets 65.1 tokens per second there versus 13.9 on native Linux. ## Why Linux loses On laptops like the MSI Prestige, plugging into wall power locks the GPU clock to about 780 MHz. Windows hits 2 GHz-plus. On GPT-OSS-120B, Windows and WSL both land around 65 tokens per second. The cause is the Nvidia Platform Controllers and Framework, a Windows-only firmware component with an encrypted power table that Linux can't decrypt. Unplug the laptop and Linux gets much closer. WSL keeps full boost either way, so it's the practical pick until Nvidia ports the power logic to its Linux driver. The irony is that the chip is the same silicon as the DGX Spark's GB10. Both report SoC revision 0x04268901 and have matching CPU layouts. The laptop gets faster memory, 9,400 MT/s versus 8,533, worth roughly 300 GB/s theoretical and an 18% lead in one GPU read test. It also exposes an NPU the GB10 leaves disabled. But Windows reserves up to 96 GB as dedicated GPU memory, while Linux on the GB10 lets CUDA use the whole pool. Each side wins something. ## Windows on Arm is still Windows on Arm One reviewer says Cyberpunk 2077 and Forza Horizon run fine through Microsoft's Prism emulator. The gaps are elsewhere: - An RTX 4080 in a Thunderbolt enclosure falls back to a basic display adapter, because Nvidia hasn't shipped an Arm64 driver with that device ID. - torch, OpenCV and cryptography lack Windows on Arm wheels on PyPI. - Steam has no official Arm client. - RTX Spark and Snapdragon X need a separate Windows 11 26H1 branch, unlike Intel's Panther Lake. Arm hit 15.3% of client PCs in Q2 2026, per Mercury Research, mostly Macs and Chromebooks. ## The flex The same reviewer found the laptop still lacked Copilot+ features, so a local Qwen 3.8 Flash Next model (125B parameters) built its own from one prompt. It ran unattended for 5 hours, 41 minutes and produced "Spark Copilot", recreating Recall, Click to Do, Live Captions and file search. That took 370,010 output tokens across 524 requests and 60 tests. It wrote its own COM interface to get around missing Windows on Arm libraries, and fixed a bug that cut search from 25 seconds to half a second. ## The stunt And for pure spectacle: a demo has an RTX 5090 showing up as "GPU 1" on a small ASUS laptop over Wi-Fi, confirmed by nvidia-smi, and playing Cyberpunk with no cables. The creator also claims it can run large models using 4 TB of memory, though not the way you'd expect. ## Questions this post answers ### Why is my Nvidia RTX Spark laptop so much slower running LLMs on native Linux than on Windows? Native Linux on the RTX Spark gets throttled because the GPU clock locks to about 780 MHz on AC power, versus 2 GHz-plus on Windows. On GPT-OSS-120B, Windows and WSL both reach roughly 65 tokens per second while native Linux manages only 13.9, about 35% of the speed. The cause is the Nvidia Platform Controllers and Framework, a Windows-only firmware component with an encrypted power table Linux can't decrypt. _Anyone benchmarking local LLMs on RTX Spark hardware tracks driver fixes like this via daily.dev._ ### Is the RTX Spark laptop the same chip as the Nvidia DGX Spark? Yes, a teardown comparison found the RTX Spark and the DGX Spark's GB10 share the same SoC revision (0x04268901), CPU layout and GPU specs. The laptop differs by using faster memory, 9,400 MT/s versus 8,533 MT/s, worth roughly 300 GB/s theoretical bandwidth and an 18% lead in one GPU read test, and it exposes an NPU that the GB10 leaves disabled. _Developers comparing Spark hardware for local inference follow teardown details like this on daily.dev._ ### Does Nvidia support Windows on Arm for external GPUs like the RTX 4080? No, an RTX 4080 connected through a Thunderbolt enclosure falls back to a basic display adapter on Windows on Arm because Nvidia hasn't shipped an Arm64 driver with that device ID. Beyond GPU drivers, Torch, OpenCV and cryptography lack Windows on Arm wheels on PyPI, and Steam has no official Arm client, limiting what runs on Arm-based Windows machines like the RTX Spark or Snapdragon X. _Anyone weighing Windows on Arm for GPU workloads tracks driver and wheel support gaps on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 1 discussion and 121 comments across hackernews (as of 2026-11-06). **TL;DR:** Commenters see this as yet another instance of Nvidia's historically poor Linux support, with many suspecting the encrypted power table is either deliberate market segmentation/DRM-in-waiting or simple Windows-first negligence rather than a pure bug. **Sentiment:** 5% positive · 25% mixed · 70% skeptical **The case for** - A few note Nvidia's closed Linux drivers have historically still matched Windows in features and performance, unlike AMD's pre-open-source era. - Some expect the issue will likely get reverse-engineered or patched fairly quickly given community interest. **The pushback** - Many suspect this is deliberate market segmentation or a precursor to DRM/licensing schemes for AI compute rather than an innocent bug. - Nvidia is repeatedly criticized as historically the worst major vendor for Linux support. - Several call it basic incompetence that a multi-trillion-dollar company shipped a laptop with broken power management on Linux. - Some argue running local inference on a laptop at all is a questionable use case versus a home server plus remote access. - Concern that encrypted/locked-down firmware previews a broader trend toward DRM for AI models and hardware, further entrenching cloud dependency. **By community** - hackernews (skeptical): Discussion is dominated by distrust of Nvidia's motives, broader worries about AI centralization/DRM, and familiar complaints about Nvidia's Linux track record, alongside a side debate about whether laptops are even the right form factor for local inference. **Hottest debate:** Whether the power table lockout is intentional market segmentation/DRM-style gatekeeping or just another case of Nvidia's Windows-first negligence toward Linux. **Open questions** - Will Nvidia port the encrypted power management logic to its Linux driver, and how quickly? - Can the encrypted firmware/power table realistically be reverse-engineered given it's a 'Windows-only' Nvidia Platform Controllers and Framework component? - Is the GB10/DGX Spark driver stack similar enough to be ported over to fix the RTX Spark issue? **Highlights** > >Oddly, the direction of the problem flips depending on power source. Unplugged, Linux hit around 61 to 67 tokens per second on LM Studio, comparable to Windows. Plugged in, that same test dropped to about 15 tokens per second. Absolutely hilarious. This is just incompetence. No reason Nvidia with its TRILLIONS of market cap couldn't of tested and fixed this on Linux before shipping. Or a day one patch. I have no doubt that'll it'll work right eventually. Long after RTX Spark laptops flop and can be purchased for half off. Last night I pulled out my 5070TI laptop and installed CachyOS. I was running various models within about 20 minutes. That 12GB of VRAM is a hard limitation, Unsloth can't split jobs between vram and system memory. > — [999900000999 on hackernews · 1 comments](https://news.ycombinator.com/item?id=50365264) > No, we are just seeing a push towards Nvidia wanting to make money. Nothing stops you from giving Nvidia half a million for their latest and greatest AI GPUs, unless you are in China I guess, but China does very well despite the export control. It is just classic market segmentation, the thing Nvidia and just about every company that can benefit from it has done for decades. Cloud providers are ready to pay a lot of money, consumers wanting local AI have a lower budget, and gamers even lower. But a low budget is greater than zero, and there is still money to be made on these markets. But they also don't want to compete against themselves, so they are going to find ways of crippling enterprise-class workloads on consumer hardware. It used to be that they crippled high precision arithmetic on consumer hardware, because it was a differentiating factor, but AI doesn't care about high precision, so now, it is something else. > — [GuB-42 on hackernews · 2 comments](https://news.ycombinator.com/item?id=50364018) > Years ago I would think I was being paranoid to think they would patch hardware retroactively to make it slower so you need to pay more to use your own hardware. In 2026, I think this is what all car companies are doing after you pay the full car price. I don't see a reason why they wouldn't do it with so little competition > — [motbus3 on hackernews · 1 comments](https://news.ycombinator.com/item?id=50363845) > Nvidia doing what Nvidia has always done; proprietary closed drivers. Weird choice to cryptographically protect the power settings. Key problem is that we've allowed a company that's always been this way to form a monopoly. Other brands exist, perhaps it is now worth the effort to migrate and rebuild elsewhere. > — [jleask on hackernews · 4 comments](https://news.ycombinator.com/item?id=50363520) > It's not weird at all, they know people will buy these laptops, put linux on them and stash them in data-centers, so they are trying to prevent this. > — [dist-epoch on hackernews · 1 comments](https://news.ycombinator.com/item?id=50364308) **Source threads** - [hackernews](https://news.ycombinator.com/item?id=50362905) · 210 points · 121 comments ## Similar posts on daily.dev - [Nvidia's RTX Spark is a chip unlike any other, and it could change Windows laptops forever](https://daily.dev/posts/nvidia-s-rtx-spark-is-a-chip-unlike-any-other-and-it-could-change-windows-laptops-forever-lgrsetnyj) · XDA Developers · 0 upvotes · 0 comments - [Microsoft built its most powerful Surface ever on Nvidia's new chip, and we saw every RTX Spark competitor coming this fall](https://daily.dev/posts/microsoft-built-its-most-powerful-surface-ever-on-nvidia-s-new-chip-and-we-saw-every-rtx-spark-comp-51dx7bfrq) · XDA Developers · 0 upvotes · 0 comments --- Tags: [#linux](https://daily.dev/tags/linux), [#nvidia](https://daily.dev/tags/nvidia), [#gpu](https://daily.dev/tags/gpu) [View this post on daily.dev](https://daily.dev/posts/rtx-spark-laptops-are-fast-until-linux-and-arm-get-involved-fuzpeirbw) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"RTX Spark laptops are fast, until Linux and Arm get involved","url":"https://daily.dev/posts/rtx-spark-laptops-are-fast-until-linux-and-arm-get-involved-fuzpeirbw","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/rtx-spark-laptops-are-fast-until-linux-and-arm-get-involved-fuzpeirbw"},"datePublished":"2026-11-04T11:21:43.200Z","dateModified":"2026-11-06T23:52:47.828Z","description":"Nvidia's RTX Spark laptops run fast on Windows but hit major snags on Linux and Windows on Arm. On an MSI Prestige, plugging into wall power on Linux locks the...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/8d1b4cb4c190fb91a2a9c09306eb6cd5?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/8d1b4cb4c190fb91a2a9c09306eb6cd5?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Trends","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Trends","logo":"https://media.daily.dev/image/upload/s--ZfSp3asX--/f_auto,q_auto/v1780996004/logos/trends?_a=BAMAMiWQ0","url":"https://daily.dev/sources/trends"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/rtx-spark-laptops-are-fast-until-linux-and-arm-get-involved-fuzpeirbw","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"linux,nvidia,gpu","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Trends","item":"https://daily.dev/sources/trends"},{"@type":"ListItem","position":3,"name":"RTX Spark laptops are fast, until Linux and Arm get involved"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/rtx-spark-laptops-are-fast-until-linux-and-arm-get-involved-fuzpeirbw#faq","mainEntity":[{"@type":"Question","name":"Why is my Nvidia RTX Spark laptop so much slower running LLMs on native Linux than on Windows?","acceptedAnswer":{"@type":"Answer","text":"Native Linux on the RTX Spark gets throttled because the GPU clock locks to about 780 MHz on AC power, versus 2 GHz-plus on Windows. On GPT-OSS-120B, Windows and WSL both reach roughly 65 tokens per second while native Linux manages only 13.9, about 35% of the speed. The cause is the Nvidia Platform Controllers and Framework, a Windows-only firmware component with an encrypted power table Linux can't decrypt. Anyone benchmarking local LLMs on RTX Spark hardware tracks driver fixes like this via daily.dev."}},{"@type":"Question","name":"Is the RTX Spark laptop the same chip as the Nvidia DGX Spark?","acceptedAnswer":{"@type":"Answer","text":"Yes, a teardown comparison found the RTX Spark and the DGX Spark's GB10 share the same SoC revision (0x04268901), CPU layout and GPU specs. The laptop differs by using faster memory, 9,400 MT/s versus 8,533 MT/s, worth roughly 300 GB/s theoretical bandwidth and an 18% lead in one GPU read test, and it exposes an NPU that the GB10 leaves disabled. Developers comparing Spark hardware for local inference follow teardown details like this on daily.dev."}},{"@type":"Question","name":"Does Nvidia support Windows on Arm for external GPUs like the RTX 4080?","acceptedAnswer":{"@type":"Answer","text":"No, an RTX 4080 connected through a Thunderbolt enclosure falls back to a basic display adapter on Windows on Arm because Nvidia hasn't shipped an Arm64 driver with that device ID. Beyond GPU drivers, Torch, OpenCV and cryptography lack Windows on Arm wheels on PyPI, and Steam has no official Arm client, limiting what runs on Arm-based Windows machines like the RTX Spark or Snapdragon X. Anyone weighing Windows on Arm for GPU workloads tracks driver and wheel support gaps on daily.dev."}}]} ```