Last week, AMD fired a warning shot at Nvidia’s RTX Spark. Today, Team Green’s payload has arrived. Ahead of the highly anticipated October 16 launch, Nvidia just unleashed a tidal wave of software for its audacious Arm-based chip, the massive Blackwell RTX GPU.
They aren’t just selling a new chip; they’re trying to prove that their garden of software features makes the jump to Arm completely worth it.
And on top of that, I currently have the Microsoft Surface Laptop Ultra with RTX Spark sitting on my desk (actually typing this story on it right now). Now, before I lock myself in my lab for the review, I want to know exactly what you want to see this thing do.
Here is the cheat sheet of the craziest software witchcraft Nvidia just announced. Read it over, drop your most chaotic testing requests in the Q&A box just below, and I will push this hardware to the absolute limit.
Reminder: We’re a friendly bunch, and this community is a safe space for anyone who wants to learn about RTX Spark. Please treat everyone (including me) with respect and avoid any nasty language.
AI without the computer science degree
Running a Large Language Model (LLM) on your PC used to require opening terrifying command-line terminals and pretending you knew how to code (I’m guilty of that one with my D in Computer Science). That friction is well on its way to being killed, and Nvidia is taking another big step forward.
- Hugging Face Llama Windows: Available today is a free, idiot-proof Windows app that lets you download and run local AI models with a standard graphical interface.
- Nvidia PAIR: Now out of beta, this free, open-source tool can essentially turn your home into its own AI data center — allowing PCs and MacBooks to share AI compute across a local network. Basically, if your prompt needs more horsepower, it can find some from your desktop in the other room.
- Getting agent-ready: New skills are coming online so your AI agents in Hermes, OpenClaw, and Perplexity can use Nvidia’s apps (Broadcast, RTX Remix, etc.).
Also, Nvidia Broadcast is getting a big upgrade with RTX video super resolution and frame generation—both for webcam and offline file processing.
The “Nemotron” Intern
On October 15th, Nvidia is dropping the Nemotron 3.5 Bolt 74B A9B model. That’s a whole lot of words, but this is basically an improved key to agentic AI.
This is a “mixture of experts” AI agent built specifically to fit inside 24GB of memory on the RTX Spark — running constantly in the background, writing code, doing research, and monitoring apps autonomously.
Skyrim NPCs that actually talk back
That Nemotron model style has a different use too — making game NPCs have personalities. Nvidia’s built a smaller, ultra-responsive model called Nemotron 3.5 Bolt E4B (again, more alphabet spaghetti), but on October 23rd, you’ll see what this can do in SkyrimNET.
Basically, this is an AI NPC mod for Elder Scrolls V, where instead of clicking through pre-written dialogue trees, you can literally talk into your microphone, and NPCs will generate dynamic, context-aware conversations and reactions on the fly.
I can’t wait to gaslight Nazeem with this…
DLSS 5 just found another gear (and a new cheat code)
Nvidia’s DLSS 5 has split opinions in NBA 2K27 (I see you in the comments). In my mind, this 3D-guided neural rendering model works well in photorealistic games — but maybe leave it off in more artistic ones.
But on October 16th, it’s set to get a speed boost with a more efficient model that Team Green claims will push frame rates up by an additional 1.4x over what I tested. Oh, and for the millennial-coded individuals out there, Hogwarts Legacy is getting the DLSS 5 treatment in November.
Over to you: Build my benchmark hitlist
Nvidia is making massive promises about Arm-based gaming, offline AI agents that don’t charge you a cloud tax, and real-time AI video game characters. AMD claims this whole platform will falter under the weight of Windows emulation.
I have the hardware. I have the games. I have the local AI models. Let’s see what this thing can do together.
Now, while I can do plenty of testing and report on my findings soon, some of this won’t be available for a little while. So this will be an ongoing thing as models and updates become available.
More from Tom's Guide
- I didn’t believe an Nvidia RTX Spark laptop would have the same performance both on and off the charger — so I put it to the test
- Forget RTX Spark: The Surface Laptop Ultra’s magnetic USB-C charger is a game-changer that makes Apple’s MagSafe look obsolete
- AMD just flaunted its own 192GB monster chip to spoil Nvidia’s RTX Spark party — Ryzen AI Max+ Pro 495 should worry Apple, too