The NVIDIA RTX Spark is no longer a secret, and it’s not an exaggeration to say that the brand is very clearly taking the fight to its rivals. And that list of rivals is the full Monty – I’m talking Intel, AMD, and Qualcomm, and all in ways that are surprising.
We got to spend some time with the coveted chipset, thanks to NVIDIA inviting us to a demo session with them. Here is the breakdown of what we saw and learned.
The Chipset
Let’s address the elephant in the room first. Yes, the RTX Spark is a variation of the DGX Spark. Well, to be absolutely simple about it: it’s obviously a cut down or watered down version of it.
Once again, here’s the skinny on the specs: It’s an SOC based on the ARM architecture, and fabricated and manufactured through a partnership with MediaTek. Beneath the hood, you’re looking at 20 Grace CPU cores; 6,144 Blackwell RTX GPU cores, 5th generation Tensor Cores, and 128GB of Unified Memory. All those components are linked through NVIDIA’s NVLink-C2C Interconnect.
The Performance
To point out another obvious point, it’s clear that NVIDIA intends for the RTX Spark to serve the masses in a general purpose capacity. During the demo sessions, NVIDIA gave us the usual fare of performance metrics that we’ve seen with each technological update it’s created over the years: the benefits of creators, gaming, and of course, as it is its bread and butter, Generative AI and the programs it uses.
With the RTX Spark, those talking points felt a little different this time.
In the years before, NVIDIA was hyping up the prowess of its GPUs, primarily from the time of Ada Lovelace and more recently, Blackwell. That’s just it: those tasks- all those AI-related functions were all executed purely from a dedicated GPU.
For a self-serving, all-in-one package like the RTX Spark to do all that is nothing short of impressive. Think rendering fully detailed, 3D models of cities in real-time and the ability to zoom into a specific street without the downtime or overhead. Think of using Blender and Agentic AI to create a render with NVIDIA’s Ray Reconstruction (RR), greatly reducing the noise generated from switching the camera position around. It is truly the kind of power Qualcomm wishes its Snapdragon X Elite series could do.
Then there is, of course, the subject that most of us hold near and dear to our hearts: gaming on the RTX Spark. And based on what I experienced, it does it well.
A couple things I should note. One, NVIDIA made it clear it wasn’t talking numbers or specs of the RTX Spark in gaming, nor other details about the laptops serving as its cradle. Sure, it has 128GB of Unified Memory, but I would’ve liked to see the rest of the laptop specs first hand.
Two, NVIDIA also wasn’t allowing us to see the average frames of the games that were on tap, nor did it allow us to see what sort of graphics presets were being used with them. The reason it gave was that all the “working” units were still prototypes and far from the final version of the finished product, and as such, the numbers that we’d be seeing were basically not the games running at their full potential.
That said, I managed to play Pragmata (for the first time, even) on a Microsoft Surface Ultra (more on this in the next part) laptop with the RTX Spark and it was actually a very smooth experience. Fun fact: the laptop with Pragmata was also running through an x86 translation layer, and on another laptop, NVIDIA was running Alan Wake 2 on an ARM-native version of the game. Oh, and both titles were loaded with all the bells and whistles, such as ray tracing DLSS 4.5, Multiframe Generation, and even Ray Reconstruction.
And that brings us nicely to the next topic: the laptops housing the RTX Spark.
I Thought These Laptops Would Be Lighter
For a laptop powered by an SoC, I honestly thought that the laptops would be lighter because…well, you’d have less components. But that proved not to be the case with not just one, but all of NVIDIA’s participating laptop partners. ASUS, Lenovo, HP, Dell, Microsoft: all their RTX Spark-powered laptops were hefty.
I want to say that all that extra weight comes from the cooling solution and battery, but I’ve seen conventional laptops of the 16-inch variety that don’t weight as much as these RTX Spark laptops did. Well, alright, maybe the MacBook Pro 16 that I reviewed a few years ago.
That’s another gripe that I have with these laptops. Nearly all of the brands only have 16-inch models, with ASUS being the only one that manufactures both 16-inch and 14-inch ProArt models. Personally, I’m all for a 14-inch laptop due to its size-to-weight ratio, but again, even this version of the laptop, dummy model notwithstanding, had some heft to it.
To be fair, the physical hardware these laptops are packing isn’t lightweight either. I’m talking Tandem OLED displays with a base resolution of QHD (2,560 x 1,440), with one laptop model actually shipping out with a 4K OLED panel. At the end of the day, these are going to be some premium devices.
Which brings us to our final subject…
These RTX Spark Laptops Aren’t Going To Come Cheap
As Jensen Huang, NVIDIA CEO, said on stage, RTX Spark laptops will be launching by the end of this year. What he didn’t mention was just how much they’d cost, and in my opinion, I don’t think they’re going to come cheap for obvious reasons.
Those reasons are nothing you haven’t heard before: the ongoing memory shortage crisis, which ironically was brought on by companies like NVIDIA who have been gobbling up advanced DRAM chips it can get its hands on, which in turn have caused the prices of available memory products in the consumer market to skyrocket.
Considering that NVIDIA has a tendency to charge a premium, I don’t expect starting prices for RTX Spark laptops to start any lower than RM9,000, and I am still just lowballing that. Honestly, I think that NVIDIA may just be running its usual marketing strategy: introduce its highest end SKU first, followed by lower-tier variations of the chipset shortly after.
