NVIDIA RTX Spark laptops ship October 16, and the spec sheet gives both N1X chips the same 45-80 W

NVIDIA opened preorders for RTX Spark laptops on October 7, at Microsoft's Windows and Surface event in San Francisco. In NVIDIA's words, the platform comes "with laptop preorders open today and available October 16. Compact desktops will be available for sale in November."

RTX Spark pairs an NVIDIA Grace CPU with a Blackwell RTX GPU on one pool of unified LPDDR5X memory. The launch post and the product page together give a fairly complete picture of the silicon. They also leave a few things out.

Two laptop chips, one power range

The product page's specification table lists two laptop configurations under the N1X name:

Both carry fifth-generation Tensor cores, fourth-generation RT cores, one ninth-generation encoder and one sixth-generation decoder, PCIe Gen 5, and HDMI 2.1b and DisplayPort 2.1b support (footnotes tell buyers to check each laptop maker for what a given model actually exposes). A second table, under the page's Desktops tab, lists the 6,144-core, 20-core, 128 GB part at 140 W.

That shared 45-80 W range is the number to watch in reviews. The page gives the two chips identical power envelopes, so how a given laptop performs will depend on where in that range its maker sets the limit, and NVIDIA's page does not say how far apart the two configurations land at the same wattage.

The launch post adds the interconnect figure: "an up to 20-core NVIDIA Grace CPU connected at 600 GB/s." What neither document publishes is the memory bandwidth of the LPDDR5X pool itself, which is the figure that bounds token generation for large local models. Until reviewers measure it, any tokens-per-second claim for a 100B-class model on these machines is an estimate.

The model-size claims, side by side

NVIDIA makes two slightly different promises about what fits. The product page: "Run models up to 120B parameters with 1M context." The launch post names a specific model that is larger than that: Qwen 3.8 Flash Next, "a 125B model with 51B n-gram that matches the intelligence of many cloud models." The post does not explain what the 51B n-gram component is or how much of the 128 GB it occupies.

Both claims assume the top 128 GB configuration. Neither document says which laptops will ship with 128 GB, and the lower chip tops out at 64 GB by NVIDIA's own table.

Who is shipping

The launch post says "Systems are coming from Acer, ASUS, Dell, HP, Lenovo, Microsoft, MSI and Gigabyte." The product page's laptop list is shorter, six models: ASUS ProArt P16, Dell XPS 16, HP OmniBook Ultra 16, Lenovo Pro 9n, Microsoft Surface Laptop Ultra and MSI Prestige N16 Flip AI+. Acer and Gigabyte appear in the post but have no laptop on the preorder page. Microsoft's Pavan Davuluri, quoted by NVIDIA, framed the Surface model this way:

We built Surface Laptop Ultra around NVIDIA RTX Spark. With up to 128 gigs of unified memory and up to a petaflop of AI compute, you can run models on this laptop that simply don't fit on a traditional machine.

Pavan Davuluri, Microsoft, as quoted on the NVIDIA Blog

Neither NVIDIA document gives an RTX Spark price.

The same month, the other Spark

RTX Spark is not NVIDIA's only unified-memory box this month. On October 2 it announced a DGX Spark configuration with 64 GB, "available from Acer, ASUS, Dell, Gigabyte, HP and MSI on Friday, Oct. 23, starting at $4,999," with the GB10 superchip and DGX OS unchanged from the 128 GB model. NVIDIA says it "supports up to 100-billion-parameter models," and that two units joined over their ConnectX-7 ports pool to 128 GB, with "up to 1.7x" the performance of one unit in its own Qwen 3.8 27B test.

The two products now overlap: a Linux developer box at 64 GB and a Windows laptop chip at up to 128 GB, both sold on local inference. NVIDIA's own documents do not yet give the bandwidth figures needed to compare them on that job.

Primary sources: NVIDIA Blog: NVIDIA, Microsoft Kick Off a New Beginning for Windows PCs With RTX Spark and AI Agents, NVIDIA RTX Spark product page, NVIDIA Blog: NVIDIA DGX Spark 64GB, read 2026-10-09.