Nvidia RTX Spark PCs Ship in October With Up to 128GB Unified Memory
Nvidia used IFA 2026 in Berlin to put a month on its first Arm-based Windows PCs: RTX Spark machines from more than half a dozen manufacturers start shipping in October. The top configuration pairs a 20-core Grace CPU with 6,144 Blackwell RTX cores and up to 128GB of unified memory, which Nvidia rates at roughly one petaflop of local AI compute.
Two N1X versions, not one
The part inside these machines is the N1X, and it ships in two distinct tiers. That split matters, because the cheaper one is capped far lower on memory than the headline figure suggests.
- Higher tier — 20 CPU cores, 6,144 Blackwell RTX cores, memory configurable up to 128GB. Goes into both compact desktops and laptops.
- Lower tier — 18 CPU cores, 5,120 Blackwell RTX cores, offered with 24GB or 32GB. Laptops only, according to the configuration details circulated after the announcement.
Both tiers use a unified memory pool shared between the CPU and GPU rather than separate system RAM and video memory, which is why the 128GB number is quoted as a single figure. Nvidia has not published memory bandwidth for either variant.
The machines announced so far
Lenovo used the show to name two systems: the Yoga Pro 9n laptop and the Yoga 9n 2-in-1. Acer showed a compact small-form-factor desktop built around the chip. Nvidia says those join six manufacturers already lined up to ship in October, and reporting from the show lists Asus, Dell, Gigabyte, HP, MSI and Microsoft among the wider partner group.
One claim worth flagging as unverified: coverage from the show says RTX Spark laptops will hold roughly the same performance on battery as when plugged in, with no power-saving throttle. Nvidia has not put a figure on that in its own material, and nobody has run an independent test yet.
What Nvidia still has not said: the price
No pricing has been announced for any RTX Spark system, in any market, and no manufacturer has attached a figure to a named model. Availability is described as a global OEM launch rather than a staged regional rollout, but until the partners post their own listings there is no way to know what these cost where you live. Anyone budgeting now is guessing.
The timing lands in a rough stretch for PC buyers. Memory prices have been climbing all year and have already pushed up the cost of phones and budget graphics cards, and a machine specced to 128GB of unified memory is exposed to exactly that pressure.
A free tool you can use before any of this ships
The more immediately useful part of the announcement needs no new hardware. Nvidia PAIR — Personal AI Router — is a free, open-source tool now in beta for Windows, macOS and Linux. It finds compatible machines on your local network and routes AI inference requests to whichever one has spare capacity, so a desktop sitting idle can take work off a laptop.
Supported hardware is broader than you might expect: GeForce RTX 20 Series and newer, RTX PRO workstations, DGX Spark, and Apple silicon from the M4 onward. That last one means a Mac can join the pool.
Nvidia also published throughput gains from software optimisation alone: up to 1.9x higher throughput in llama.cpp on a GeForce RTX 5090, 1.2x in vLLM on an RTX PRO 6000 Blackwell, and up to 1.4x across two clustered DGX Spark units. Those are existing cards getting faster, not RTX Spark numbers.
What to watch between now and October
Three things will decide whether these are worth waiting for. First, the price — nothing else can be judged without it. Second, which tier each laptop actually ships with, since an 18-core, 32GB machine and a 20-core, 128GB machine will be sold under similar branding. Third, how well Windows software runs on Arm hardware, the same question that has followed every Arm-based Windows PC to date. Nvidia's own mobile graphics work, like the on-device AI push we covered in Qualcomm's Adreno Neural Fusion, shows how fast this part of the market is moving.
Frequently Asked Questions
When do Nvidia RTX Spark PCs go on sale?
Nvidia said at IFA 2026 that the first RTX Spark Windows PCs start shipping in October 2026. More than half a dozen manufacturers are involved, including Lenovo and Acer. Nvidia has not given a specific day or a staggered regional schedule.
How much does an Nvidia RTX Spark PC cost?
No price has been announced for any RTX Spark laptop or desktop, in any market. Nvidia did not disclose pricing at IFA and no manufacturer has attached a figure to a named model. Expect prices to appear when partners such as Lenovo and Acer publish their own listings closer to the October launch.
What is the difference between the two RTX Spark N1X configurations?
The higher tier has a 20-core CPU, 6,144 Blackwell RTX cores and can be configured with up to 128GB of unified memory, and it goes into both desktops and laptops. The lower tier has an 18-core CPU, 5,120 Blackwell RTX cores and is offered with 24GB or 32GB, and it is limited to laptops. Both share a single unified memory pool between the CPU and GPU.
Which RTX Spark laptops have been announced?
Lenovo named the Yoga Pro 9n laptop and the Yoga 9n 2-in-1, and Acer showed a compact small-form-factor desktop. Nvidia says these join six manufacturers already set to ship in October. Asus, Dell, Gigabyte, HP, MSI and Microsoft have also been reported as partners.
What is Nvidia PAIR and what hardware does it need?
PAIR, or Personal AI Router, is a free open-source tool that finds compatible PCs on your local network and sends AI inference work to whichever machine has spare capacity. The beta runs on Windows, macOS and Linux. It supports GeForce RTX 20 Series and newer, RTX PRO workstations, DGX Spark, and Apple silicon from the M4 onward.
