Nvidia announces RTX spark as ‘the most efficient pc chip ever built’

At a glance:

• Nvidia unveils RTX Spark, an Arm‑based chip that combines CPU, GPU and up to 128 GB of unified memory for thin‑and‑light laptops.
• The flagship version matches a DGX Spark GB10 die with 20 CPU cores, 6,144 GPU cores and 128 GB LPDDR5X, while lower‑tier parts will start at 16 GB RAM.
• Eight premium laptops are confirmed for fall 2024, including Microsoft’s Surface Laptop Ultra, with more than 30 laptops and 10 desktops from partners such as Acer, Asus, Dell, Gigabyte, HP, MSI and Lenovo slated for later this year.

What Nvidia announced

Nvidia said it is moving from a pure graphics‑card supplier to a full‑stack consumer PC chipmaker, joining the ranks of Intel, AMD, Apple and Qualcomm. The company introduced the RTX Spark family, built on an Arm‑based silicon design that merges CPU, GPU and AI accelerators on a single die. The announcement came after months of speculation and follows the DGX Spark “personal AI supercomputer” released last year, which used the same GB10 chip.

Technical specifications

The flagship RTX Spark is spec‑to‑spec identical to the DGX Spark GB10: 20 high‑performance CPU cores, 6,144 GPU cores and up to 128 GB of LPDDR5X unified memory. Nvidia says lower‑tier variants will ship with as little as 16 GB of RAM, targeting more price‑sensitive segments. The chip is fabricated on TSMC’s 3‑nm process in partnership with MediaTek, though the company declined to disclose manufacturing locations. Power consumption is said to range from “low, low single‑digit” watts up to 80 W at full load, promising all‑day battery life on thin‑and‑light laptops.

Partner ecosystem

Nvidia confirmed eight specific laptop models for the initial launch, the most notable being Microsoft’s Surface Laptop Ultra, which the Surface boss described as “the most powerful thing we’ve ever made.” In addition, the company said partners are already developing over 30 laptops and more than 10 desktops, naming Acer, Asus, Dell, Gigabyte, HP, MSI and Lenovo as participants. This broad OEM support suggests Nvidia is aiming for a wide price‑point coverage, from premium ultrabooks to more affordable thin‑and‑light devices.

Software and AI capabilities

Because RTX Spark runs Arm‑based code, Windows applications built for x86 must use Microsoft’s Prism emulation layer. Nvidia highlighted that Microsoft has been optimizing Windows and Prism for Arm for years, and that the chip’s native AI and graphics engines will mitigate typical emulation overhead. Nvidia also showcased use‑case scenarios: an esports streamer could automate lighting, microphone muting and broadcast mode; a designer could ask Adobe software to turn a sketch into a full‑rendered image and an AI‑generated video; developers could have an AI agent monitor GitHub and fix QA issues autonomously. All AI processing happens locally, keeping data private and avoiding token costs.

Performance claims and market positioning

While Nvidia did not provide benchmark charts, Mark Aevermann claimed the chip’s graphics performance is roughly comparable to an RTX 5070 mobile GPU, and the CPU portion is “competitive with anything else out there in the Windows space.” Nvidia also said the chip can render a 90 GB 3D scene, edit 12K video, or run a graphically intensive game—Indiana Jones and the Great Circle—at 100 fps in 1440p on a 14 mm‑thick laptop without being plugged in. These statements position RTX Spark as a potential “personal computing paradigm where AI is the UX,” aiming to replace traditional mouse‑and‑keyboard interactions with conversational commands.

Open questions and outlook

Several details remain vague. Pricing has not been disclosed, though Nvidia indicated the first batch targets premium price points. Battery‑life estimates are qualitative, and the 80 W ceiling suggests heavy workloads could drain a laptop in about an hour. Nvidia declined to comment on Linux driver support, US versus overseas manufacturing, or the possibility of pairing RTX Spark with discrete GPUs in larger desktops. Nonetheless, the company’s strategy mirrors Apple’s 2020 silicon rollout—initially opaque, then validated by real‑world performance.