NVIDIA Expands RTX Spark Roadmap Through 2030

NVIDIA unveiled a multi-generation roadmap for its RTX Spark PC platform at Computex 2026, showing follow-up architectures through 2030, sources report. PC Gamer and Tom's Hardware report the slide places Grace/Blackwell RTX Spark in 2026, a Vera Rubin Spark generation in 2027/2028 with LPDDR6 memory, and a Rosa Feynman Spark (sometimes reported as Rosa or Feynman) in 2029/2030. Tom's Hardware and PC Gamer note NVIDIA CEO Jensen Huang displayed the roadmap during his keynote. PC Gamer reports the Blackwell RTX Spark package includes 6144 GPU cores, 20 Grace CPU cores, and up to 128 GB of unified LPDDR5X memory. Multiple outlets including Wccftech and Heise add that Vera Rubin will adopt LPDDR6, while VideoCardz, ServeTheHome, and others corroborate the multi-generation plan.
What happened
NVIDIA presented a roadmap for its RTX Spark PC platform at Computex 2026, showing multi-generation releases stretching to 2030, multiple outlets report. PC Gamer and Tom's Hardware report the slide places the initial Grace/Blackwell RTX Spark in 2026, a Vera Rubin Spark generation in 2027/2028, and a Rosa Feynman Spark generation in 2029/2030. Tom's Hardware reports NVIDIA committed that every future generation of its platform will include a Spark chip. PC Gamer reports the Blackwell RTX Spark package includes 6144 GPU cores, 20 Grace CPU cores, and up to 128 GB of unified LPDDR5X memory. Wccftech and Heise note the Vera Rubin roadmap entry explicitly references support for LPDDR6 memory.
Technical details
Editorial analysis - technical context
Moving from LPDDR5X to LPDDR6 typically implies higher memory bandwidth and efficiency for mobile and compact systems. Industry hardware vendors adopting LPDDR6 aim for improved sustained throughput and lower energy per bit, which matters for unified memory architectures where CPU and GPU share the same pool.
The published Blackwell Spark hardware figures (as reported by PC Gamer) and Wccftech's mention of a 600 GB/s unified memory bandwidth for Spark-class devices point to system designs prioritizing on-package memory bandwidth rather than discrete DRAM channels. For machine learning workloads that rely on large shared tensors and frequent CPU-GPU transfers, those bandwidth characteristics materially affect real-world throughput and latency.
Context and significance
Public reporting frames NVIDIA's roadmap as an explicit push into Windows on Arm and a multi-year commitment to Spark platforms, with Tom's Hardware and Wccftech describing the move as part of a broader effort to bring agentic AI capabilities to laptops and desktops. For OEMs and software vendors, a multi-generation hardware roadmap reduces uncertainty around platform continuity and can influence decisions about driver investments and system designs.
Industry context
The choice to name CPU architectures (Vera Rubin, Rosa Feynman) and to indicate a memory-generation change signals a hardware cadence that could affect timelines for OS tuning, compiler toolchains, and ISV optimization. Historically, shifts in memory standards and CPU microarchitectures require updates to power delivery, thermal design, and memory controllers, which ripple into system validation cycles.
What to watch
For practitioners
monitor OEM design wins and announced notebooks/mini-PCs that adopt the Blackwell Spark parts; early silicon samples and thermal/benchmark data will reveal how the unified memory bandwidth performs in mixed CPU/GPU ML workloads. For the memory ecosystem: track LPDDR6 module availability, JEDEC finalization status, and vendor pricing, since those factors shape time to market. For software: watch for Microsoft and ISV support signals for Windows on Arm optimizations targeting the Spark unified memory model.
Key Points
- 1Industry context: NVIDIA published a multi-generation RTX Spark roadmap through 2030, giving OEMs a visible hardware cadence for Windows on Arm adoption.
- 2Industry context: The move from LPDDR5X to LPDDR6 implies higher sustained memory bandwidth and efficiency, important for unified CPU/GPU ML workloads.
- 3Industry context: Public roadmap visibility lowers platform uncertainty for ISVs and OEMs, accelerating optimizations for agentic AI use cases on laptops and mini-PCs.
Scoring Rationale
A multi‑generation hardware roadmap affects OEM planning, memory ecosystem timelines, and developer optimization work for Windows on Arm. It is notable for practitioners integrating AI workloads into client devices, but it is not a frontier model or paradigm shift.
Sources
Public references used for this report.
View 6 more sources
- 04Nvidia has released the RTX Spark roadmap and with it, a rare hint of when to expect new RTX cardspcgamer.com
- 05Nvidia shows PC roadmap until 2030heise.de
- 06NVIDIA confirms RTX Spark roadmap with Rubin in 2027 and Rosa Feynman in 2029videocardz.com
- 07NVIDIA Computex 2026 Keynote Live Coverage - ServeTheHomeservethehome.com
- 08RTX Spark Superchip: Nvidia greift AMD und Intel im Windows-PC-Markt ancomputerbase.de
- 09RTX Spark To Jumpstart New Lineup As NVIDIA Will Expand With Vera Rubin Spark & Rosa Feynman Spark Architectures With LPDDR6 RAM Support, Arriving 2027 Onwardswccftech.com
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems

