Subscribe

NVIDIA RTX Spark

NVIDIA’s first Windows PC superchip pairs a Blackwell RTX GPU with a 20-core Grace CPU and went on sale in laptops on 16 Oct 2026.

RTX Spark is a single superchip for Windows laptops and compact desktops that joins an NVIDIA Blackwell RTX GPU (6,144 CUDA cores, fifth-generation Tensor Cores with FP4) to a 20-core NVIDIA Grace CPU over NVLink-C2C. MediaTek collaborated with NVIDIA on the custom CPU design. NVIDIA announced it on 31 May 2026 alongside Microsoft and says it delivers up to 1 petaflop of FP4 AI performance with up to 128GB of unified memory.

Laptops from ASUS, Dell, HP, Lenovo, Microsoft and MSI went on sale globally on 16 Oct 2026. NVIDIA has not published the process node, memory bandwidth, or whether the 1 petaflop figure is sparse or dense.

About NVIDIA All AI chips

Silicon from NVIDIA

11 chips on record, newest first.

  1. 2026

    NVIDIA RTX Spark

    Not stated by NVIDIA NVIDIA Blackwell RTX GPU, 6,144 CUDA cores, fifth-generation Tensor Cores with FP4 precision; 20-core NVIDIA Grace CPU via NVLink-C2C

    up to 1 petaflop of FP4 AI performance

    Absolute
  2. 2026

    NVIDIA Groq 3 LPX (LP30 LPU)

    Not stated by NVIDIA LP30 language processing unit, 500 MB on-chip SRAM per chip, 256 LPUs per rack

    315 PFLOPS AI inference per LPX rack

    Absolute
  3. 2026

    NVIDIA Rubin GPU

    Not stated by NVIDIA Third-gen Transformer Engine, 336bn transistors across two dies, NVLink 6 at 3.6 TB/s

    Rubin GPU delivers 50 petaflops of NVFP4 compute for AI inference

    Absolute
  4. 2026

    NVIDIA Vera Rubin Superchip

    Not stated by NVIDIA One Vera CPU (88 custom Olympus Arm cores) joined to two Rubin GPUs over NVLink-C2C

    100 PFLOPS NVFP4 inference

    Absolute
  5. 2025

    NVIDIA Rubin CPX

    Not stated by NVIDIA Monolithic GPU for long-context inference, 128 GB GDDR7, hardware attention acceleration

    up to 30 petaflops of compute with NVFP4 precision

    Absolute
  6. 2025

    NVIDIA B300 (Blackwell Ultra)

    TSMC 4NP Enhanced fifth-gen Tensor Cores with 2x attention-layer acceleration, 208bn transistors across two dies

    15 PetaFLOPS dense NVFP4 compute

    Absolute
  7. 2024

    NVIDIA Jetson AGX Thor (T5000)

    Not stated by NVIDIA Blackwell GPU on the Thor SoC, 130 W, 128 GB LPDDR5X

    up to 2,070 FP4 teraflops of AI compute

    Absolute
  8. 2024

    NVIDIA GB200 Grace Blackwell Superchip

    TSMC 4NP (GPU dies) One Grace CPU joined to two B200 GPUs over 900 GB/s NVLink-C2C

    40 | 20 PFLOPS NVFP4 Tensor Core

    Absolute
  9. 2024

    NVIDIA B200

    TSMC 4NP Fifth-gen Tensor Cores, second-gen Transformer Engine, 208bn transistors across two dies joined at 10 TB/s

    Blackwell delivers 2.5x its predecessor's performance in FP8 for training, per chip, and 5x with FP4 for inference.

    Relative only
  10. 2023

    NVIDIA H200

    Not stated by NVIDIA Hopper GPU, the first with HBM3e (141 GB)

    3,958 TFLOPS FP8 Tensor Core (H200 SXM)

    Absolute
  11. 2022

    NVIDIA H100

    TSMC 4N Fourth-gen Tensor Cores with the first Transformer Engine (FP8), 80bn transistors

    3,958 teraFLOPS FP8 Tensor Core (H100 SXM)

    Absolute

Compare against every other vendor in the silicon tracker

NVIDIA in the news