Description
For the GB300, Nvidia is shifting to a more modular approach. Instead of delivering a fully assembled motherboard, Nvidia is said to provide the B300 GPU on an SXM Puck module, the Grace CPU in a separate BGA package, and the hardware management controller (HMC) from Axiado. Customers now source the remaining motherboard components themselves, and the CPU memory uses standard SOCAMM memory modules that are obtainable from various vendors. Nvidia continues to provide the switch tray and copper backplane as before. The reuse eliminates the need for a complete redesign, streamlining production processes and reducing risk.
With GB200, Nvidia provides the complete Bianca motherboard, which includes the B200 GPU, Grace CPU, 512GB of LPDDR5X memory, and power delivery components — all integrated on a single printed circuit board (PCB). Nvidia also supplies the switch tray and copper backplane for this system.
As the latest member of the NVIDIA Blackwell architecture family, the NVIDIA Blackwell Ultra GPU builds on core innovations to accelerate training and AI reasoning. It fuses silicon innovations with new levels of system-level integration, delivering next-level performance, scalability, and efficiency for AI factories and the large-scale, real-time AI services they power.
With its energy-efficient dual-reticle design, high bandwidth and large-capacity HBM3E memory subsystem, fifth-generation Tensor Cores, and breakthrough NVFP4 precision format, Blackwell Ultra is raising the bar for accelerated computing. This in-depth look explains the architectural advances, why they matter, and how they translate into measurable gains for AI workloads.
Dual-reticle design: one GPU
Blackwell Ultra is composed of two reticle-sized dies connected using NVIDIA High-Bandwidth Interface (NV-HBI), a custom, power-efficient die-to-die interconnect technology that provides 10 TB/s of bandwidth. Blackwell Ultra is manufactured using TSMC 4NP and features 208B transistors–2.6x more than the NVIDIA Hopper GPU—all while functioning as a single, NVIDIA CUDA-programmed accelerator. This enables a large increase in performance while also maintaining the familiar CUDA programming model that developers have enjoyed for nearly two decades.
Benefits
Unified compute domain: 160* Streaming Multiprocessors (SMs) across two dies, providing 640 fifth-generation Tensor Cores with 15 PetaFLOPS dense NVFP4 compute.
Full coherence: Shared L2 cache with fully coherent memory accesses.
Maximum silicon utilization: Peak performance per square millimeter.
GB300 Liquid Cooling Revolution
The GB300’s formidable performance comes with a steep power demand. With a single-GPU TDP soaring to 1,400W and full-rack system power consumption reaching a staggering 125-130kW, NVIDIA has adopted a fully liquid-cooled thermal solution to tackle this challenge—significantly advancing liquid cooling technology.
While the GB300 retains the GB200’s design for components like manifolds, Cooling Distribution Units (CDUs), and cartridges, its core cooling module has been re-engineered to deliver breakthroughs in efficiency.
Sourced from NADDOD.com
