DUAL-DIE GPU ARCHITECTURE
Dual-die GPU architecture with a high-bandwidth chip-to-chip interconnect
Dual-die GPU architecture with a high-bandwidth chip-to-chip interconnect
Described as a two-reticle design manufactured on TSMC 4NP
AI & HPC · Data Center
Public source
NVIDIA Blackwell is a publicly named data-center GPU architecture for AI and HPC accelerators.
Silicon Showcase illustrative visual
Illustrative representation based on publicly available technical information.
Blackwell is NVIDIA’s current data-center GPU architecture, built as a dual-die GPU with a 10 TB/s chip-to-chip interconnect. Official materials describe a two-reticle design manufactured on TSMC 4NP.
Official NVIDIA newsroom materials present Blackwell as a dual-die GPU with a 10 TB/s chip-to-chip interconnect and a two-reticle design manufactured on TSMC 4NP. This listing records the architecture rather than individual accelerator SKUs. Products such as B200, GB200 or Blackwell Ultra remain implementations of this architecture rather than separate listings. NVLink and NVLink-C2C, already listed, are interconnect technologies used with NVIDIA GPUs, not this compute architecture. Transistor counts, die-to-die bandwidth and process-node details should be confirmed on NVIDIA’s current Blackwell pages.
Key characteristics associated with this technology.
Dual-die GPU architecture with a high-bandwidth chip-to-chip interconnect
Described as a two-reticle design manufactured on TSMC 4NP
Positioned for data-center AI training and inference accelerators
Distinct from NVLink scale-out fabric and NVLink-C2C package links
Explore other semiconductor technologies.
Ultra-low-power connected-device system-on-chip that Sensesemi describes with wireless connectivity, a microcontroller, and an edge AI accelerator.
Consumer ElectronicsMedicalView Technology
Kulicke & Soffa APTURA is a thermo-compression bonder for advanced packaging, not a wire bonder catalog and not OSAT assembly.
AI & HPCData CenterView Technology
Photronics advanced IC photomasks are lithography reticles for integrated circuits, not a foundry node and not FPD masks.
AutomotiveAI & HPCView Technology
AlphaSGT is Alpha and Omega’s shielded-gate trench MOSFET technology, not a GaN HEMT and not a power-module substrate.
AI & HPCData CenterView Technology
