Blackwell (microarchitecture)

Blackwell
LaunchedQ4 2024
Designed byNvidia
Manufactured by
Fabrication processTSMC 4NP (Datacenter[1])
TSMC 4N (Consumer[2])
Codename(s)GB100
GB20x
Product Series
Desktop
Specifications
Memory supportGDDR7 (Consumer)
HBM3e (Datacenter)
PCIe supportPCIe 5.0 (Consumer)
PCIe 6.0 (Datacenter)
Supported Graphics APIs
DirectXDirectX 12 Ultimate (Feature Level 12_2)
Direct3DDirect3D 12
Shader ModelShader Model 6.8
OpenCLOpenCL 3.0 (64-bit only, 32-bit support removed)[3]
OpenGLOpenGL 4.6
VulkanVulkan 1.4
Supported Compute APIs
CUDACompute Capability 10.x (64-bit only, 32-bit support removed)[3]
Compute Capability 12.x (64-bit only)[3]
DirectComputeYes
Media Engine
Encoder(s) supportedNVENC
History
PredecessorAda Lovelace (consumer)
Hopper (datacenter)
SuccessorRubin

Blackwell is a graphics processing unit (GPU) microarchitecture developed by Nvidia as the successor to the Hopper and Ada Lovelace microarchitectures.

Named after statistician and mathematician David Blackwell, the name of the Blackwell architecture was leaked in 2022 with the B40 and B100 accelerators being confirmed in October 2023 with an official Nvidia roadmap shown during an investors presentation.[4] It was officially announced at Nvidia's GTC 2024 keynote on March 18, 2024.[5]

  1. ^ "NVIDIA Blackwell Architecture Technical Brief". NVIDIA. Retrieved February 2, 2025.
  2. ^ "NVIDIA RTX BLACKWELL GPU ARCHITECTURE" (PDF). NVIDIA. Retrieved February 2, 2025.
  3. ^ a b c Deakin, Daniel R (February 26, 2025). "Strange GeForce RTX 50-series benchmark issues solved but new problem for gamers arises". Notebookcheck. Retrieved February 26, 2025.
  4. ^ "Nvidia Corporation - Nvidia Investor Presentation October 2023". Nvidia. Retrieved March 19, 2024.
  5. ^ "Nvidia Blackwell Platform Arrives to Power a New Era of Computing". Nvidia Newsroom. Retrieved March 19, 2024.

From Wikipedia, the free encyclopedia · View on Wikipedia

Developed by Nelliwinne