Jump to content

Draft:GeForce 50 series

From Wikipedia, the free encyclopedia

GeForce 50 series
Release dateSeptember 2025
Manufactured byTSMC
Designed byNvidia
Marketed byNvidia
CodenameBlackwell
ModelsGeForce RTX series
Transistors
  • (GB207)
  • (GB206)
  • (GB205)
  • (GB203)
  • (GB202)
Fabrication processTSMC 4NP
Cards
Entry-levelRTX 5050
RTX 5050 Ti
Mid-rangeRTX 5060
RTX 5060 Ti
High-endRTX 5070
RTX 5070 Ti
RTX 5080
RTX 5080 Ti
EnthusiastRTX 5090
RTX 5090 Ti
API support
DirectXDirect3D 12.0 Ultimate (feature level 12_2)
Shader Model 6.8
OpenCLOpenCL 3.0
OpenGLOpenGL 4.6
VulkanVulkan 1.3
History
PredecessorGeForce 40 series
SuccessorGeForce 60 series
Support status
Unreleased

The GeForce 50 series is an announced family of graphics processing units developed by Nvidia, succeeding the GeForce 40 series. The series was announced in June 2023 during the Nvidia's AI training demonstration.[1]

Details

[edit]

Architectural highlights of the Blackwell architecture include the following:

  • CUDA Compute Capability 10.x[2]
  • TSMC 4NP process (custom designed for Nvidia)
  • Fifth-generation Tensor Cores with FP4, FP8, FP16, bfloat16, TensorFloat-32 (TF32) and sparsity acceleration[3]
  • ?-generation Ray Tracing Cores, along with concurrent ray tracing, shading and compute
  • PCI Express 5.0
  • GDDR7 SDRAM memory support[4]
  • DisplayPort 2.1a UHBR20 80Gbps native implementation

Products

[edit]

Desktop

[edit]

Mobile

[edit]

See also

[edit]

References

[edit]
  1. ^ "NVIDIA's Ada Lovelace successor, the GeForce RTX 50 series, set to launch in 2024 and 2025". TweakTown. 2023. Retrieved 2023-07-24.
  2. ^ "CUDA C++ Programming Guide". NVIDIA Developer Zone.
  3. ^ https://www.nvidia.com/en-us/data-center/tensor-cores/#blackwell
  4. ^ https://www.jedec.org/news/pressreleases/jedec-publishes-gddr7-graphics-memory-standard
[edit]