Description:
- GPU memory: 4 GB GDDR6
- Architecture: NVIDIA Ampere
- CUDA cores: 768
- Tensor Cores: 24 (3rd Generation)
- RT Cores: 6 (2nd Generation)
- Single-Precision performance: 2.7 TFLOPS
- RT Core performance: 5.4 TFLOPS
- FP16 Tensor performance: 21.7 TFLOPS*
- Peak INT8 Tensor performance: 43.3 TOPS**
- Memory interface: 64-bit
- Memory bandwidth: 96 GB/s
- Max. simultaneous displays: 4× 4,096 × 2,160 @ 120 Hz / 4× 5,120 × 2,880 @ 60 Hz
- System interface: PCI Express 4.0 x8
- Encode / Decode: 1× / 1× (+ AV1 Decode)
- Power consumption: 50 W
- Power connector: None (slot powered)
- Thermal solution: Active
- Form factor: Single slot, LP, 2.7" × 6.4"
- Graphics API: DirectX 12, Shader Model 6.6, OpenGL 4.6, Vulkan 1.3
- Compute API: CUDA 11.6, OpenCL 3.0, DirectCompute
* Effective FP16 TFLOPS using the sparsity feature ** Peak INT8 TOPS using the sparsity feature
