Teraflop (TFLOP)

wikipedia:Teraflop

• Google TPU Ironwood (2025): 4614 TFLOPs
• Tesla D1 chip, 362 TFLOPs
• A100 (2020): 19,5 TFLOPs
• H100 (2022)
• Nvidia T4: 8.1 TFLOPs at FP32
• L4: 30.3 NVIDIA L4 Tensor Core (2023)

Vanilla:

• Apple M1 GPU: 2,6 TFLOPs at FP32
• Apple M2 GPU: 3,6 TFLOPs at FP32

Pro:

• Apple M1 GPU: 5,2 TFLOPs at FP32
• Apple M2 GPU: 6,8 TFLOPs at FP32

Max:

• Apple M1 GPU: 10,4 TFLOPs at FP32
• Apple M2 GPU: 13,6 TFLOPs at FP32
• Apple M2 GPU: 40 Cores

Related

• GPU
• Nvidia A100
• TOPS (NPU)
• Snapdragon
• TOPS

See also

• FLOP, Exaflop, Teraflop, TFLOP, GFLOP, TOPS
• Floating point, FP4, FP6, FP8, FP16, FP32, FP64, bc, Teraflop, INT8, bfloat16 floating-point format (bf16), mpdecimal
• GPU, eGPU, OpenCL, ROCm, Nvidia, CUDA, Nvidia Tesla, AMD, Intel Arc, RTX, gpustat, Teraflop, Ethash, Pytorch, Pascal GPU, TBDR, A100, H100, TRIP, AMD Instinct MI100, AMD Radeon, LPU, TFLOP, Cloud GPUs