AI hardware
Google TPU v3
FP16 throughput
1.2×10¹⁴ FLOP/s
Memory
29.8 GB
TDP
450 W
Google TPU v3 is an AI accelerator designed by Google (TPU), released in May 2018. It is built on a 16 nm process.
It delivers 1.2×10¹⁴ FLOP/s of tensor FP16/BF16 throughput. It carries 29.8 GB of memory at 838 GB/s of bandwidth. Its thermal design power is 450 W.
Full record
- Manufacturer
- Type
- TPU
- Release date
- May 18, 2018
- Tensor FP16/BF16
- 1.2×10¹⁴ FLOP/s
- Memory
- 29.8 GB
- Memory bandwidth
- 838 GB
- TDP
- 450 W
- Process node
- 16 nm
More from Google
SourceEpoch AI, 'Machine Learning Hardware'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/data/machine-learning-hardware. Licensed under CC BY 4.0.
← All ai hardware