Google TPU v3 32GB vs Google TPU v4 32GB
The TPU v4 delivers 2.2x the BF16 throughput of the TPU v3 (275 vs 123 TFLOPS dense).
Google TPU v3 32GB
Full specs →Google TPU v4 32GB
Full specs →Specifications
Performance (TFLOPS)
FLOPS by Precision
What actually differs
The TPU v4 is the newer part: TPU, launched in 2020, against the TPU v3's TPU from 2018. Newer architectures typically add lower-precision formats and better throughput per watt, so check the precision rows your workload actually uses.
Memory is 32 GB against 32 GB, fed at 900 GB/s versus 1.2 TB/s. More memory per GPU means larger models fit before you have to shard across cards, which often matters more than raw TFLOPS for inference.
For LLM training and inference, weight the FP8 and FP16 rows and memory capacity most heavily. For scientific and HPC workloads, the FP64 row is the one to read.
Frequently asked questions
Is the TPU v3 faster than the TPU v4?
At BF16 precision the TPU v4 reaches 275 TFLOPS dense against 123 TFLOPS for the TPU v3. The performance table on this page lists every published precision for both GPUs.
Which has more memory, the TPU v3 or the TPU v4?
Both GPUs carry 32 GB of VRAM. Memory bandwidth is 900 GB/s for the TPU v3 and 1.2 TB/s for the TPU v4.
Get Comparison Updates
New GPUs added weekly. Be the first to see how they compare.