Google TPU v5e Pod

pod
256× TPU v5e
2D torus ICI, 400 GBps bidirectional per chip
2023
BF16
50.4
PFLOPS
INT8
100.6
PFLOPS
Power
kW Total
Memory
4194
GB Total

A full TPU v5e pod is 256 chips in a 2D torus. Per chip Google publishes 197 TFLOPS bf16 and 393 TOPS int8, giving 50.4 PFLOPS and 100.6 POPS across the pod. Memory is 256 x 16 GB HBM at 800 GiBps per chip, with 400 GBps of bidirectional inter-chip interconnect per chip. v5e is Google's cost-and-efficiency oriented generation, which is why its pod is a thirty-fifth the size of the v5p pod launched alongside it. Google publishes no pod-level compute or power figure; the compute above is chip count times published per-chip peak, the convention Google's own published v4 and Ironwood pod figures both confirm. No sparse figures are published, so all figures are dense.

We will point you at suppliers who have it. Free, and no signup.

BF16
50.43
PFLOPS
INT8
100.61
PFLOPS

System Details

GPU Configuration

GPU Count: 256 GPUs
Architecture: TPU
Interconnect: 2D torus ICI, 400 GBps bidirectional per chip

System Specifications

Form Factor: pod
Total Power:
Total Memory: 4194 GB
Memory Bandwidth: 204800 GB/s

Precision Performance Breakdown

PrecisionSystem PerformancePer GPUEfficiency
50.432 PFLOPS 197.0 TFLOPS
100.608 PFLOPS 393.0 TFLOPS

All figures are dense. Vendors commonly headline the number, which is twice the dense one.

Powered by Google TPU v5e

This system utilizes 256 × Google TPU v5e 16GB GPUs, each delivering exceptional performance for AI and HPC workloads.

Per GPU TDP

Per GPU Memory

16 GB

Process Node

Architecture

TPU

Documentation & Resources

Official Datasheet

Google TPU v5e Pod technical specifications

Download PDF

Cloud TPU v5p

Google • 2023-12-07

View Document ↗

Typical Use Cases

AI/ML Training
High-Performance Computing
Data Analytics

The Google TPU v5e Pod runs 256× TPU v5e GPUs over 2D torus ICI, 400 GBps bidirectional per chip, delivering 50.4 PFLOPS BF16 dense.