Google TPU v5e
The smaller-memory TPU branch for economical training and serving. Eight chips provide eight 16 GB budgets, not one shared 128 GB allocation.
HBM
Published peak, not measured application throughput
The current v5e table labels memory bandwidth as 800 GiB/s, while older summaries use roughly 0.82 TB/s. This table converts the current label to about 0.859 TB/s and keeps the source visible.
When this is a sensible choice
Start here if…
Start with v5e for a supported smaller model and an explicit sharding plan. An 8B BF16 weight floor consumes 16 GB before runtime; a single chip is already too tight for that naive plan.
Choose another configuration if…
Use v5p or later hardware when per-chip memory or communication dominates. A training allocation and a serving allocation also have different availability and latency goals.
Specifications with their boundaries attached
- Architecture
- TPU v5e
- Memory
- 16 GB HBM
- Memory bandwidth
- 0.86 TB/s per accelerator
- Peak compute
- 197 TFLOPS BF16; 393 TOPS INT8
- Scale-up interconnect
- 400 GB/s bidirectional per chip
- Host attachment
- Managed TPU host; 2D torus, 256-chip pod
- Power
- Not published in checked per-chip reference
- Partitioning
- Slice/chiplet allocation; not NVIDIA MIG
- Catalogue status
- Documented product
Compute figures are theoretical peaks at the stated precision. Dense and structured-sparse rates must not be mixed. Bandwidth labelled bidirectional combines both directions. See the source documents.
Three bandwidths, three different jobs
400 GB/s bidirectional per chip
DCN between slices
Where it appears in provider documentation
| Provider / machine | Network scope | What changes the decision |
|---|---|---|
Google Cloud TPU v5e ↗ | ICI within a slice; DCN between slices | Select a documented slice shape and supported region. Chip, chiplet, TensorCore, host and VM counts are not interchangeable. |
Model fit and software support
These publisher or serving-engine documents mention this hardware family. They have not been reproduced on our machines.
No model-specific recipe in our reviewed set certifies this exact hardware. The memory calculator can narrow candidates, but it cannot establish software support. Read the model register.
What is the memory floor?
Start with total parameters, then add the memory the workload needs. This arithmetic does not certify a serving configuration. All output sizes below are decimal GB.
Override device capacity with the memory exposed by your allocation, particularly for cloud B300 and partitioned devices. The starting 32 GB budget and 15% reserve are editable teaching assumptions. They are not measurements for the selected model. Mixed-precision tensors, quantization scales, vision encoders and draft models can increase the weight payload.
Raw weights only
320B × 8 bits ÷ 8
Budget per device
16 GB × (1 − 15%)
Arithmetic lower bound
round up ((weights + 32) ÷ budget)
This assumes perfectly balanced sharding. A result of three does not prove that a three-device parallel layout is supported. Check layer/expert divisibility, actual allocatable memory, precision kernels and the fabric before renting.
320B is the model card’s total; 18B active is not its storage size. The vLLM recipe reports about 306 GiB for the native FP8 checkpoint. Hopper requires BF16 KV for this model; the documented ROCm path targets gfx950, not every Instinct GPU. Read the model source ↗
For full training, also budget gradients, optimizer states, activations and communication buffers. The inference weight estimate above is insufficient.
Tokens per second, TTFT and TPOT are not certified for this hardware in our reference. Use the benchmark checklist to compare an exact model, software revision and workload.
Nearby memory capacities, different tradeoffs
These are comparison candidates selected by memory capacity, not performance rankings or drop-in replacements.
Google TPU v6e (Trillium)
Trillium doubles v5e’s nominal memory to 32 GB and increases BF16 compute. It retains a 256-chip, 2D pod design.
Google TPU v4
A mature training-oriented TPU with 32 GiB per chip. Existing XLA jobs can be a better starting point than porting a CUDA-only workload.
Google TPU v5p
A large-pod training TPU with 95 GiB per chip, much more memory than v5e. The suffix identifies a different design, not a larger v5e VM.
Sources and review scope
Manufacturer and cloud documentation checked 2026-09-12. We reviewed specifications and the stated product boundaries; we did not run training or serving benchmarks on this device. Cloud memory and availability can differ by configuration.
- Google TPU v5e architecture ↗ · checked 2026-09-12
Cite this reference
AI Infra Interviews. Google TPU v5e: specifications and workload fit. Reviewed . https://aiinfrainterviews.com/hardware/tpu-v5e
Include your access date when citing a changing specification. Link to the specification section for hardware figures or the explanation for a sizing or workload decision.
Research method and limits
We compile manufacturer specifications, cloud documentation, model cards and serving recipes. The reference preserves source units and distinguishes individual devices from nodes and racks. Conflicting figures and unknown fields remain labelled.
Our contribution is the comparison, unit reconciliation, worked arithmetic and workload explanation. Published peaks are vendor specifications. Calculator results are estimates under the displayed assumptions. Neither is a measurement from our own accelerator lab.
For a manufacturer’s specification, consult the original source documents. Cite our page when using its analysis, and retain primary-source attribution for underlying figures. This curated reference does not establish market share, live availability or a universal performance ranking.
Found a discrepancy? Send a correction with the page URL, exact variant, disputed figure and supporting primary document.