vs

Google TPU v5p 95GB vs NVIDIA H100 PCIe 80GB

The H100 delivers 3.3x the FP8 throughput of the TPU v5p (1,513 vs 459 TFLOPS dense). The TPU v5p counters with 15 GB more memory (95 GB vs 80 GB).

TPU v5p: TPU v5p, 2023 H100: Hopper, 2022
FP8 dense lead
3.3x
H100: 1,513 vs 459 TFLOPS
Memory
95 vs 80 GB
15 GB more for the TPU v5p
Bandwidth
2.8 TB/s vs 2.0 TB/s
TPU v5p moves data faster
TDP
450 W vs 350 W
H100 draws 100 W less

Specifications

TPU v5p
H100
Architecture
TPU v5pHopper
Launch Year
20232022
Form Factor
CustomPCIe
Memory
95 GB80 GB
Memory Bandwidth
2.8 TB/s2.0 TB/s
TDP
450 W350 W
Process Node
5nm4nm

Performance (TFLOPS)

TPU v5p
H100
FP64
No verified data available26 TFLOPS
FP32
No verified data available51 TFLOPS
TF32
No verified data available378 TFLOPS
sparse not published
BF16
459 TFLOPS 757 TFLOPS
sparse not published
FP16
No verified data available757 TFLOPS
sparse not published
FP8
459 TFLOPS 1,513 TFLOPS
sparse not published
FP6
No verified data available No verified data available
FP4
No verified data available No verified data available
INT8
No verified data available1,513 TOPS
sparse not published

FLOPS by Precision

What actually differs

The TPU v5p is the newer part: TPU v5p, launched in 2023, against the H100's Hopper from 2022. Newer architectures typically add lower-precision formats and better throughput per watt, so check the precision rows your workload actually uses.

Dense throughput for the TPU v5p against the H100: FP8 459 vs 1,513 TFLOPS. Sparse figures, where the vendor publishes them, appear under each dense number in the performance table.

Memory is 95 GB against 80 GB, fed at 2.8 TB/s versus 2.0 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 v5p faster than the H100?

At FP8 precision the H100 reaches 1,513 TFLOPS dense against 459 TFLOPS for the TPU v5p. The performance table on this page lists every published precision for both GPUs.

Which has more memory, the TPU v5p or the H100?

The TPU v5p carries 95 GB of memory versus 80 GB for the H100. Memory bandwidth is 2.8 TB/s for the TPU v5p and 2.0 TB/s for the H100.

How much power do the TPU v5p and the H100 draw?

The TPU v5p is rated at 450 W TDP and the H100 at 350 W. On FP32 throughput per watt, the H100 is the more efficient part.

Can I rent the TPU v5p or the H100 in the cloud?

Yes. Live cloud listings tracked by Flopper start at $1.99 per GPU hour. The rental pricing section on this page lists current providers and rates for both GPUs.

Where to Rent

NVIDIA H100 PCIe 80GB

ProviderConfigurationPrice/GPU-hrChecked
RunPod
1× H100 PCIe Community $1.99 4h ago View →
Hyperstack
1× H100 PCIe Spot $2.00 4h ago View →
Hyperstack
1× H100 PCIe $2.50 4h ago View →
Vast.ai
1× H100 PCIe Community $2.53 4h ago View →
RunPod
1× H100 PCIe $2.89 4h ago View →
Lambda Labs
1× H100 PCIe $3.29 4h ago View →
All H100 listings and price history →

Get Comparison Updates

New GPUs added weekly. Be the first to see how they compare.