Prices Providers Compare

…

© 2026 PetaPrice

Providers GPU comparisons Architectures Price report Glossary FAQ Methodology Corrections JSON API llms.txt Privacy Imprint

Price disclaimer: All prices are indicative list prices in USD, collected automatically from provider APIs and public catalogs and refreshed about every 2 hours. They exclude taxes, storage, egress and other fees unless a provider bundles them, may be out of date, and don't guarantee availability. The provider's own website is authoritative at the time of booking. PetaPrice does not rent out compute and is not a party to any contract with a provider. FLOPS are theoretical peak values from vendor datasheets. More

GPU glossary/Compute

FP8

Also known as: E4M3, E5M2, Float8

8-bit floating point for training and inference, with roughly twice the throughput of FP16. Native on Hopper, Ada, Blackwell and AMD CDNA 3 and newer.

GPUs with FP8, best value first

Ranked by $ per PFLOP-hour of dense FP8 at each GPU's lowest on-demand price.

1Instinct MI355X288 GB$2.59/h$0.51/PFLOP·h
2Instinct MI300X192 GB$1.85/h$0.71/PFLOP·h
3B200 SXM180 GB$3.20/h$0.71/PFLOP·h
4Instinct MI325X256 GB$2.00/h$0.76/PFLOP·h
5RTX 409024 GB$0.29/h$0.89/PFLOP·h
6RTX 6000 Ada48 GB$0.65/h$0.89/PFLOP·h
7H100 SXM80 GB$1.82/h$0.92/PFLOP·h
8RTX 5000 Ada32 GB$0.49/h$0.94/PFLOP·h

Compare every GPU →

Related terms

PrecisionHow many bits each number uses. Fewer bits mean more throughput and less memory, at some cost in accuracy. The same GPU has a different peak for each precision.
FP16 and BF1616-bit formats and the standard for training and running LLMs. BF16 keeps FP32's range with less precision, which makes training more stable; it needs Ampere, CDNA 2 or newer.
FP44-bit floating point, used for quantized inference. Only Blackwell and AMD CDNA 4 run it natively.

Compute

FLOPS, TFLOPS, PFLOPSFloating-point operations per second: how much arithmetic a chip can do. 1 TFLOPS is a trillion per second, 1 PFLOPS (PetaFLOPS) is a thousand TFLOPS. PetaPrice shows all compute in PFLOPS.
Dense vs. sparse FLOPSVendors often quote tensor throughput with 2:4 structured sparsity, which doubles the headline number. Almost no training or inference uses it, so PetaPrice always shows dense figures.
$ per PFLOP-hourThe price of one GPU-hour divided by the GPU's dense peak PFLOPS at a chosen precision. It shows how much compute a dollar buys, so fast expensive GPUs and slow cheap ones can be compared directly. Lower is better.
TF32NVIDIA's 19-bit format that runs FP32 matrix math on tensor cores. PyTorch can use it for FP32 matmuls on Ampere and newer.
INT8 and TOPS8-bit integers for quantized inference. Throughput is counted in TOPS (trillions of operations per second) rather than FLOPS.
Tensor cores / matrix coresUnits that multiply small matrices in one step, where nearly all of a GPU's AI throughput comes from. NVIDIA calls them tensor cores, AMD matrix cores.
MFU (model FLOPS utilization)The share of peak FLOPS a real workload achieves. Large training runs typically reach 30–60%, so datasheet peaks are for comparing GPUs, not for predicting run time.