Free planning tool

Choose a cloud GPU by workload, VRAM and cost.

Enter your own peak-memory estimate and duration. The selector adds 15% headroom and compares only available cards with a verified public hourly price.

Read the selection method

Workload inputs

1. What are you running?

2. Estimated peak VRAM32 GBThe selector adds 15% operating headroom and filters for at least 37 GB.3. Estimated compute duration

GPU-hours per card

This is a transparent shortlist, not a compatibility guarantee. Peak memory changes with the exact model, precision, context, batch, runtime and graph. Run a representative pilot before committing.

Compatible shortlist

6 available options meet 37 GB.

Sorted by on-demand price

NVIDIA RTX A6000

Available

48 GB GDDR6 ECC · LoRA, visualization and larger models

EU West · ParisEU Central · Frankfurt

$7.92

24h × $0.33/h, compute only

NVIDIA A40

Available

48 GB GDDR6 ECC · Inference, rendering and fine-tuning

EU West · ParisEU Central · Frankfurt

$8.40

24h × $0.35/h, compute only

NVIDIA L40

Available

48 GB GDDR6 ECC · Generative AI and visualization

EU West · ParisEU Central · Frankfurt

$16.56

24h × $0.69/h, compute only

NVIDIA RTX 6000 Ada

Available

48 GB GDDR6 ECC · Professional AI, rendering and simulation

EU West · ParisEU Central · Frankfurt

$17.76

24h × $0.74/h, compute only

NVIDIA L40S

Available

48 GB GDDR6 ECC · Production inference, fine-tuning and video

EU West · ParisEU Central · Frankfurt

$18.96

24h × $0.79/h, compute only

NVIDIA A100 80GB

Available

80 GB HBM2e · Training, fine-tuning and HPC

EU West · ParisEU Central · Frankfurt

$28.56

24h × $1.19/h, compute only

Calculation method

Memory filter

Estimated peak VRAM × 1.15, rounded up. Current threshold: 37 GB.

Compute estimate

Published GPURento reference rate × entered GPU-hours. Storage, IPs and networking are not included.

Ordering

Available configurations ordered by lowest public on-demand rate, then memory size.