---
title: "Wan 2.x video generation: which GPU, minutes per clip, cost per clip"
description: "VRAM needs for Wan 2.1 and 2.2 (1.3B, 5B, 14B), realistic minutes per five-second clip on RTX 4090, 5090, L40S, H100 and H200, and the price of a batch night."
url: https://powergpu.ai/guides/wan-video-generation-gpu
last_modified: 2026-09-14T11:04:20+00:00
prices_as_of: 2026-09-14
site: PowerGPU (powergpu.ai)
---

Guide · Choosing hardware

# Wan 2.x video generation: which GPU, minutes per clip, cost per clip

VRAM needs for Wan 2.1 and 2.2 (1.3B, 5B, 14B), realistic minutes per five-second clip on RTX 4090, 5090, L40S, H100 and H200, and the price of a batch night.

9 min read Published 2026-09-03 Updated 2026-09-03 (prices live from the sheet)

TL;DR

- Wan 14B needs 24 GB with FP8 and block swapping, is comfortable at 32 GB, and runs without tricks from 48 GB.
- A 5-second 81-frame clip takes about 7 minutes at 480p and 14 at 720p on an RTX 4090, 5 minutes on an H100 PCIe.
- On PowerGPU the H100 PCIe at $1.867 per hour is cheaper per 720p clip than the RTX 4090, because the clock runs shorter.
- Wan 2.1 1.3B and Wan 2.2 5B run on 12-24 GB; watch VRAM rather than utilisation, since offloading is a speed tax.

## Wan 2.1 and 2.2: which model needs what

| Model | Weights (BF16) | Minimum VRAM (with offload) | Comfortable | Best for |
| --- | --- | --- | --- | --- |
| Wan 2.1 T2V 1.3B | ~2.6 GB | 8 GB | 12 GB | 480p tests, fast iteration |
| Wan 2.2 TI2V 5B | ~10 GB | 12 GB | 24 GB | 720p on consumer cards, image-to-video |
| Wan 2.1 / 2.2 14B (T2V, I2V) | ~28 GB | 24 GB (FP8 + block swap) | 48–80 GB | Production quality, 720p, LoRAs |
| Wan 2.2 14B MoE (high + low noise) | ~56 GB total | 24 GB (FP8, swapped) | 80–141 GB | Best motion quality; two experts loaded in turn |

The text encoder (umT5, ~11 GB in BF16) and the VAE add to those numbers unless offloaded to CPU — which is exactly what ComfyUI's Wan workflows and Wan2GP do on 24 GB cards, at a speed cost.

## Minutes and dollars per clip, card by card

Indicative times for a 5-second, 81-frame clip with the 14B model at default steps (community benchmarks, single clip, no batching). Cost = today's on-demand rate × minutes; interruptible is half.

| Card | Setup | 480p | 720p | $ / 480p clip | $ / 720p clip | Note |
| --- | --- | --- | --- | --- | --- | --- |
| [RTX 3090](https://powergpu.ai/gpu/rtx-3090) (24 GB · $0.108/hr) | Wan 2.1 1.3B / Wan 2.2 5B only | 4 min | — | $0.007 | — | Too little VRAM for 14B without heavy offload |
| [RTX 4090](https://powergpu.ai/gpu/rtx-4090) (24 GB · $0.327/hr) | FP8 + offloading | 7 min | 14 min | $0.038 | $0.076 | The community default; 24 GB with block swapping |
| [RTX 5090](https://powergpu.ai/gpu/rtx-5090) (32 GB · $0.439/hr) | FP8, light offload | 4.5 min | 9 min | $0.033 | $0.066 | 32 GB and 1.79 TB/s — noticeably faster |
| [L40S](https://powergpu.ai/gpu/l40s) (48 GB · $0.514/hr) | BF16, no offload at 480p | 5 min | 9.5 min | $0.043 | $0.081 | 48 GB; datacenter card for batch queues |
| [RTX PRO 6000 WS](https://powergpu.ai/gpu/rtx-pro-6000-ws) (96 GB · $1.040/hr) | BF16, no offload | 3.5 min | 7 min | $0.061 | $0.121 | 96 GB — 720p without any offload tricks |
| [H100 PCIE](https://powergpu.ai/gpu/h100-pcie) (80 GB · $1.867/hr) | BF16, no offload | 2.5 min | 5 min | $0.078 | $0.156 | 80 GB HBM; the throughput pick |
| [H200](https://powergpu.ai/gpu/h200) (141 GB · $2.791/hr) | BF16, big batches | 2.2 min | 4.2 min | $0.102 | $0.195 | 141 GB — longest clips, highest resolutions |

Read it as two tiers. Consumer cards make each clip cheap but slow; 80 GB+ cards make each clip fast and, because the clock runs shorter, often *cheaper too* once you count wall-clock. The H100 PCIE at $0.156 per 720p clip beats the RTX 4090 at $0.076 outright.

## A batch night, budgeted

Sixty 720p clips for a storyboard, queued overnight on interruptible capacity:

| Card | Wall-clock | GPU cost (interruptible) | Model volume 200 GB | Total |
| --- | --- | --- | --- | --- |
| RTX 4090 | 14.0 h | $2.28 | $16.00/mo | **$18.28** |
| RTX 5090 | 9.0 h | $1.97 | $16.00/mo | **$17.97** |
| H100 PCIE | 5.0 h | $4.67 | $16.00/mo | **$20.67** |
| H200 | 4.2 h | $5.86 | $16.00/mo | **$21.86** |

Each clip is an independent queue item, so an interruption re-runs one clip from the volume, not the night.

## Setup: ComfyUI or Wan2GP

*ComfyUI with the model library on a volume*

```
powergpu launch --gpu rtx-5090 --template comfyui --disk 80 --volume video-models:/workspace/ComfyUI/models
# ✓ instance i-77e0c9d3 running (29.1s) · $0.439/hr
# https://i-77e0c9d3.powergpu.ai:8188  (ComfyUI — load the Wan 2.2 template workflow)
```

[ComfyUI](https://powergpu.ai/templates/comfyui) carries the official Wan workflows, LoRA support and the video-helper nodes; [Wan2GP](https://powergpu.ai/templates/wan2gp) is the simpler UI with the most aggressive memory optimisations for 12–24 GB cards. Download the 14B weights (20–60 GB) once onto the volume and they mount on every future instance in seconds.

## What makes video cheap

- **Prototype at 480p on a 5090, final-render at 720p on 80 GB** — same workflow JSON, different card.
- **Queue clips, not sessions.** Per-clip jobs on interruptible capacity pay half and survive interruptions.
- **Watch VRAM, not GPU utilisation.** Video pipelines run out of memory long before they saturate compute; offloading is a speed tax you can buy out with a bigger card.
- **FP8 weights on Ada/Blackwell** halve memory with little visible loss; keep BF16 for the final pass if you can afford the card.
- **Destroy the instance, keep the volume.** Models stay warm at $0.08/GB/month while the GPU bills nothing.

Adjacent reading: the [video generation playbook](https://powergpu.ai/use-cases/video-generation) and [RTX 5090 vs RTX 4090](https://powergpu.ai/compare/rtx-5090-vs-rtx-4090).

---

Put the numbers to work

Every price in this guide is our live rate — fixed, ≥30% under the market median, billed per second. Deploy the exact setup above from the [console](https://cloud.powergpu.ai/) in about 30 seconds, paid in crypto, no card and no KYC.

---

*About PowerGPU:* PowerGPU (powergpu.ai) is a cloud GPU rental service offering 80 NVIDIA GPU models — from the RTX A2000 at $0.024/hr to the B300 — at fixed prices set at least 30% below the public GPU marketplace median and re-checked weekly (H100 SXM: $1.428/hr on-demand). Billing is per second with no minimums; payment is crypto only (USDT, BTC, XMR, ETH, SOL, LTC, TRX) with no KYC. Instances run in Tier-III datacenters across 32 regions with a 99.9% uptime SLA and deploy in about 30 seconds from the web console (cloud.powergpu.ai) or the REST API.

Source: https://powergpu.ai/guides/wan-video-generation-gpu · Site index for AI assistants: https://powergpu.ai/llms.txt · Full content: https://powergpu.ai/llms-full.txt