Two RTX 5080s failed to beat one RTX 5090 in ComfyUI single-render tests
Geekdomo · reddit · 2026-07-27
A detailed ComfyUI benchmark concludes that two RTX 5080s do not meaningfully speed up a single render compared with one RTX 5090.
The author tested multi-GPU setups for image generation and LTX 2.3 video generation on a high-end workstation, including multi-GPU ComfyUI configurations, Raylight, workflow splitting, offloading, device-specific execution, peer-to-peer transfer, and separate instances. The key findings were:
- VRAM does not automatically become one usable pool for a standard workflow.
- Splitting one render across two cards added overhead and made jobs slower.
- The only useful dual-GPU setup was running independent jobs in parallel.
- In the author’s tests, the RTX 5090 was about 3.23× faster for image batches, 1.82×–1.89× faster on production-quality video models, and 2.14× faster with NVIDIA NVFP4, though that checkpoint still produced poor-quality output.
The author concludes that if your goal is faster single-job generation, a second 5080 is not a substitute for a 5090.
More from Infra
- Nvidia signs $1.5B multi-year Amkor deal to expand US chip packaging capacity — Beth_Kindig · 2026-07-27
- PyTorch DDP misses a tiny-parameter NVIDIA GPU optimization out of the box — gordic_aleksa · 2026-07-27
- A fully local AI girlfriend runs on 15GB VRAM with Whisper, llama.cpp, and Qwen3-TTS — max_paperclips · 2026-07-27
- OpenAI reportedly plans to spend over $30B on a 3.2GW Georgia data center — Beth_Kindig · 2026-07-27
- Open-source profiler tracks every STT, LLM, and TTS call in self-hosted voice agents — mahimairaja · 2026-07-27
- SemiAnalysis says better memory and storage can beat a faster GPU in modern inference — rwang07 · 2026-07-27