Hermes Agent hardware question: swap RTX 30 cards for P40s?
A Reddit user asked whether to trade an RTX 3080 10GB and RTX 3070 8GB for two P40 24GB GPUs. Their use case is Hermes Agent, small model loading, and occasional tests with larger models. Commenters said the P40 has more memory but is old, slower for some AI use, and may bring support, power, and setup issues.
Key points
- The user is comparing RTX 3080 10GB plus RTX 3070 8GB against two P40 24GB GPUs.
- Their stated use is Hermes Agent, small models, and occasional larger model tests.
- Several replies warned that P40 is an old GPU with limited future support.
- Commenters said P40 may be slow for inference despite having more VRAM.
- One reply said mixing Pascal and Ampere cards may work on Windows, but not reliably on Linux.
Quick term guide
- Hermes Agent
- It appears to be a tool or community for building and managing AI agents.
- Hermes
- A service for letting an AI agent use web tools and complete tasks.
- models
- Different AI engines that can power answers or code suggestions inside a tool.
- memory
- A ChatGPT feature that lets it use details from past chats in future chats.
- locally
- Running on your own computer or server instead of a remote company server.
- benchmark
- A test used to compare speed, quality, or cost.
- inference
- The step where a trained AI model actually produces answers or results in real use.
- Windows
- Microsoft’s operating system for many personal computers.