# Is anyone running dual GPUs with x570 or x870 Taichi? — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/0542ad4e6d288e3d

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-09-07 17:16 ingested 2026-09-08 02:42

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1w9y7ii/is_anyone_running_dual_gpus_with_x570_or_x870/)

## retained text



```
[X570 Taichi Specs](https://www.asrock.com/mb/AMD/X570%20Taichi/#Specification) AM4

[X870 Taichi Specs](https://www.asrock.com/mb/AMD/X870E%20Taichi/#Specification) AM5

I'm considering these boards as it seems that with the correct CPU, you can run 2 GPUs in x8/x8 PCIe.

I use Linux full-time, I'm comfortable "trouble-shooting" or configuring.

The largest models I'd consider running (for coding) are:
- Qwen3.x-27b (Dense)
- Qwen3.x-3xb-a3b (MoE)

I'm sure I could use smaller models for other tasks and pleasant t/s speeds.

**Is it effective to have either MoBo and a combo like this?**
- 2 x 7900 XT  : (20GB VRAM each, total, 40)
- 2 x 7900 XTX : (24GB VRAM each, total, 48)

By "effective" I mean, that by splitting layers or tensors, whatever/etc., I could have a model and context fully-loaded in VRAM.

---
**Setup misc**

- I'd obviously get a 1200W+ PSU and the biggest case I can find, lots of fans, etc.
- I'm avoiding NVIDIA GPUs because they're significantly more (usually) for the same amount of VRAM.
- If I don't build an AM4 setup, I think I'll settle with an eGPU Dock (laptop) and a 7900 XTX (for starters); I'm aware of the bandwidth limitations of USB4.

---
If you've personally run a setup like this, please let me know!

```



document fetched 2026-09-08 00:40 · canonical https://reddit.com/r/LocalLLaMA/comments/1w9y7ii/is_anyone_running_dual_gpus_with_x570_or_x870
