Hacker News Reader

Incredibly good. A lot of main GPU board partners are making LLM-focused boards. There was a few previews on Gamers Nexus from... Computex I think? I wanna say lots of memory, dual GPU boards that looked very well built for local LLM usage.

With unified memory?

https://www.servethehome.com/maxsun-intel-arc-pro-b60-dual-g...

It's two B60 GPUs on a single x16 PCIe card. Nothing unified.

Working with the vram of both gpus on the same card is aeons faster than taking a system ram roundtrip.

Looks like they are independent x8 GPUs on bifurcated x16 PCIe slot.

I've also heard Intel GPUs make good cheap hardware transcoders for home media servers.

Very curious about this as well.

Though I would say the fact that they were not involved with OpenAI on gpt-oss is not a good sign.

(OAI partnered with Databricks, NVIDIA, Dell, bunch of MSPs. etc.)