Gorgon Halo is just a renamed Strix Halo with software-patched support for up to 192 GB RAM. The price doesn't matter in the sense that it's only 96 GB RAM and, I think on Windows 32 GB are reserved for the OS and can not be used/allocated, so it's really 64 GB RAM. On 128 GB RAM configs, only 96 GB can be allocated. 16 GB RAM for the OS would have been enough. On Linux, a higher percentage of the total RAM/memory can be allocated. But maybe the situation on Windows is better by now.
The rest of any AI LLM model that doesn't fully fit into RAM and/or VRAM, can be offloaded to a SSD (speed will be slower, but at least it will run) (see also BigMoeOnEdge): reddit.com/r/LocalLLaMA/comments/1ow0jj0/running_a_1_trillion_parameter_model_on_a_pc_with/ reddit.com/r/LocalLLaMA/comments/1vdbix4/deepseekv4flash_284b_on_53gb_of_memory/
An unannounced HP workstation has surfaced on Geekbench, pairing AMD's newest flagship mobile silicon with a chassis size HP hasn't offered in this series before.
Deutsch
English
Español
Français
Italiano
Nederlands
Polski
Português
Русский
Türkçe
Svenska
Chinese
Magyar