MS-S1 MAX

Hi all, about 2 months ago I picked up the MS-S1 MAX and have been having an amazing time adapting my skillset from a web developer to tinkering AI lunatic. I've been custom building an inference platform with Proxmox running an LXC w/ a custom setup that manages "slots" each running a type of strix

The “Strix Halo is slow” argument is only really true for dense models, not for your setup. Since your workload is RAG plus summarization, MoE models are the better fit, and 128GB of unified memory is a very strong combination for that. GPT-OSS 120B can run at around 30 tokens per second, Qwen3-30B-

Minisforum MS-S1 Max (128GB / 2TB, has a PCIe x16 slot + USB4v2 + dual 10GbE) — seems like the most expandable I just received one of these last week and the expandability is the key reason I chose it over any of the other Strix Halo options. You do want to note that the PCIe is a x16 slot but has x

If 1080P gaming is your goal, it is an exceptional piece of hardware, but that is where the exceptional part ends when it comes to gaming. An RTX 5080 laptop has the absolute raw TDP and graphics hardware to obliterate strix halo. We aren't just talking double digit gaines here. we are talking about

It may be ignorance on my part, but with my Minisforum MS-S1 max 128GB, which I believe i the same thing as Strixhalo, I found that I cannot really use CUDA productively with my RTX5070ti 16GB on eGPU and USB4. On windows11, i found if I have a model that is larger than 16GB, it spillover to ram, at

Isn’t minisforum ms-s1 max based on the AMD Ryzen AI Max+ 395 combined with a powerful iGPU (AMD Radeon 8060S) - 40 cores ? I guess if you research for LLM inference benchmarks using vulkan on AMD Radeon 8060S - you should be seeing good numbers in terms of output tokens / second - which should be p

Take note that I ask an LLM to compose this, but all is on point. 💥 MS-S1 Max/Strix Halo eGPU Users: I Found the Fix! 💥 For everyone struggling with getting an eGPU to reliably connect (especially via hotplug) to the Minisforum MS-S1 Max' rear ports on Linux, I've confirmed the absolute core workaro

MS-S1 MAX with Oculink 4x pci card + DEG1 Oculink + RTX5080 tested.I managed to get it working (Linux and Windows see both the integrated gpu AND the RTX5080). Really hard to make it work. It requires plugging the Oculink cable at some brief moment after the S1 MAX has booted, but not too long. It i

Minisforum MS-S1 MAX But why TB5? A MiniPC could use a permanent Oculink connection. This is much faster.

What a ridiculous post. It's not "a bit", and nothing justifies buying it at $3999 vs a Minisforum S1 Max at $2399. Training using the full Nvidia professional stack absolutely justifies it for a certain demographic. You're not it. Sounds gimmicky, for local, NVMe drives are fine and for connectivit