From Newsgroup: comp.os.linux.advocacy
Jack Wallen tries a Windows rCLAIrCY machine <
https://www.zdnet.com/article/should-you-buy-a-windows-mini-pc-in-2025-my-verdict-after-a-week-of-testing/>.
First, a test of its claimed rCLAIrCY capabilities, running an identical
local LLM compared with a Linux box:
After the model was pulled, I asked each the same question (both
being used from the command line:
What is Linux?
It's a fairly straightforward and simple question. So, how did
they fare? Here's a breakdown.
Even though both machines used the same LLM, deepseek-r1:8b
running on Linux produced an answer that was considerably longer
and more in-depth. On top of that, the answer produced by the A9
Max AI suffered from terrible grammar and created far more
mistakes than that the Thelio's.
Verdict on AI use:
From an overall experience, my System76 Thelio was much more up to
the task of running AI. To be fair, the A9 Max AI was hampered by
Windows 11, which would make a big difference, given that Windows
11 consumes far more system resources than Linux.
Is it any wonder that AI apps seem to be developed more for Linux than
Windows?
Just for fun, he installed VirtualBox and then created an Ubuntu VM
under the Windows host:
I should add this: The Ubuntu guest OS ran circles around Windows
11, even though it had a third of the RAM and a sixth of the CPU
cores. That alone validated my supposition that Windows was a
hamper.
Anybody want to suggest he should have tried Windows 10 instead?
Probably wouldnrCOt install, though.
--- Synchronet 3.21a-Linux NewsLink 1.2