I'm not pulling off any special tricks here or sabotaging llama.cpp by having some load in the background, Both engines are running with the same parameters same conditions, etc..
Unfortunately (Or rather fortunately) this isn't the end, There's still plenty of headroom left.
@abdimoalim_ That's a nice what if you got there.
Realistically speaking, Modern computers are faster than what we give them credit for.
It's just that software is extremely unoptimized.
So even if such CPU existed, Someone would figure out a way to make programs *not* finish in one cycle..
@M0_R4_H4@bee_fumo Yes but we all know Nvidia to be a little funky, And speaking from personal experience, The linux on ARM situation is not that great vendor wise, Pretty much everything is a proprietary legacy mess that nobody maintains.
@bee_fumo Also no..
You need the drivers themselves to support ACPI, What good is a GPU driver if it was built with the assumption of running via device trees only?