The AMD Strix Halo mini PC lineup has seen another new addition, the Sixunited AXB35-02. It will be available with up to Ryzen AI Max+ 395, which brings the Radeon 8060S iGPU. The system can also be configured at up to 120 W TDP.https://www.notebookcheck.net/Sixunited-AXB35-02-debuts-as-new-Ryzen-AI-Max-395-mini-PC-with-up-to-120-W-TDP.984007.0.html
I hope the reviewers to stop calling this a mini-pc anymore! Instead, they should call it a "Compact PC" for the sake of distinction from the real mini-pc form factor.
Mini pcs should be easy to carry in back-bags not like this humongous freak!, I appreciate its 8060s iGPU but it's clear how much far Apple has advanced in terms of iGPU performance and efficiency, and I hoped this one would not need a bulky heatsink to cool! It hurts me to say the truth because I'm not a fan of Apple and their ways to cage its users. so I'm really frustrated!
And how many tops of NPU, XDNA, the last 8000 series I owned, had 16 tops, my 3 year old iPad mini 6, had 16 tops, my iPad mini 7, has 35 tops, my Mac mini M4, has 38 tops, for $630 US. Oh, it uses graphics AI, unless it's for training, it's not power efficient, NPU, TPU, machine learning, is efficient for inference, which is what 99% of us use AI for, we're not developers. Yes, Apple generally has less RAM, but it's unified RAM, on the very chip die, soldered in, right next to the power efficient ARM APU. In a year, the Mac mini M4, will run Apple intelligence, Gemini and Copilot, those having decrypted deepseek, for good power efficiency.
ARM and Machine Learning architectures, memory efficiency from deepseek, at 3nm, unified memory speeds, without updating problems, that's why Apple, is worth $3 trillion, patience and focus.
Quote from: StuPCIe5AI3nmGDDR6 on March 28, 2025, 18:58:31And how many tops of NPU, XDNA, the last 8000 series I owned, had 16 tops, my 3 year old iPad mini 6, had 16 tops, my iPad mini 7, has 35 tops, my Mac mini M4, has 38 tops, for $630 US. Oh, it uses graphics AI, unless it's for training, it's not power efficient, NPU, TPU, machine learning, is efficient for inference, which is what 99% of us use AI for, we're not developers. Yes, Apple generally has less RAM, but it's unified RAM, on the very chip die, soldered in, right next to the power efficient ARM APU. In a year, the Mac mini M4, will run Apple intelligence, Gemini and Copilot, those having decrypted deepseek, for good power efficiency.
ARM and Machine Learning architectures, memory efficiency from deepseek, at 3nm, unified memory speeds, without updating problems, that's why Apple, is worth $3 trillion, patience and focus.
395 has 50 tops on the APU and total of 126 tops. It also has unified memory. The biggest bottleneck of the 395 is limited to 256bit bandwidth or it could probably do more.