ZDNET’s key takeaways
- The Geekom A9 Max AI is accessible for $999 on Amazon.
- This tiny PC has tons of energy to spare and might even work effectively with the inefficiencies of Home windows.
- Nevertheless, its followers could be fairly loud.
Comply with ZDNET: Add us as a preferred source on Google.
After I assessment small form factor PCs, I are likely to maintain my expectations in test. On condition that I exploit a monstrous System76 Thelio desktop for many of my work, it is usually a problem for such a tiny, shiny field to match. Actually, I’ve by no means met a tiny PC that might outperform my Thelio.
Additionally: How to upgrade your ‘incompatible’ Windows 10 PC to Windows 11 – for free today
The GEEKOM A9 Max AI Mini actually gave it a run for its cash. And whereas being hampered by an inferior OS. Observe to all firms sending me PCs to assessment: Please ship me machines with Linux preinstalled. Not solely do I discover it far simpler to assessment computer systems with Linux, nevertheless it additionally reveals shoppers that you simply do provide open-source choices.
With that request out of the way in which, let’s discuss in regards to the A9 Max AI.
My expertise
Naturally, as a result of this machine sported the Home windows working system, it took a very long time earlier than I might even use it. There have been updates aplenty, logging in with my Microsoft account, blah, blah. the drill.
As you most likely seen, the title of the PC contains AI, and I am certain you may guess what meaning. That is proper, this little buddy was constructed for “unmatched AI efficiency.” What does that even imply?
Additionally: 10 open-source apps I recommend every Windows user try – for free
Think about this: most individuals who use AI aren’t utilizing it domestically; they’re going to be utilizing ChatGPT or one of many many cloud-based companies.
Based on the outline, the AMD Ryzen AI 9 HX 370 processor “delivers an industry-leading 80 TOPS of whole AI acceleration, that includes a next-gen XDNA 2 AI engine that pumps out 50 TOPS of devoted NPU efficiency, enabling seamless Copilot+ experiences, native massive mannequin deployment, and superior content material era.”
Yeah, after all. And why would not it, proper?
The rear of the A9 Max AI PC.
Jack Wallen/ZDNET
The factor that jumped out at me is the “native massive mannequin deployment,” so I did precisely what you may anticipate: I put in Ollama and downloaded one of many larger LLMs to see how effectively it could carry out. The mannequin in query was the 9GB deepseek-r1:8b, and I pulled it for each the A9 Max AI and my System76 Thelio, working Pop!_OS.
After the mannequin was pulled, I requested every the identical query (each getting used from the command line:
What’s Linux?
It is a pretty easy and easy query. So, how did they fare? This is a breakdown.
Despite the fact that each machines used the identical LLM, deepseek-r1:8b working on Linux produced a solution that was significantly longer and extra in-depth. On high of that, the reply produced by the A9 Max AI suffered from horrible grammar and created much more errors than that the Thelio’s.
So far as the occasions, they appeared like this:
- Thelio: 2:42.38
- A9 Max AI: 01:07.69
Sadly, the occasions do not actually inform an correct story, given how rather more in-depth the Thelio reply was.
Additionally: How much RAM does your Linux PC really need in 2025? I did the math so you don’t have to
From an total expertise, my System76 Thelio was rather more as much as the duty of working AI. To be truthful, the A9 Max AI was hampered by Home windows 11, which might make an enormous distinction, provided that Home windows 11 consumes much more system assets than Linux.
I then tried to essentially push the A9 Max AI and use the gpt-oss:120b mannequin, which is significantly bigger and extra succesful than the DeepSeek mannequin I initially used. I believed this is able to process the A9 Max AI past what anybody would use it for.
The outcomes? Effectively, it was very gradual, however that ought to be anticipated by such an enormous mannequin on a small machine. The excellent news is that the outcomes have been significantly extra correct. I might hear the followers working at excessive speeds to maintain the CPU cool, and the top end result took practically 5 minutes to finish.
However provided that this machine does not have a strong NVIDIA GPU, it was nonetheless fairly spectacular that it was in a position to run the mannequin to be with.
Let’s get past the AI problem and take a look at this from a extra conventional person expertise.
Putting in and working purposes
You will discover loads of USB ports on this child.
Jack Wallen/ZDNET
I put in a number of purposes on the A9 Max AI and located all of them put in in a short time and ran fairly effectively. I put in LibreOffice, GIMP, Msty, and VirtualBox to see how effectively the A9 Max stood as much as what I am accustomed to.
Ultimately, every thing put in rapidly and ran easily. With the purposes prepared, I made a decision to create a digital occasion of Ubuntu utilizing VirtualBox to see how effectively it carried out. I gave the VM 10GB of RAM and 4 CPU cores (out of 32GB of RAM and 24 CPU cores).
Additionally: ChatGPT just got several work-friendly updates – what it can do now
After the set up completed (took about 2 minutes), I began the VM to see how effectively it could run. Through the set up of the Ubuntu visitor, the followers on the A9 Max AI have been at their loudest, so underneath load, this isn’t a quiet machine. I used to be in a position to hear the followers working from inside one other room of the home. If quiet is a necessity (and you actually push your machine), this won’t be the pc for you. If sheer energy and a small measurement are what you are in search of, this mini PC is a good choice.
After the VM was created, I fired it up. It ought to come as no shock (particularly given the assets I allotted for the machine) that the Ubuntu visitor ran very effectively. Even with the VM up and working, put in purposes carried out fairly effectively.
I ought to add this: The Ubuntu visitor OS ran circles round Home windows 11, though it had a 3rd of the RAM and a sixth of the CPU cores. That alone validated my supposition that Home windows was a hamper.
ZDNET’s shopping for recommendation
At first blush, a mini PC promoting for $999 may look like a foul cut price, however when you think about the efficiency you get with the A9 Max AI, the associated fee turns into much less of an element. This little machine dealt with every thing I might throw at it, and the one cause it blinked was due to Home windows 11. If I had taken the time to put in Linux on this machine, I can think about it could carry out insanely effectively. Sadly, Home windows is a useful resource hog and does not obtain practically the effectivity of Linux.
That being stated, the Geekom A9 Max AI mini PC significantly impressed me. If I have been on the hunt for a strong PC that took up subsequent to no house, this may effectively be my first selection.
Additionally: I finally found a Windows mini PC powerful enough to replace my desktop (and it’s on sale)
Geekom A9 Max AI mini PC tech specs
- Processor: AMD Ryzen AI 9 HX 370 (configurable TDP as much as 54W)
- Graphics: AMD Radeon 890M (RDNA 3.5, 16 compute items)
- AI Efficiency: As much as 80 TOPS of whole AI efficiency, with a devoted AI engine
- Reminiscence: As much as 128GB of dual-channel DDR5-5600MHz RAM
- Storage: 1 x M.2 2280 PCIe 4.0 NVMe SSD slot, 1 x M.2 2230 PCIe 4.0 NVMe SSD slot
- Connectivity: Wi-Fi 7 and Bluetooth 5.4 2 x 2.5 GbE RJ45 ports A number of USB ports, together with USB 4.0 Sort-C ports with DP-Alt-Mode and Energy Supply
- 2 x HDMI 2.1 ports
- Full-size SD card reader
- Working System: Home windows 11 24H2
- Dimensions: 135 × 132 × 46.9 mm
