
MINISFORUM AI X1 Pro vs MINISFORUM UM890 Pro
Two mini PCs side by side: what each one suits, where they differ, and the price of each on amazon.com.
Updated


MINISFORUM UM890 Pro
The short answer
These two compact MINISFORUM boxes share two 2.5GbE ports and integrated Radeon graphics, then part ways on expansion and memory as sold. The AI X1 Pro brings a 12-core Ryzen AI processor, a 50 TOPS NPU, Wi-Fi 7 and up to four displays, and its 96 GB model is the one for local AI and workstation duty. The UM890 Pro suits you if you want an OCuLink port for an external graphics card later, four USB-A ports, or a barebones kit to fill yourself.
Specifications side by side
CPU, GPU, installed memory and storage describe the configurations shown above. Memory limits, slots and ports describe the chassis; other configurations can differ.
| Specification | MINISFORUM AI X1 Pro | MINISFORUM UM890 Pro |
|---|---|---|
| Processor | AMD Ryzen AI 9 HX 470 | AMD Ryzen 9 8945HS |
| Graphics | AMD Radeon 890M | AMD Radeon 780M |
| Installed memory | 32 GB | Not included in this barebones kit |
| Installed storage | 1 TB | Not included in this barebones kit |
| Maximum memory | 96 GB | 96 GB |
| Memory slots | 2 | 2 |
| M.2 slots | 2 | 2 |
| LAN speed | 2.5GbE | 2.5GbE |
| USB4 ports | 2 | 2 |
| Thunderbolt ports | Not confirmed | 0 |
Workload by workload
Everyday work and coding come out level. Both are well suited, the AI X1 Pro in either form and the UM890 Pro in its two built configurations. The UM890 Pro barebones kit is not assessed until built, since it ships without memory or storage.
Gaming is workable on both. The AI X1 Pro's Radeon 890M is fine for esports titles at 1080p but not for current AAA games at 1440p. The UM890 Pro's Radeon 780M suits esports and older titles at 1080p as sold, and its PCIe 4.0 x4 OCuLink port can drive an external GPU enclosure, which the AI X1 Pro lacks.
Homelab use depends on memory for the AI X1 Pro: workable with 32 GB, well suited with 96 GB. The UM890 Pro is workable, with no 10GbE.
For local AI, capacity decides which model sizes fit. The 32 GB versions of both are workable. For local AI, check the exact model release, quantization, context length, runtime overhead and memory available after the operating system. System RAM and GPU VRAM are different limits; shared-memory allocation depends on firmware, drivers and the backend. CPU inference or partial offloading can change what loads and its speed. Installed capacity alone does not establish compatibility or useful generation speed. Check the current Ollama or LM Studio hardware documentation for the exact GPU and operating system.
Workstation use is where they part company. The 96 GB AI X1 Pro is well suited and the 32 GB one workable, while the UM890 Pro is a poor fit, with no discrete graphics, no x8 or x16 card slot and fewer than four memory slots.
Expansion and upgrades
Memory headroom matches on paper: each has two SO-DIMM slots, memory that is not soldered, and a 96 GB maximum. The 32 GB versions of both can grow to 96 GB, the 64 GB UM890 Pro has a smaller step left, and the 96 GB AI X1 Pro is already full.
For storage, both machines have two M.2 slots that take 2280 drives. The AI X1 Pro ships with 1000 GB or 2000 GB, the UM890 Pro with 1000 GB. MINISFORUM does not state either machine's M.2 slot PCIe generation or SATA support, and a drive runs at its slot's speed, so check before paying for a fast one. Neither has a PCIe card slot.
Ports and connectivity
Wired networking matches at two 2.5GbE ports each. Wireless does not: the AI X1 Pro includes a Wi-Fi 7 module with Bluetooth 5.4, while MINISFORUM does not state the UM890 Pro's Wi-Fi standard, Bluetooth version or whether a wireless module is included. Both have two USB4 ports and two USB-C ports, and the UM890 Pro's USB4 ports run at 40Gbps. The UM890 Pro adds four USB-A ports and has no Thunderbolt; MINISFORUM does not state either count for the AI X1 Pro.
The AI X1 Pro drives up to four displays from one HDMI output, one DisplayPort output and DisplayPort over USB-C. The UM890 Pro has one HDMI and one DisplayPort output, and MINISFORUM does not state its display maximum, so plan around those outputs. Neither HDMI version is stated, so 4K above 60 Hz is not assured.
Which configurations to compare
The even match is the 32 GB AI X1 Pro with the Ryzen AI 9 HX 470 and 1000 GB of PCIe 4.0 NVMe storage against the 32 GB UM890 Pro with the Ryzen 9 8945HS and 1000 GB. Memory and storage line up, so the gap is 12 cores and 24 threads against 8 cores and 16 threads, plus the graphics and ports.
Higher up, the 64 GB UM890 Pro meets the 96 GB AI X1 Pro with the Ryzen AI 9 HX 370 and 2000 GB: both are well suited to local AI, but only the AI X1 Pro to workstation use. The UM890 Pro barebones kit has no counterpart, since the AI X1 Pro comes built.
If neither fits
If you need more than 96 GB of memory, neither machine gets you there, because both stop at that maximum. The MINISFORUM MS-02 Ultra is the step to consider: it uses an Intel Core Ultra 9 285HX instead of the Ryzen AI 9 HX 470 and takes up to 256 GB of memory rather than 96 GB. That higher ceiling matters for local language models, since memory capacity governs which model sizes fit.
Workload fit
The best configuration of each machine, as our fit rules judge it.
| Workload | MINISFORUM AI X1 Pro | MINISFORUM UM890 Pro |
|---|---|---|
| Everyday | Good | Good |
| Coding | Good | Good |
| Gaming | Workable | Workable |
| Home lab | Good | Workable |
| Local AI | Good | Good |
| Workstation | Good | Not suited |
Common questions
Which one can take an external graphics card?
The UM890 Pro, through its OCuLink port, which carries PCIe 4.0 x4 to an external GPU enclosure. The AI X1 Pro has no OCuLink and neither machine has a PCIe card slot, so its Radeon 890M is what you game on.
Can I upgrade the memory later?
Yes on both. Each has two SO-DIMM slots with memory that is not soldered, up to 96 GB. The 32 GB versions have the most room to grow, while the 96 GB AI X1 Pro already sits at that maximum.
Which is the better choice for running local language models?
For local AI, check the exact model release, quantization, context length, runtime overhead and memory available after the operating system. System RAM and GPU VRAM are different limits; shared-memory allocation depends on firmware, drivers and the backend. CPU inference or partial offloading can change what loads and its speed. Installed capacity alone does not establish compatibility or useful generation speed. Check the current Ollama or LM Studio hardware documentation for the exact GPU and operating system. MINISFORUM does not state memory bandwidth for either, and bandwidth governs generation speed.
Is the UM890 Pro barebones kit ready to use out of the box?
No. It is sold without memory or storage, and MINISFORUM does not state an operating system for it, so you fit SO-DIMM memory and an M.2 2280 drive before it runs. What it suits is not assessed until built.
The two machines
Fields we could not confirm from the manufacturer are marked not confirmed rather than guessed. Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon.com at the time of purchase will apply to the purchase of this product.