AnswersPublished27 sources
Lenovo ThinkCentre X Ultra vs. Acemagic F9A Pro: Local AI in a Mini Workstation
The key advantage of Lenovo’s 1.6L ThinkCentre X Ultra and Acemagic’s roughly 2L F9A Pro is large unified memory, not the 131 TOPS headline: Lenovo offers up to 128GB and Acemagic up to 192GB, giving their integrated... Choose Lenovo for a business oriented, cluster ready deployment—up to four units and 512GB aggreg...
Published byEdited with GPT-5.6 TerraImages generated with GPT Image 2
Research answer
Key takeaways
- The key advantage of Lenovo’s 1.6L ThinkCentre X Ultra and Acemagic’s roughly 2L F9A Pro is large unified memory, not the 131 TOPS headline: Lenovo offers up to 128GB and Acemagic up to 192GB, giving their integrated...
- Choose Lenovo for a business oriented, cluster ready deployment—up to four units and 512GB aggregate memory—with November 2026 availability from $3,699.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: How do Lenovo’s ThinkCentre X Ultra and Acemagic’s F9A Mini Workstation, unveiled at IFA 2026, use AMD’s Ryzen AI Max+ Pro 495 processor to. Article summary: Lenovo and Acemagic are putting workstation-class local-AI capacity into 1.6–2 L desktops by pairing the Ryzen AI Max+ PRO 495’s 16-core Zen 5 CPU, Radeon 8065S integrated GPU, and 55-TOPS NPU with unusually large shared. Topic tags: general, general web. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fake numbers, click
IFA 2026 made one trend unmistakable: compact desktops are becoming credible local-AI machines when they have enough shared memory. Lenovo’s ThinkCentre X Ultra and Acemagic’s F9A Pro both use AMD’s Ryzen AI Max+ PRO 495 platform, combining a 16-core CPU, Radeon 8065S integrated graphics and a 55-TOPS NPU. AMD rates the platform at up to 131 total AI TOPS. 3
5
The meaningful distinction is memory. Rather than pairing an integrated GPU with a small fixed graphics-memory allocation, these systems use a large unified LPDDR5X pool shared across CPU, GPU and NPU. That makes them potentially useful for running larger local language models, longer-context inference and development workloads in a desk-friendly enclosure.
At a glance
| Feature | Lenovo ThinkCentre X Ultra | Acemagic F9A Pro 495 |
|---|---|---|
| Form factor | 1.6L desktop |
Roughly 2L; 158.5 × 158.5 × 81.5 mm |
| Processor | Up to Ryzen AI Max+ PRO 495 |
Ryzen AI Max+ PRO 495 |
| Platform AI rating | Up to 131 TOPS; 55-TOPS NPU |
Up to 131 TOPS; 55-TOPS NPU |
| Unified memory | Up to 128GB |
Up to 192GB LPDDR5X-8533 |
| GPU memory allocation | Up to 96GB of graphics memory |
Up to 160GB can be allocated to graphics, according to reports on Acemagic’s announcement |
| Scale-out | Up to four systems in a cluster |
No comparable clustering announcement identified |
| Availability | Expected November 2026 |
Not announced |
| Starting price | $3,699 expected |
Not announced |
Why unified memory changes local LLM work
A local model needs room not only for its weights but also for its context cache and runtime overhead. In these designs, the Radeon integrated GPU can draw from the same large LPDDR5X pool as the CPU instead of being constrained by the fixed VRAM capacity of a conventional discrete GPU.
That is why the F9A Pro’s 192GB ceiling is its standout specification. Acemagic says the system has demonstrated local operation of DeepSeek V4 Flash with 284 billion parameters, while other reports say up to 160GB can be allocated as graphics memory. Those are capacity-oriented claims, not guarantees of a particular response speed, framework compatibility or fine-tuning performance. 3
5
In practice, model usability still depends on quantization, memory bandwidth, model architecture, context length, batch size, cooling and software support. A machine may be able to load a large model yet remain less practical for interactive work than a system with faster accelerator hardware.
Lenovo’s 128GB maximum is lower, but still unusually large for a 1.6L desktop. Its reported 96GB graphics-memory allocation leaves substantial room for the operating system and applications while targeting local AI and workstation workloads. 18
32
Lenovo’s advantage: four-node scale-out
The ThinkCentre X Ultra is the more enterprise-oriented design. Lenovo says up to four systems can connect in a cluster, expanding compute and memory resources for larger models, longer context windows, multi-agent workflows and concurrent requests. A fully populated group reaches up to 512GB of aggregate unified memory. 32
36
There is an important caveat: four machines do not automatically behave like one GPU with a single 512GB memory pool. Workloads must be designed for distributed execution, and performance will depend on the software stack and interconnect. The practical appeal is capacity and parallelism in a small footprint, rather than a simple way to turn four 96GB graphics allocations into one seamless accelerator.
Lenovo expects the X Ultra to ship in November 2026, starting at $3,699. The company’s announcement confirms Windows 11 and Linux compatibility, while reporting also points to up to two M.2-2280 SSDs, with configurations reaching two 4TB drives. 37
38
44
Acemagic’s advantage: maximum memory per box
Acemagic’s F9A Pro targets users who want the largest announced unified-memory configuration in one compact machine. The Ryzen AI Max+ PRO 495 configuration supports up to 192GB of LPDDR5X-8533 memory in a roughly two-liter chassis. 2
5
The announced connectivity is unusually broad for a mini workstation: two USB4 ports, dual 2.5GbE, Wi-Fi 7, Bluetooth 5.4, HDMI 2.1, DisplayPort 2.1 and OCuLink expansion. OCuLink provides a PCIe-based route for external expansion, including an external GPU; it does not, however, convert the system’s unified LPDDR5X into discrete GPU VRAM. 4
5
Storage details need care. Acemagic’s product pages for the earlier Ryzen AI Max+ 395 F9A list dual PCIe 4.0 x4 M.2 NVMe slots, while reporting on the PRO 495 version has varied between one and two M.2 slots. Buyers should verify the final regional specification before assuming a particular storage layout. 9
12
15
Acemagic had not announced the F9A Pro 495’s final SKU lineup, regions, price or retail date at the time of its IFA showcase. That makes the 192GB configuration compelling on paper, but not yet a like-for-like shipping alternative to Lenovo’s announced November product. 2
Where these systems fit against RTX Spark and other AMD mini PCs
These are not conventional high-end GPU workstations shrunk into smaller cases. They are large-unified-memory x86 desktops designed to keep more AI work on-device. Their appeal is model capacity, compactness and conventional desktop flexibility—not necessarily raw performance leadership over systems using Nvidia’s CUDA ecosystem.
Nvidia RTX Spark systems are also due in mini PCs and laptops from October 2026, with Nvidia describing versions that use a Grace CPU and Blackwell GPU. That creates a direct category battle for desk-side AI development, particularly among teams that prioritize Nvidia’s GPU software stack. 24
25
AMD’s side of the market is expanding as well. GMKtec announced an EVO-X5 Pro based on the same Ryzen AI Max+ PRO 495 processor, although it had not provided a price at announcement. 17 The result is a growing choice of mini systems that prioritize memory-rich local inference over the traditional pattern of pairing a small desktop with limited graphics memory.
Which one makes more sense?
ThinkCentre X Ultra is the clearer choice for organizations that value a defined rollout schedule, Windows and Linux support, compact deployment and the option to build a four-node cluster. Its 128GB maximum remains substantial, and its platform is designed around multi-system local-AI use. 32
44
Acemagic F9A Pro is the more interesting option for developers whose first requirement is the greatest announced per-machine memory capacity. Its 192GB unified-memory target and OCuLink expansion are distinctive, but purchasing decisions should wait for final storage, operating-system, regional availability and pricing details. 2
4
5
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "Lenovo ThinkCentre X Ultra vs. Acemagic F9A Pro: Local AI in a Mini Workstation"?
The key advantage of Lenovo’s 1.6L ThinkCentre X Ultra and Acemagic’s roughly 2L F9A Pro is large unified memory, not the 131 TOPS headline: Lenovo offers up to 128GB and Acemagic up to 192GB, giving their integrated...
What are the key points to validate first?
The key advantage of Lenovo’s 1.6L ThinkCentre X Ultra and Acemagic’s roughly 2L F9A Pro is large unified memory, not the 131 TOPS headline: Lenovo offers up to 128GB and Acemagic up to 192GB, giving their integrated... Choose Lenovo for a business oriented, cluster ready deployment—up to four units and 512GB aggregate memory—with November 2026 availability from $3,699.
Sources
boerse-express.comACEMAGIC F9A Pro: Mini-Workstation mit 131 TOPS für KI lokal
finance.yahoo.comACEMAGIC First to Introduce Ryzen AI Max+ PRO 495 AI Mini Workstation at IFA 2026
news.lenovo.comLenovo Advances Hybrid AI for Business with New Think-Family Devices, Displays, and Security Solutions - Lenovo StoryHub
windowsforum.comACEMAGIC F9A PRO 495: 192GB Mini PC, What’s Known