AnswersPublished18 sources
IFA 2026 Local AI PCs: GEEKOM, Acemagic and Minisforum’s AMD Gorgon Point and Gorgon Halo Systems
IFA 2026 split AMD’s compact local AI story in two: Gorgon Point machines such as GEEKOM’s GeekBook X16 Pro and Acemagic’s F2A handle mainstream on device AI, while Minisforum’s Gorgon Halo systems scale to 192GB of u... GEEKOM’s AMD GeekBook X16 Pro pairs a Ryzen AI 9 HX 470 with Radeon 890M graphics; Minisforum’s...
Published byEdited with GPT-5.6 TerraImages generated with GPT Image 2
Research answer
Key takeaways
- IFA 2026 split AMD’s compact local AI story in two: Gorgon Point machines such as GEEKOM’s GeekBook X16 Pro and Acemagic’s F2A handle mainstream on device AI, while Minisforum’s Gorgon Halo systems scale to 192GB of u...
- GEEKOM’s AMD GeekBook X16 Pro pairs a Ryzen AI 9 HX 470 with Radeon 890M graphics; Minisforum’s MS S1 MAX P495 workstation and N5 MAX P495 AI NAS use Ryzen AI Max+ PRO 495, with up to 160GB assignable as graphics memo...
- Acemagic confirmed both Intel Core Ultra X7 358H and Ryzen AI 9 HX 470 versions of its F2A mini PC, but final source backed AMD memory and storage configurations remain unclear.[3]
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What AMD Gorgon Point and Gorgon Halo devices did GEEKOM, Acemagic, and Minisforum unveil or demonstrate at IFA 2026—including the GeekBook. Article summary: IFA 2026 showed a two-tier AMD local-AI strategy: Gorgon Point systems such as GEEKOM’s laptop and Acemagic’s F2A target portable/on-device AI, while Gorgon Halo mini workstations use unusually large unified memory to ru. 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 did not produce one single type of “AI PC.” Instead, GEEKOM, Acemagic and Minisforum used AMD platforms for two distinct jobs: portable or conventional mini PCs built around the Ryzen AI 9 HX 470, and compact local-AI systems whose defining feature is far more unified memory.
The latter category is led by Minisforum’s Ryzen AI Max+ PRO 495 machines. With up to 192GB of unified LPDDR5x memory and up to 160GB allocated to graphics workloads, they are designed for people who want to keep larger models and data on-premises rather than rely entirely on cloud inference.1
2
GEEKOM GeekBook X16 Pro: a thin-and-light Gorgon Point laptop
GEEKOM showed an AMD version of its 16-inch GeekBook X16 Pro at IFA, built around AMD’s Ryzen AI 9 HX 470 and Radeon 890M integrated graphics.4
5 GEEKOM describes the HX 470 platform as combining Zen 5 CPU cores, an XDNA 2 NPU and Radeon 890M graphics, with up to 86 total platform TOPS for local AI, creative applications and 1080p gaming.
6
The Ryzen AI 9 HX 470 is reported as a 12-core, 24-thread processor with boost clocks up to 5.2GHz.5 In practical terms, this positions the GeekBook as a conventional premium notebook that can accelerate AI-enabled applications locally alongside creative work, multitasking and gaming—not as a dedicated large-language-model server.
Display, chassis and battery
GEEKOM’s X16 product material describes a 16-inch, 16:10, 2.5K, 120Hz display with 100% sRGB coverage. It also lists a CNC magnesium chassis at about 2.8 lb and 0.27 inches thick.7 GEEKOM’s IFA material for the AMD version separately identifies a 16-inch magnesium-alloy design and a 77Wh battery.
4
There is an important qualification: the detailed X16 product page is for the established Intel configuration. The available IFA material confirms the AMD processor, Radeon graphics, magnesium body and 77Wh battery, but does not publish a complete, final configuration sheet for the AMD X16 Pro’s RAM capacity, SSD capacity or storage-slot layout. Those values should not be inferred from the Intel X16 or from GEEKOM’s A9 Max mini PC.
Acemagic F2A: one chassis, Intel or AMD platform
Acemagic’s F2A is offered in two significantly different processor directions: Intel’s Core Ultra X7 358H, based on Panther Lake, or AMD’s Ryzen AI 9 HX 470.3
The Intel model is specified with a 16-core, 16-thread Core Ultra X7 358H, Intel Arc B390 graphics based on Xe3, an NPU rated up to 50 TOPS, and a claimed total platform AI rating of up to 180 TOPS.3 Intel’s Panther Lake platform combines its CPU architecture with Xe3/Celestial integrated graphics.
8
The AMD option instead uses the Ryzen AI 9 HX 470 and Radeon 890M, bringing AMD’s XDNA 2 NPU and Radeon graphics design to the same compact-PC class.3
6
What is and is not confirmed
The public product material clearly confirms the two F2A processor options. It does not provide a sufficiently complete, consistent specification table for a definitive comparison of the final AMD and Intel versions’ memory ceilings, SSD bays or regional storage configurations. AI TOPS figures should likewise be treated carefully: vendors may quote NPU-only or total-platform figures, and these measurements are not interchangeable.
The supplied reporting also does not provide a reliable, source-backed specification for Acemagic’s Intel-based F7A memory limit or storage layout. Until Acemagic publishes a final F7A product sheet, those particulars remain unconfirmed.
Minisforum’s Gorgon Halo systems: MS-S1 workstation and N5 AI NAS
Minisforum formally unveiled two Ryzen AI Max+ PRO 495 systems at IFA: the MS-S1 MAX-P495 AI Mini Workstation and the N5 MAX-P495 AI Agent NAS.1
2
Both are built around AMD’s Gorgon Halo-class Ryzen AI Max+ PRO 495 processor and its Radeon 8065S integrated GPU. Minisforum rates the platform at up to 131 TOPS total AI compute, including up to 55 TOPS from the NPU.1
2
Their standout specification is the memory architecture:
- Up to 192GB of LPDDR5x-8533 unified memory
- Up to 160GB allocatable as graphics memory
- A shared CPU–GPU memory pool intended to reduce copying between separate system memory and VRAM
1
2
That capacity matters for local inference because model weights must fit into available memory. Minisforum explicitly positions the MS-S1 for local inference, LLMs, multimodal generation and other demanding AI workflows.2
The N5 extends the idea into storage infrastructure. Minisforum says the AI Agent NAS can support up to 200TB of local storage, combining an always-on storage appliance with local AI processing.1 That makes it more than a small desktop: it is intended to keep data and AI workloads close together in one private, local system.
How these systems compare with Nvidia RTX Spark-class boxes
AMD’s IFA pitch is strongest on model capacity and deployment flexibility.
A compact Gorgon Halo system configured with 192GB of unified memory—160GB of which can be available to graphics workloads—can hold substantially more model data than a compact system restricted to a smaller dedicated-VRAM pool.1
2 The MS-S1 offers that capability in a mini-workstation form factor, while the N5 incorporates it into a NAS. GEEKOM and Acemagic, meanwhile, put the more mainstream Gorgon Point platform into laptop and mini-PC designs.
That does not establish a universal performance victory over Nvidia hardware. CUDA ecosystem maturity, the inference runtime, model format, quantization, memory bandwidth and software optimization all materially affect real-world LLM throughput. TOPS ratings also cannot be compared directly across vendors or even across different portions of a platform.
The clearest takeaway from IFA is therefore not “AMD is faster than Nvidia.” It is that AMD-based compact systems are expanding the set of local-AI choices: a portable notebook for everyday AI-assisted work, a conventional mini PC with either Intel or AMD silicon, and unusually memory-rich mini workstations or NAS appliances for private local-model deployments.
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 "IFA 2026 Local AI PCs: GEEKOM, Acemagic and Minisforum’s AMD Gorgon Point and Gorgon Halo Systems"?
IFA 2026 split AMD’s compact local AI story in two: Gorgon Point machines such as GEEKOM’s GeekBook X16 Pro and Acemagic’s F2A handle mainstream on device AI, while Minisforum’s Gorgon Halo systems scale to 192GB of u...
What are the key points to validate first?
IFA 2026 split AMD’s compact local AI story in two: Gorgon Point machines such as GEEKOM’s GeekBook X16 Pro and Acemagic’s F2A handle mainstream on device AI, while Minisforum’s Gorgon Halo systems scale to 192GB of u... GEEKOM’s AMD GeekBook X16 Pro pairs a Ryzen AI 9 HX 470 with Radeon 890M graphics; Minisforum’s MS S1 MAX P495 workstation and N5 MAX P495 AI NAS use Ryzen AI Max+ PRO 495, with up to 160GB assignable as graphics memo...
What should I do next in practice?
Acemagic confirmed both Intel Core Ultra X7 358H and Ryzen AI 9 HX 470 versions of its F2A mini PC, but final source backed AMD memory and storage configurations remain unclear.[3]