GMKtec / Enthusiast tier
GMKtec EVO-X2 (128GB)
Enthusiast-tier mini workstation: AMD Ryzen AI Max+ 395 with 128 GB of unified memory. Runs Llama 3.3 70B and Mistral Small 3.1 at roughly 15-25 tokens/s — 70B-class models without a discrete GPU, in a small-form-factor chassis.
- Sku
- CS-GEVOX2
- Chip
- AMD Ryzen AI Max+ 395
- Hdmi
- HDMI 2.1 ×1, up to 8K@60Hz
- Name
- GMKtec EVO-X2 AMD Ryzen AI Max+ 395 128GB RAM + 2TB SSD Windows 11 Pro Mini PC 2 years warranty (CS-GEVOX2)
The decision
Where it wins
- 128GB massive memory pool
- Ultra-compact form factor
- Dual NVMe storage slots
Where it doesn't
- APU graphics bottleneck token generation speed
- Audible fan noise under sustained load
Specifications
Structured product data, grouped by system.
Core hardware03
| Chip | AMD Ryzen AI Max+ 395 |
|---|---|
| Processor | AMD Ryzen™ AI Max+ 395, 16 cores / 32 threads, up to 5.1GHz |
| Runs | Llama 3.3 70B, Mistral Small 3.1 |
Memory & compute05
| Memory | 128GB LPDDR5X 8000MHz |
|---|---|
| Memory Form | Onboard LPDDR5X, typically not user-upgradeable |
| Memory Speed | 8000MHz |
| Memory Capacity | 128GB LPDDR5X |
| Unified memory | 128 GB |
AI & acceleration02
| Ai Npu | Based on AMD XDNA 2 architecture; AI TOPS performance varies by application, model, and settings |
|---|---|
| Throughput | 15-25 tok/s |
Display & media03
| Hdmi | HDMI 2.1 ×1, up to 8K@60Hz |
|---|---|
| Displayport | DisplayPort 1.4 ×1, up to 8K@60Hz |
| Display Output | HDMI 2.1 ×1, DisplayPort 1.4 ×1, USB4 ×2, supports up to 8K@60Hz display output, actual output depends on display, cables, and settings |
Physical & connectivity02
| Usb A | USB 3.2 ×3, USB 2.0 ×2 |
|---|---|
| Dimensions | Approximately 186 × 77 × 192 mm |
Other specifications12
| Sku | CS-GEVOX2 |
|---|---|
| Name | GMKtec EVO-X2 AMD Ryzen AI Max+ 395 128GB RAM + 2TB SSD Windows 11 Pro Mini PC 2 years warranty (CS-GEVOX2) |
| Audio | 3.5mm headphone / audio jack ×1 |
| Storage | 2TB M.2 2280 SSD; up to 2 M.2 2280 slots for expansion, actual capacity depends on SKU and hardware support |
| Warranty | 2 years |
| Bluetooth | Bluetooth 5.4 |
| Wired Network | 2.5G LAN ×1 |
| Sd Card Reader | Some official / retail data indicates SD Card Reader, actual shipping version varies |
| Operating System | Windows 11 Pro |
| Wireless Network | Wi-Fi 7 |
| Preloaded Storage | 2TB M.2 2280 SSD |
| Storage Expansion | Up to 2 M.2 2280 slots; expansion limit depends on actual unit and manufacturer data |
Read the record
- Source on record
- gmktec.com.hk
Packing 128GB of RAM into a mini-PC chassis enables running massive models that simply will not fit in standard consumer VRAM. While inference speed via CPU/APU is slower than discrete GPUs, the sheer memory capacity unlocks use cases previously restricted to cloud servers.
The EVO-X2 packs an AMD Ryzen AI Max+ 395 with 128GB of unified memory into a mini-PC chassis — enough to run Llama 3.3 70B and Mistral Small 3.1 at roughly 15–25 tokens/second without a discrete GPU.
What the memory unlocks
Models in the 70B class simply do not fit in standard consumer VRAM. 128GB addressable by the APU makes them loadable on a machine you can hold in one hand, drawing a fraction of the power a multi-GPU rig would. Dual NVMe slots handle the model library.
The trade-off is speed, as always with this architecture
The APU graphics bottleneck token generation — 15–25 tok/s is usable for interactive chat and retrieval, not for high-throughput serving or batch work. And the fan is audible under sustained load.
Who should buy
High-capacity context-retrieval workloads, CPU/APU-based LLM inference, and anyone who needs 70B-class capability in a small, quiet-ish box. If your target models fit in 16–24GB of VRAM, a discrete GPU runs them several times faster for less money.
Frequently Asked Questions
Who is GMKtec EVO-X2 (128GB) for?↓
What are the drawbacks of GMKtec EVO-X2 (128GB)?↓
What does GMKtec EVO-X2 (128GB) do well?↓
How much does GMKtec EVO-X2 (128GB) cost?↓
What is the Fathom Layer design score for GMKtec EVO-X2 (128GB)?↓
Alternatives to consider
See all alternatives →Memory on record
- Another option Mac Mini M4 Pro (64GB)
64GB unified memory accessible by GPU
64 GBDesign score 9.5/10 - Another option Ryzen 5 + RTX 4060 Ti Build
16GB VRAM at entry-level pricing
16 GBDesign score 7.5/10 - Another option Nvidia GeForce RTX 3090 (Used)
24GB of GDDR6X memory for handling large datasets and complex scenes
Explore its evidenceDesign score 8.5/10
Where to buy
Purchase links may earn Fathom Layer a commission, at no extra cost to you. As an Amazon Associate we earn from qualifying purchases.
Further reading
Featured badge
Building this product? Add the badge to your site to show it’s in the index.
<a href="https://fathomlayer.com/compute/local-ai-workstations/gmktec-evo-x2-128gb" target="_blank" rel="noopener noreferrer"><img src="https://fathomlayer.com/fathom-badge.svg" alt="Featured on Fathom Layer" width="250" height="54" /></a>