- Ryzen AI Max+ 395128 GB unified memory
Full specifications
- Operating system
- Windows 11 Pro
- Processor
- AMD Ryzen AI Max+ 395
- CPU cores
- 16 cores / 32 threads (Zen 5), up to 5.1 GHz
- Graphics
- AMD Radeon 8060S integrated, 40 RDNA 3.5 compute units
- NPU
- AMD XDNA 2, up to 50 TOPS
- Memory
- 128 GB LPDDR5X-8000 unified memory (soldered)
- Storage
- 2 × M.2 2280 PCIe 4.0 slots
- Networking
- 2 × 10 Gigabit Ethernet
- USB4
- 2 ports
- Wireless
- Wi-Fi 7
Good for
- Running 70B models at 8-bit, or 120B-class models at 4-bit, on one small machine
- A quiet local-AI server shared over 10 Gigabit Ethernet
- Developers who want large memory without building a multi-GPU tower
Not ideal for
- Models of 30B and below: a single 24–32 GB graphics card runs them faster
- CUDA-only workflows
- Anyone who wants to add memory later; it is soldered
Beelink
Beelink GTR9 Pro
The GTR9 Pro is a compact desktop built around AMD's Ryzen AI Max+ 395, with 128 GB of unified memory shared between the CPU and a strong integrated GPU. That shared pool lets it load models far larger than a typical graphics card can hold, and dual 10 Gigabit Ethernet makes it easy to put on a fast network. Ships with Windows 11 Pro.
AED 17,589Charged in US dollars at checkout: $4,790. Import duties and taxes included.In stock. Dispatched in 1 to 3 business days.
- Free shipping to the UAE
- Ships from our warehouse in Malaysia
- 30-day returns
- Import duties and taxes included
- Shipping, returns and warranty
On Windows the GPU can be given up to 96 GB of the 128 GB. Check that the model you plan to run fits in that before you buy.
Who it's for
Anyone who wants to run 70B-class models at home in a small, quiet box without building a multi-GPU tower.
Full specifications
- Operating system
- Windows 11 Pro
- Processor
- AMD Ryzen AI Max+ 395
- CPU cores
- 16 cores / 32 threads (Zen 5), up to 5.1 GHz
- Graphics
- AMD Radeon 8060S integrated, 40 RDNA 3.5 compute units
- NPU
- AMD XDNA 2, up to 50 TOPS
- Memory
- 128 GB LPDDR5X-8000 unified memory (soldered)
- Storage
- 2 × M.2 2280 PCIe 4.0 slots
- Networking
- 2 × 10 Gigabit Ethernet
- USB4
- 2 ports
- Wireless
- Wi-Fi 7
About the Beelink GTR9 Pro
The Beelink GTR9 Pro is a small desktop built on AMD's Ryzen AI Max+ 395, with 128 GB of unified memory that the CPU and the integrated GPU share. That pool is what lets it load 70B models at 8-bit and 120B-class models at 4-bit, which no single consumer graphics card can hold. Buy it if you want to run large open models at home in one quiet box, with two 10 Gigabit Ethernet ports for serving them to the rest of your network.
Key features
- AMD Ryzen AI Max+ 395: 16 Zen 5 cores / 32 threads, up to 5.1 GHz
- Radeon 8060S integrated graphics with 40 RDNA 3.5 compute units
- 128 GB LPDDR5X-8000 unified memory on a 256-bit bus (soldered)
- Up to 126 TOPS of combined AI performance; the XDNA 2 NPU alone gives up to 50 TOPS
- 2 × 10 Gigabit Ethernet and 2 × USB4 (40 Gb/s)
- Two M.2 2280 PCIe 4.0 slots, up to 8 TB each; ships with a 2 TB SSD
- Built-in 230 W power supply; Beelink rates it at 140 W TDP at 32 dB
What it runs
On unified-memory machines we plan for about 75% of the pool to be usable by models, because the operating system keeps a share and the GPU has its own limits. On 128 GB that leaves roughly 96 GB, which covers every tier up to 70B at 8-bit and 120B-class models at 4-bit. A 120B model at 8-bit needs around 140 GB and does not fit. Our model fit tool shows the same check for every machine we sell.
| Model size | 4-bit | 8-bit |
|---|---|---|
| 7–8B | Fits | Fits |
| 13–14B | Fits | Fits |
| 30–34B | Fits | Fits |
| 70B | Fits | Fits |
| 120B+ | Fits | Does not fit |
How much of the 128 GB the GPU may address is set in firmware and by the operating system, so check the current BIOS and driver notes before planning around the very largest models.
Performance in context
Capacity is the strength here; raw speed is not. The memory bandwidth of a shared LPDDR5X pool is well below that of a high-end graphics card, so a model that fits on a 24 or 32 GB card will generate text faster there. The GTR9 Pro wins when the model is too big for those cards. Mixture-of-experts models suit it well, since only part of the weights is active for each token while all of them stay in memory. For a side-by-side with the other large-memory options, read Mac vs NVIDIA vs Ryzen AI Max+.
Ports and expansion
- 2 × 10 Gigabit Ethernet for serving models on the local network or linking machines
- 2 × USB4 at 40 Gb/s, also usable for displays
- HDMI 2.1 and DisplayPort 2.1; up to four 8K displays in total
- Two M.2 2280 PCIe 4.0 slots, up to 16 TB of storage
- Built-in microphone and dual speakers
Memory is soldered, so 128 GB is what you keep for the life of the machine. Storage is the part you can grow.
Power, cooling and noise
The power supply sits inside the all-metal case, so there is no external brick, and it accepts 100–240 V. Two turbine fans and a vapor chamber cool the chip at its 140 W setting; Beelink quotes 32 dB under that load. Expect it to be audible in a silent room during long jobs, and close to silent at idle.
Software
AMD lists Windows 11, Ubuntu and RHEL as supported for this processor. Common local-model tools such as llama.cpp, Ollama and LM Studio run on Ryzen AI Max systems. This is not a CUDA machine: if your workflow depends on CUDA-only libraries, look at the NVIDIA DGX Spark instead.
What to pair it with
A second fast NVMe drive keeps model files off the system disk, and a 10 Gigabit switch makes use of both network ports. If you need a PCIe slot for a faster network or capture card, the Minisforum MS-S1 Max has the same chip with one. Our Ryzen AI Max+ 395 mini PC comparison covers the differences in detail.
Good for
- Running 70B models at 8-bit, or 120B-class models at 4-bit, on one small machine
- A quiet local-AI server shared over 10 Gigabit Ethernet
- Developers who want large memory without building a multi-GPU tower
Not ideal for
- Models of 30B and below: a single 24–32 GB graphics card runs them faster
- CUDA-only workflows
- Anyone who wants to add memory later; it is soldered
What it can run
Up to 120B+ at 4-bit, 70B at 8-bit (approximate)
Based on 128 GB of unified memory, about 75% of which the GPU can use. See every machine by model size
Guides for this product
Mini PCs · 4 min read
Ryzen AI Max+ 395 Mini PCs Compared
Ryzen AI Max+ 395 mini PCs compared: Beelink GTR9 Pro or Minisforum MS-S1 Max, 128 GB of unified memory, ports, expansion and the models each runs.
Read the guide →Local AI · 5 min read
How Much Memory Do You Need to Run a Local LLM?
How much memory a local LLM needs: 7B to 120B+ models at 4-bit and 8-bit, how unified memory differs from VRAM, and which machines hold each size.
Read the guide →Reviews
What reviewers say
ServeTheHome
A big step up in design for Beelink. The reason to buy it is running large models such as 120B-class ones on a Windows machine; for gaming alone you can build something cheaper.
“Overall, this system is really good.”
Read the review
Atomrig customer reviews
No customer reviews yet. Bought this from us? Tell us how it runs.
Related products

Minisforum
Minisforum MS-S1 Max
- 128 GB unified
- AMD Ryzen AI Max+ 395
Runs up to 120B+ at 4-bit
The PCIe slot takes low-profile cards and runs at x4 lanes. It is ideal for a faster network card, not for adding a full-size graphics card.
AED 9,989
NVIDIA
NVIDIA GeForce RTX 5090
- 32 GB
- NVIDIA Blackwell (GB202)
Runs up to 30–34B at 4-bit
32 GB runs 30B-class models comfortably at 4-bit, but a 70B model will not fit on one card. For 70B, plan for two cards or a large unified-memory system.
AED 7,339
PowerColor
PowerColor AMD Radeon AI PRO R9700 32GB
- 32 GB
- AMD RDNA 4 (Navi 48)
Runs up to 30–34B at 4-bit
Check that your tools support ROCm on your operating system before buying. PyTorch and llama.cpp do; some niche tools still assume CUDA.
AED 4,769
ASRock
ASRock Intel Arc Pro B60 Creator 24GB
- 24 GB
- Intel Xe2 (Battlemage)
Runs up to 30–34B at 4-bit
This is a single-GPU card with 24 GB. It uses a PCIe 5.0 slot wired at x8 and also works in PCIe 4.0 slots; check ASRock's spec sheet for the exact length and power connector before you order.
AED 2,199