AMD · Local AI machine
Beelink GTR9 Pro (128 GB)
AMD machine: AMD Ryzen AI Max+ 395, 128 GB memory, 96 GB for the GPU, 256 GB/s memory speed, about 3.9 tok/s on a 70B model (estimate).
78.5Score
- #31 of 41 ranked
- #15 of 18 AMD machines
Why this score
The score adds up these parts. Each part has a weight. A part counts as the average of its sources, pulled toward 80 when there are few sources.
AI power
50% weightcounts as 80.7Average of 2 sources: 81.1
Experts
30% weightcounts as 80.0No sources found yet. No data yet. This part counts as 80.
Users
20% weightcounts as 70.4Average of 1 source: 60.8
Shown, not counted
These sit inside an average we already count, or come from the maker's own shop.
Thin: only 1 review or buyer rating, so it is pulled toward the average.
80 is the field average. A part counts as 80 + (average − 80) × N ÷ (N + 1), where N is its number of sources. One source counts half of the gap from 80. No sources count as 80.
Price
€4619 at Beelink EU (eu.bee-link.com) · Maker's EU shop, ships to Austria
128 GB LPDDR5X + 2 TB SSD, EU plug, 3-year warranty; struck-through list price EUR 5,199; shipping to Austria not stated on the page
Prices checked 8 Oct 2026.
Specs
- Chip
- AMD Ryzen AI Max+ 395
- CPU
- 16 cores / 32 threads
- Memory
- 128 GB
- GPU memory
- 96 GBWindows/BIOS: AMD Variable Graphics Memory, up to 75% of RAM = 96 GB on 128 GB (AMD rule; the largest share the vendor tools allow). ServeTheHome: out of the box 96 GB video / 32 GB CPU. Tuned Linux maxima need kernel parameters and are not used: up to 124 GiB (kyuz0 toolboxes on a Framework Desktop, Fedora, amdgpu.gttsize=126976, ttm.pages_limit=32505856, 4 GiB left to the OS), about 110 GB (runaihome), about 108 GB stable (Jeff Geerling).
- Memory speed
- 256 GB/s
- FP16 TFLOPS
- 60 TFLOPS
- NPU
- 50 TOPS
- OS
- Linux, Windows
- Software
- ROCm, Vulkan
- Ethernet
- 10 Gbit/s
- Thunderbolt / USB4
- USB4
- QSFP port
- No
- Idle power
- 16 W
- Load power
- 128 W
- Noise
- 41 dB(A)
- Volume
- 2.9 L
- Weight
- 2.3 kg
- SSD
- 2 TB
- Release
- 2025-08
Measured
- gpt-oss-120b, not stated
- 31.4 tok/s writing · LM Studio (Windows 11)
- Llama 3.3 70B, not stated
- 4.9 tok/s writing · not stated (ServeTheHome out-of-box test)
- Qwen3.6-35B-A3B, UD-Q4_K_M
- 62.6 tok/s writing, 1059.5 tok/s prompt · llama.cpp Vulkan/RADV (llama-bench, Ubuntu 24.04, b9049)
Notes
One of the best-connected Strix Halo boxes: dual 10 GbE, dual USB4 and a built-in 230 W power supply in 2.94 L, which suits clustering several boxes. ServeTheHome measured 31 tok/s on gpt-oss-120b at about 128 W and 39-41 dBA, but the EU shop asks EUR 4,619 for 128 GB.
Noise: maker claims 32 dB at 140 W TDP (test conditions not given), ServeTheHome measured 36-37 dBA idle and 39-41 dBA during gpt-oss-120b in a 34 dBA studio (distance not stated), a reader comment on that review reports 52 dBA on a returned unit (unverified). Notebookcheck's teardown article repeats 36-41 dBA. Idle power: ServeTheHome 16-25 W under Windows 11 Pro. Price: ServeTheHome quoted USD 1,999 in October 2025, the EU shop now asks EUR 4,619. The Beelink page shows 'OpenClaw and local LLM pre-installed' variants with the same EUR 4,619 listing. candidates.json cites USD 2,399 for 128 GB/2 TB from Beelink's international shop; not re-found, the EU shop shows EUR 4,619. Geizhals lists the 128 GB/2 TB model since April 2026 with no Austrian offers. GPU memory follows AMD's Variable Graphics Memory rule for every box with this chip (75% of RAM: 48 GB on 64 GB, 96 GB on 128 GB); tuned Linux figures are in the GPU memory note. FP16 TFLOPS is AMD's figure for the chip.
More reading:servethehome.comservethehome.comservethehome.comnotebookcheck.netturbovs.com