A company best known for CPU coolers and thermal paste has shipped one of the most aggressively specced mini PCs on the market. Thermalright's HydroNous R1 pairs an AMD Ryzen AI Max+ 395 processor with 128GB of LPDDR5x-8000 unified memory and an integrated liquid cooling loop, making it one of the few small form factor machines that can genuinely sustain heavy GPU workloads without throttling. The closed-loop system lets the Strix Halo chip pull up to 176W at peak and hold 153W continuously, numbers that would be unthinkable in a passively or fan-cooled chassis of this size.
The Ryzen AI Max+ 395's Radeon 8060S integrated graphics can address the full 128GB memory pool as shared VRAM, a setup that is increasingly attractive for local AI inference and creative workloads that choke on the 8 to 24GB limits of discrete GPUs. The Strix Halo platform already has a growing open-source inference stack behind it, with AMD's ROCm documentation providing a dedicated system optimization guide and community projects such as amd-strix-halo-toolboxes and llama-cpp-for-strix-halo shipping pre-built llama.cpp binaries that target the gfx1151 GPU and can address the full memory pool for inference. Thermalright ships the R1 with a 2TB PCIe 4.0 SSD and three M.2 2280 expansion slots, so storage is unlikely to be a bottleneck either. Connectivity is similarly generous: dual USB4 Type-C ports at 40 Gbps, dual USB 3.2 Gen 2 Type-A, HDMI, DisplayPort, WiFi 7 via a Qualcomm FastConnect 7800 module, and both 10 Gigabit and 2.5 Gigabit Ethernet jacks. A 4.6-inch LCD panel on the front of the chassis rounds out the build, offering system status at a glance.
The whole package measures 23.5 x 8.3 x 13.3 cm (9.2 x 3.3 x 5.2 inches) and is available now in China for approximately $3,500 (€3,220). Thermalright has not announced pricing or availability for North America or Europe. For anyone running local LLMs, Stable Diffusion, or heavy compilation workloads on Linux, the combination of massive unified memory, sustained thermal headroom, and 10GbE networking makes this a compelling purpose-built node, provided it eventually reaches broader markets.
