GTR9 Pro (Ryzen AI Max+ 395) - Hard Freeze & Instant Reboot

TLDOS

System Specifications:

Model: Beelink GTR9 Pro (Revision Version 1)

Processor: AMD Ryzen AI Max+ 395 (32 CPUs) ~ 3.0GHz
Graphics: Integrated Radeon 8060S
Memory Configuration: 128GB RAM (Configured as 64GB System RAM / 64GB Dedicated VRAM via AMD Adrenalin)
OS: Windows 11 Pro 64-bit (Build 26200)
Current BIOS: GTRP112 (Updated from GTRP110)

Issue Summary:

The system suffers from a sudden hard freeze and an instantaneous reboot only during the very first local AI workload execution after the mini PC is powered on. This occurs during intensive neural-network tasks like Topaz Gigapixel AI upscaling and Adobe Photoshop “Remove AI”.
The issue persists on the latest GTRP112 BIOS.

The Exact Reproducible Pattern:

The Cold-Start Crash: Turn on the PC from a powered-off state and open Photoshop or Topaz. Run any local AI model. The system instantly freezes, cuts power, and reboots.
The “Perfect Run” Post-Reboot: Once the machine automatically reboots and returns to Windows, running the exact same local AI tool on the exact same file works flawlessly. It will continue to work perfectly for hours without a single crash, regardless of how many heavy AI tasks are thrown at it.
The Reset: The crash only returns if the device is shut down completely and turned back on again.

Technical Analysis:

This highly specific symptom strongly indicates a hardware power delivery initialization fault rather than a software or driver bug:

In-Rush Current / Capacitor Droop: On a fresh boot from a powered-off state, the motherboard VRMs (Voltage Regulator Modules) or power rail capacitors are likely not fully conditioned or optimized for sudden transient power spikes.
The Surge Trigger: When the APU attempts to pull maximum wattage instantly from an idle state to process the heavy 64GB VRAM/NPU matrix calculations, it causes a severe voltage drop. This trips the motherboard’s Over-Current Protection (OCP) or Under-Voltage Protection (UVP), forcing an instant reset.
Why it works after reboot: Following the hard reset, the power rails remain energized, or the power management controllers adjust their voltage offsets, allowing the system to handle the massive burst load successfully from that point forward.

Troubleshooting Attempted So Far:

Successfully flashed the GTRP112 BIOS.
The issue occurs regardless of VRAM allocation (tested at both default and maximum 64GB configurations).
Windows Power Plan is forced to “High Performance” to try and stabilize power transitions.

Requested Support:

Is there an updated AGESA / CPU microcode patch or an internal beta BIOS that alters the VRM power-up states to prevent this cold-boot voltage drop?
Are there specific power management or telemetry settings in the Advanced -> AMD CBS BIOS menu that can be tweaked to soften the initial load spike?

IainT

This sounds like the Intel Ethernet chip hang and bluescreen. Try going to device manager and uninstalling +deleting the driver for both Intel network ports. Leave them as unknown ethernet adaptor. I suggest the crash probably go away. You can only do this on the first boot; after the reboot the PCIe root port disappears along with both ethernet ports. The reboot seems to be caused by the Intel driver reacting badly to the ethernet chip going away.
There is an updated BIOS (P112, which you’ve got) with updates AGESA which made zero difference. There have been many NVM updates for the Intel NIC chip. None have helped either.
Beelink keep suggesting there may be a software fix later, but have been doing so for a year or so and the only fix that seems to work is a warranty swap for the rev 2 board with realtek ten-gigabit ethernet. But that means sending the whole thing back and waiting 6+ weeks for a replacement to be shipped. They won’t do advance replacement.

Beelink CS-Ian Chan

Hi , There

Please try some solutions here :
CLR CMOS
First, please unplug the power supply and HDMI cable.
Secondly, Please hit the “CLR CMOS” button/hole for about 10 seconds.
For some models the CMOS hole could be in the Audio port .
Please wait until 10 minutes later (THIS IS VERY IMPORTANT),
Then plug the power supply and HDMI cable back in, turn on the PC again,
once turn on the PC, you will need more than one minute to wait for the PC to enter the system.
See if it works. Please try few more times.

BIOS reset
Please go to BIOS save&exit page, reset to the default .

Please update this driver : https://url.bee-link.cn/GRtq

Go to the windows terminal (admin) > type in : sfc /scannow > then wait until finished.

TLDOS

Beelink CS-Ian Chan

Hi Ian,

Thank you for the troubleshooting steps. I have followed all of your instructions carefully. Here is the status of each item:

CLR CMOS: Done. (Unplugged cables, held button for 10 seconds, waited over 10 minutes before rebooting.)
BIOS Reset to Default: Done
(Reset via the Save & Exit page.)Driver Update via provided link: Attempted. (Failed with Error 1603: A General Error Has Occurred During the Installation of the AMD Software Package. The chipset driver refuses to install.)
SFC /Scannow via Admin Terminal: Done. (System files verified successfully.)

Unfortunately, none of these steps resolved the issue. In fact, the system behavior has severely degraded:

Increased Frequency / Random Crashing: The issue is no longer limited to the initial local AI workload during a cold start. The system now randomly freezes and reboots every 10 to 30 minutes, even when completely idle or doing basic tasks. This frequent instability started immediately after updating to the GTRP112 BIOS.
Missing LAN Drivers Post-Crash: I noticed a highly specific hardware symptom. Immediately following one of these random freezes and automatic restarts, the LAN (Ethernet) controller completely disappears from Windows Device Manager as if it does not exist. The network hardware only reappears if I perform a full, manual power cycle (shutting down completely and turning the PC back on).
Chipset Communication Failure: The fact that the official AMD installer triggers a critical Error 1603 strongly implies that the OS can no longer properly communicate with the motherboard chipset, likely due to corrupted PCIe lane/bus allocation in the new BIOS.
Recent Ethernet Firmware Flash Details: Previously, the LAN port lights would stubbornly stay turned on even when the device was completely powered off (as long as it was plugged into an active wall socket). To fix this, I successfully flashed the official Ethernet firmware package (E610_XAT2_FW_8000F3CD) found in your download section. While that flash successfully resolved the phantom LED light issue, I strongly suspect that this firmware update, combined with the new GTRP112 BIOS, has introduced a severe conflict on the PCIe bus or the Intel E610 controller, causing the current system crashes.

TLDOS

Hi Ian,

Thank you for the troubleshooting steps. I have followed all of your instructions carefully. Here is the status of each item:

CLR CMOS: Done. (Unplugged cables, held button for 10 seconds, waited over 10 minutes before rebooting.)
BIOS Reset to Default: Done
(Reset via the Save & Exit page.)Driver Update via provided link: Attempted. (Failed with Error 1603: A General Error Has Occurred During the Installation of the AMD Software Package. The chipset driver refuses to install.)
SFC /Scannow via Admin Terminal: Done. (System files verified successfully.)

Unfortunately, none of these steps resolved the issue. In fact, the system behavior has severely degraded:

Increased Frequency / Random Crashing: The issue is no longer limited to the initial local AI workload during a cold start. The system now randomly freezes and reboots every 10 to 30 minutes, even when completely idle or doing basic tasks. This frequent instability started immediately after updating to the GTRP112 BIOS.
Missing LAN Drivers Post-Crash: I noticed a highly specific hardware symptom. Immediately following one of these random freezes and automatic restarts, the LAN (Ethernet) controller completely disappears from Windows Device Manager as if it does not exist. The network hardware only reappears if I perform a full, manual power cycle (shutting down completely and turning the PC back on).
Chipset Communication Failure: The fact that the official AMD installer triggers a critical Error 1603 strongly implies that the OS can no longer properly communicate with the motherboard chipset, likely due to corrupted PCIe lane/bus allocation in the new BIOS.

jotapesse

Missing LAN Drivers Post-Crash: I noticed a highly specific hardware symptom. Immediately following one of these random freezes and automatic restarts, the LAN (Ethernet) controller completely disappears from Windows Device Manager as if it does not exist. The network hardware only reappears if I perform a full, manual power cycle (shutting down completely and turning the PC back on).

TLDOS This a well known issue since Sep 2025 where the Intel E610 NIC crashes into a “firmware recovery” state for yet not well understood reasons, scenarios or triggers. Different units of the same GTR9 Pro v1.0 have different behaviours, not all are affected and the ones affected not all are in the same way. This was extensively discussed on another thread: GTR 9 Pro Ethernet Malfunction under load. A summary recommendation of it can be read on this post.

Beelink CS-Ian Chan

TLDOS
Hi , There
Please email us for further assistance with purchase record and SN code picture.
PC after-sales service support:
Support-PC@bee-link.com

TLDOS

jotapesse
The Idea is I don’t use the LAN ports I just work on WIFI no cable is connected at all

jotapesse

TLDOS I don’t think that will fully prevent the issue from manifesting itself. And for sure won’t resolve it. If your unit can’t be resolved with those Intel NVM firmware and driver updates then my advice to you would be to have the unit replaced with the newer revision GTR9 Pro v2.2 under warranty.