In order to utilize GTR Pro full potential for AI applications, I’m required to install Ubuntu 24.04 along with AMD ROCm (which is not supported on windows 11). After installing Ubuntu, I noticed a recurring problem with the built in ethernet adapters. Whenever the computer encounters load on the GPU, the ethernet card would crash and the only way to get it to work again is to unplug my machine from power. It seems that other users of GTR Pro are also experiencing this issue: https://community.intel.com/t5/Ethernet-Products/Intel-Ethernet-Controller-E610-Firmware-Recovery-Mode-Under-Load/td-p/1719424 Linux Kernel is 24.04c oem as required by ROCm Things I’ve tried: Installing the ubuntu LAN driver version 6.0.6 provided in the beelink driver package - fails under ubuntu 24.04 Installing the latest LAN driver from Intel website - The ethernet card still crashes Installing an older version (6.1.4) - The ethernet card still crashes. Please advise as this makes the machine unusable as a server under Linux. It happens on gaming distributions such as Bazzite, as well.

I am on manjaro linux, and it seems like you can just use the intel default package for nvm updates for the 610 NIC ( https://www.intel.com/content/www/us/en/download/845408/non-volatile-memory-nvm-update-package-for-intel-ethernet-e610-series-linux.html ) to flash it to 1.30. Just make sure you can see the nic with lspci before you start (if not, coldstart).

f70w I attempted the above, no luck unfortunately. Beelink is working on a solution

Sorry for the trouble. Currently, the Lan issue of the GTR9 395 persists, and its root cause lies in the Ethernet driver. As of now, there is no new driver available to resolve the problem of the Intel E610. The current workarounds are as follows: Perform a physical power-off and re-plug the power cable. We suggest you to to use a Wi-Fi connection instead. A complete resolution to this issue will only be possible once a new driver is released

Beelink CS-George I was just about to place an order for this device… after reading these posts and the implicated lack of urgency in your response (not even and indicated time-line) to this major issue. I’ll hold off and quite likely look at the Framework product.

Oh, and the fact that your driver web page is inaccessible…

Welly Hi there, it is accessible, here is the link https://dr.bee-link.cn/?dir=uploads%2FGTR%2FGTR9-395

Welly Sorry for the inconvenience. We already urge Intel to develop a good driver. We will update it once it is released.

Beelink CS-George I will keep an eye out for this news.

Thanks for the update but what’s the confidence level that the new driver will resolve this issue? Are you able to reproduce it? What is the root cause? Could you share more details? Thank you very much.

LEPT Thank you for your understanding and patience. decryp7 Thank you for your response. The issue stems from missing drivers, which have caused the Ethernet functionality to become inoperable. New drivers are currently under development.

GTR 9 Pro Ethernet Malfunction under load

McSe

Hello Beelink,
I acknowledge Beelink’s efforts in resolving the problem.
However, I don’t think a partial solution is a good recommendation.
It would be better if Beelink provided a complete and tested solution.
Therefore, I would prefer to wait a little longer.

jotapesse

However, I don’t think a partial solution is a good recommendation.
It would be better if Beelink provided a complete and tested solution.
Therefore, I would prefer to wait a little longer.

McSe So you prefer to stay with a not working state instead of a an improved state (potentially resolved)? That doesn’t make much sense. Wouldn’t it be best for you to try it and see in your specific case how much does it improve or even if it resolves it for you? At least you try it, then if you have reasons for it you can complain about it.

jotapesse

BIOS Update (Fixes Mouse Lag Caused by Network Card)
BIOS Download Link: https://url.bee-link.cn/rsWn
Step-by-Step Tutorial: https://url.bee-link.cn/YU3p

Beelink CS-George To be correct this should not be addressed as a “BIOS” Update. It’s actually the Intel E610-XT2 Firmware (NVM) Update 1.31 previously mentioned here and here.

McSe

jotapesse

Well, there are different opinions on the possible solutions. But I expect perfect functionality from PC hardware that costs $2000..

tbocek

jotapesse The problem is, unless BeeLink mentions what “improved” is, I assume that nothing was improved with respect to the NIC crash. And when other users read, that its still crashing, i.e. when you read my message above where I tested it with the NIC crash still occurring, then the conclusion is clear: nothing improved wrt the NIC crash, and I would also not update.

(and to be fair: BeeLink mention in that update that NIC is still not fixed)

jotapesse

Well, there are different opinions on the possible solutions. But I expect perfect functionality from PC hardware that costs $2000.

McSe Well, too late for that. Although in principle and in a perfect world I would agree with you, the world isn’t perfect and in the same way no product is as well, no matter the cost. And as far as I can see all other products based on the same hardware platform are having their own issues and quirks. It’s just too new and the software for it just doesn’t seem to have matured yet. Having said this I also would like to see this resolved once and for all and as soon as possible. But as far as I can see Beelink is doing its best and no wrong doing. Which means making the improvements (possible up until now) available for everyone to apply them. It’s of course up to you not to use them - it just doesn’t make any sense for me. But that’s ok, to each their own.

jotapesse

The problem is, unless BeeLink mentions what “improved” is

tbocek Yes, this I agree. We need better communication and full transparency. A more detailed release note explaining what improvements, fixes or new features were done/added should be in place. Specifically for the Intel E610-XT2 Firmware (NVM) Update 1.31 - they mentioned a fix for “Mouse Lag Caused by Network Card” but it would be good to know more details about it. The Intel E610-XT2 Driver Update 30.5 - mentions a generic “Enhances Network Card Stability” but again it would be good to understand exactly what was enhanced and why.

Nirantali

As far as I know, it was said that the fix from Intel would come in December.
It’s not December yet, so there’s still hope.

freeflyer454

I ordered my GTR-9 directly from Beelink on Sept 30th, and haven’t received the unit or heard a peep out of them since. Anyone know if they’re waiting for the fix before sending them out? 🤷 I thought about cancelling my order because of these issues, but now that the prices of all these AMD Max+ 395 class PC’s have skyrocketed, I might as well stick with what I’ve ordered and hope it’s not a dud.

Nirantali

I now tested the it btw.
Didn’t take much… 🙁
Did run Prime95 and Furmark = Stable, No Issues.
Did run “Unigine Superposition” = NIC Crashed! No Internet, both NIC LEDs on, Reboot didn’t help, had to unplug Power cord.

So yeah i’m now also waiting for a fix, if no fix till 15. Dec 2025 (17. Nov + 30 Days = Return Deadline 17. Dec) I gonna need to return it.
An exception would be if it can be guaranteed that the fix will take place later in December and/or the return deadline will be extended.

I hope that it’s really just the firmware and/or the bios and it can be fixed and not a hardware design issue, because that would be sad.

Infos:
Bios: 1.08
Intel NIC driver: ixw.sys v1.2.43.0
NIC Firmware: 1.31

Reported Errors:

NDIS Fatal Error: The Miniport has detected an internal error.
The Network Adapter has been removed.

Interesting behavior found with GPU and NIC:
When I run Furmark with GPU Shark to monitor usage and temps and at the same time run iPerf3 full 10Gbit to another Server the GPU Core Load climbs up from 97% to 100% and as soon iperf3 sent its 10GB it goes back to 97% again. It’s like if, for whatever Reason using NIC also uses GPU Core. But interestingly the NIC didn’t crash during the 15 Mins of Furmark and sending a total of 200GB over NIC and GPU Temps stayed stable at 69°C.

I then started Unigine Superposition again, NIC instant crashed within 5 secs accompanied with a drrrzzzd sound and Unigine hanged.
I also tested Timespy, instant NIC crash and System freeze with Reboot!

So Furmark runs fine 15mins with the same time sending 200GB over NIC, but Unigine Superposition and Timespy instant crash NIC and the later even the whole System. 🤔

Another Interesting behavior found with GPU and NIC:
I managed to run Unigine Superposition in a Window without NIC crashing, but the green NIC LED blinked constantly, you know like it’s losing the connection every second. So I checked the Event Log and of course I found hundreds of “The network connection was interrupted”.

This is now looking more and more like a hardware design flaw, for example NIC uses the same PCIe lane as GPU or a power issue of some sort that causes the NIC to get not enough power or something. If that is the cause, then unfortunately no firmware, software or whatever update will help, then the board will have to be revised to fix it.

Another possibility that just came to mind is that the PSU is too weak and struggles with transient loads.

This thought came to me after the last crash when I started Unigine Superposition and the NIC lights went completely out and Windows said “A problem has occurred with a device and Windows needs to be restarted”. This would of course be bad, but could at least be done by sending a new PSU instead of all devices having to be sent back. As long as replacing the PSU is relatively easy.

Final Test:

Disable NIC (Device 3 Fun 2: Disabled)
Attached Thunderbold4 10Gbit NIC

Result:
Unigine Superposition suddenly runs smoothly without any stuttering and with much higher FPS as before and reached higher Score (+8k).
3DMark Timespy for the first time runs smooth without any stutter and most importantly without instant BSOD in the first Scene.
So the NIC when enabled, not only crashes, but also takes away performance from the GPU and leads to instabilities throughout the system.

My conclusion:
Hardware-level design flaw or Hardware-level power delivery issue or both.

McSe

?? … Should Beelink fail to resolve the highly probable hardware problem by December, then all buyers from Q3/Q4 2025 will receive a free Thunderbolt 4 docking station as compensation (ho-ho, just a joke…).

Nirantali

McSe
The issue with those Thunderbolt 4 10Gbit NICs are that they themselves have an annoying fan that still runs even when the Beelink was shutdown. I need to unplug it every time to stop that annoying fan. So yeah either I rip out that fan or get one without fan.

However, if it is really a hardware-level design flaw or a hardware-level power supply issue that can only be resolved with a new board revision, the device is unusable for its intended networking purpose while maintaining system stability and a return is required.

For me the only question now is whether to get a partial refund and keep the device or just return it and get a full refund.

McSe

Nirantali
An interesting point… Beelink is currently expanding its business in the EU/UK/US and aims to secure market share with “outstanding” PC hardware. If Beelink were to refuse to take back the defective hardware, it would significantly damage its business.

jotapesse

However, if it is really a hardware-level design flaw or a hardware-level power supply issue that can only be resolved with a new board revision, the device is unusable for its intended networking purpose while maintaining system stability and a return is required.

Nirantali If that proves to be the case then yes, that will be required to resolve and I’m sure Beelink will stand by it.

Nirantali

McSe
They provide 30-Day Risk-Free Trial, so I guess they wouldn’t refuse to take it back at least within those 30 Days.
But the 1 Year Warranty they provide is already a bit meager for European standards, as there is a minimum warranty period of 2 years.

And yeah, a refusal would be very bad for their image and business, but just the same if not admitting a unfixable hardware design flaw and stall customers with promises of an easy firmware fix until the 30-Day Risk-Free Trial are over.

Because customers don’t easily forgive or forget such things. Gigabyte as an example once disappointed me like that, that was 20 years ago. I haven’t bought a single Gigabyte product since, I basically permanently blacklisted them because of it.

Nirantali

Here also some calculation that shows that the PSU(rated 230 watts) in the GTR9 is most probably undersized.

Total Estimated Wattage
Summary of average/peak estimates:
APU: ≈140W (peak boost)
NIC: ≈5.1W (peak as of Intel)
2x NVMe(PCIe Gen4 ×4) SSDs: ≈20W (average)
RAM(128GB LPDDR5x): ≈20W (average)
Motherboard/Chipset: ≈40W (average)
Ports/Readers(2xUSB3.2, 2xTB4, SD/TF Reader, Fingerprint Sensor): ≈15W (average)
Total Estimated Peak System Load ≈ 240W

And to account for power efficiency, transient power spikes, and system longevity, choosing a PSU that provides at least 25% overhead is best practice. Therefore, a power supply rated for 250W to 300W or higher would provide sufficient, stable power for all components at full load.

So the power supply is already running above its rating and we don’t even know if it’s a bronze, gold or platinum.

jotapesse

Here also some calculation that shows that the PSU(rated 230 watts) in the GTR9 is most probably undersized. (…) Total Estimated Peak System Load ≈ 240W

Nirantali Your calculations… they don’t seem to add up. You also seem to know more than everyone else. Have you tried using an AC Power meter to confirm those? Let’s also look at other similar specced AMD Ryzen AI Max+ 395, 128GB LPDDR5X, dual M2 NVMe PCIe 4.0 ×4 mini PCs:

Brand Model	Max TDP	USB 2-3-4	LAN Ethernet	Extra I/O	PCIe Exp. Slot	PSU Type	PSU Power
Beelink GTR9 Pro	140W	0-2-3	2× 10 Gbps	No	No	Internal	230W
Bosgame M5 AI	140W	2-3-2	1× 2.5 Gbps	No	No	External	240W
Framework Desktop	140W	0-4-2	1× 5 Gbps	Yes	Yes	FlextATX	400W
GMKtec EVO-X2 AI	140W	2-3-2	1× 2.5 Gbps	No	No	External	230W
HP Z2 Mini G1a	120W	2-3-3	1× 2.5 Gbps	Yes	No	Internal	300W
Minisforum MS-1A Max	160W	2-3-4	2× 10 Gbps	Yes	Yes	Internal	320W

👉 Typically they use 230-240W Power supply units. Exceptions are the ones offering multiple extra I/O ports configurations and/or an internal PCIe 4.0 ×4 expansion slot which understandably will then use higher power rated PSUs from 300W.

👉 Also, all reviews of the Beelink GTR9 Pro stress testing 100% the CPU+GPU (worst case scenario) show a maximum power consumption (AC power meter) of about 180W only. So there seems to be a confortable leeway of Power available, even with room to populate/use all USB ports and M2 NVMe connectors. So your estimated calculations simply do not add up to what is confirmed reality.

Having said this, yes there is a chance that some of the Beelinks GTR9 Pro PSU power rails may be badly designed or shared with several high power devices devices and unable to cope with their power demands at given scenarios. But that’s speculation at this point and there is no clear indication or evidence for it.

Or it can simply be that the Intel E610-XT2 NIC is having compatibility/reliability issues with AMD platform chipsets and require Driver/Firmware fine-tuning. That wouldn’t be the first time for other Intel NICs similar cases in the past. And we know from several other reports that the Intel E610-XT2 NIC crashes on the GTR9 Pro can also happen on idle or light APU use cases (some cases even while the NICs are disconnected/not used at all) meaning very low power demands. 🤷

Nirantali

jotapesse
Now these are just estimated values (I don’t have exact numbers), but someone measured 180w from the socket with stock BIOS settings (balanced). And from the socket that is already crazy if you then take into account possible conversion inefficiency.

In addition, it only crashes for me in Unigine Superposition and 3DMark Timespy, which produces around 58°C, but not in Furmark, which produces 69°C, so it can’t be the temperature. But it could also be because Furmark doesn’t load a lot of data, whereas the other two might stream data from the NVMEs, but I would have to test it more closely.

And yes, it doesn’t matter whether the NICs are connected or you are using them, it is enough if they are simply switched on, for me another indication that it is something hardware related. Whatever it is, it is fatal because it destabilizes the entire system, leads to stuttering and then to hard crashes with system reboots. The only way to get the system to be 100% stable is to turn off the NIC completely, which is simply not acceptable for a $2k+ product.

What I found most blatant was the green Link Connection LED that started to flicker and lost the link almost every second when Unigine Superposition was running in a window. By the way, the NIC didn’t crash there. Are both devices, GPU and NIC competing for shared resources (PCIe bandwidth, power delivery via the motherboard VRMs) or are the NICs or something not adequately shielded or PCIe lanes too close together on the board and creating electromagnetic interference (EMI)? Whatever it is, it’s fatal.

Either way, I no longer believe it’s just a firmware problem that can be easily fixed with simple software updates. Rather, I now believe that this is a serious hardware-level design flaw or hardware-level power delivery problem that requires a recall of all devices and a completely redesign and revision of the board.

Of course, this is extremely annoying and will be expensive for Beelink, but if that’s the case, then it’s still better to just admit it and bite the bullet than to keep stalling customers with promises that there will soon be a simple solution that enables stable use of the System with enabled NICs as if they were switched off.

I also hate beta testing hardware for manufacturers; to be honest, it’s enough for me if I have to do that with games and software. If I buy a $2k device, I expect it to be extensively tested before it goes on the market. I only tested it for about a day and ran into various fatal errors, which makes you wonder whether the devices were properly tested before they were marketed.

I’m sticking with it, either there will actually be a fix soon or I’ll expect a 50% refund (I keep the device) or I’ll send the device back (full refund and I’ll never buy a device from Beelink again). The alternative would be a callback and complete revision and replacement, extremely annoying, but then I would perhaps, depending on how it is handled, at least consider continuing to buy products from Beelink in the future.

Nirantali

jotapesse
The table is not completely complete for my taste, but I don’t know what the other devices have.

Brand Model	Max TDP	USB 2-3-4	LAN Ethernet	Extra I/O	PCIe Exp. Slot	PSU Type	PSU Power
Beelink GTR9 Pro	140W	0-3-2	2×10Gbps + Wifi7	Yes (SD/TF, Biometrics, BT)	No	Internal	230W

*There are only 2 x USB4, 2xUSB3.2 10Gbps (USB-A) and 1xUSB3.2 10Gpbs (USB-C)

I also had to correct my Table:
APU: ≈140W (peak boost)
NIC: ≈5.1W (peak as of Intel)
2x NVMe(PCIe Gen4 ×4) SSDs: ≈20W (average)
RAM(128GB LPDDR5x): ≈20W (average)
Motherboard/Chipset: ≈40W (average estimate)
2xUSB3.2 10Gbps (USB-A) ≈9W (Minimum according to standard)
Dual turbine fans ≈1W (peak)
Total Estimated Intermediate Peak System Load ≈235W
1xUSB3.2 10Gpbs (USB-C) ≈7.5W (Minimum according to standard)
2xUSB4 ≈15W (Minimum according to standard)
1xFingerprint Sensor ≈1W
1xSD/TF Reader ≈1W
1xWifi7/BT ≈7W (average)
Total Estimated Peak System Load ≈266W

But a 230W PSU can only provide a maximum of 230W, regardless of its 80 Plus efficiency rating (Bronze, Gold, Platinum, etc.) that efficiency only affects how much is pulled from Socket for providing those 230W. But depending on whether the values for the Mobo/Chipset are correct (it can be less but also more), the power supply capacity could already been exceeded without USB4, USB-C, SD/TF, Biometrics, WiFI7/BT. In any case, there is not any reserve left for transient power spikes. And so regardless of whether this is the cause of the crashes, the power supply should be viewed as undersized. In the best case this only affects power efficiency, in the worst case it endangers system stability and leads to crashes.

If we now knew exactly how the components are connected, we could possibly determine the cause. It would also be interesting, for example: whether the 5.1w that Intel specifies is actually correct, or whether some of them might actually pull more. With the power supply already operating at the limit, that could be enough to destabilize the entire system and when this then coincides with transient power spikes also could cause crashes.

What I will now test is Furmark with iperf3 (Has been stable so far) + Crystal Disk to generate NVME traffic. If it then crashes, then it can be assumed that Timespy and Superposition are crashing because there additional data is being streamed from the NVME. This may be enough to overload the power supply and if that is the case, then the crashes would also depend on whether something is accessing the NVMEs or not, for example. random scans, updates or other background activities which would also explain the intermittent nature of the crashes that some experience.

jotapesse

Now these are just estimated values (I don’t have exact numbers), but someone measured 180w from the socket with stock BIOS settings (balanced). And from the socket that is already crazy if you then take into account possible conversion inefficiency.

Nirantali 180W measured at the mains oulet and “considering power conversion inefficencies” means that the harware itself is actually consuming less power than those 180W as some of it is wasted on the PSU. But hey, hypotetically, let’s assume that under BIOS Performance mode it goes up to 200W (measured at the mains oulet), it would still be well under the 230W rated PSU. But my point was: if this was to be an issue - then it would also be an issue for all other AMD Ryzen AI Max+ 395, 128GB LPDDR5X, dual M2 NVMe PCIe 4.0 ×4 miniPCs using similar power rated PSUs. And that’s not the case so no reason it would be the case for the GTR9 as well. Something else is at play here.

👍 Anyway, we all agree that we shouldn’t have to deal with this at this time and for so long, this should have been extensively tested well before launch and Beelink should get to the bottom of this as quickly and with upmost respect and transparency for its customers. If proven to be a hardware design issue and a hardware design review replacement will be required to resolve then it should admit it and compensate its customers for all the trouble and inconvenience in the best possible and reasonable way.

🙏 Let’s hope that’s not the case really and that an easier resolution is possible with an Intel driver/firmware upgrade, as acknowledged by Beelink up until now.

jotapesse

The table is not completely complete for my taste, but I don’t know what the other devices have.

Nirantali This table was not meant to please your taste, nor to include all specifications and details for a full comparison. That’s not its purpose. 🙂 It was meant to refer to the device capabilities and differences most relevant or likely to affect the power demands from the PSUs on the several mini PCs - which was the subject of your allegations (power estimations/calculations).

USB 2-3-4: 0-3-2 *There are only 2 x USB4, 2xUSB3.2 10Gbps (USB-A) and 1xUSB3.2 10Gpbs (USB-C)

👍 You are correct here, the 3rd USB-C port (front panel) is USB 3.2 10Gpbs compliant only. My bad. But in the end not really relevant for the discussion of the hardware power demands. Total number of ports is the same with no significant power demand difference.

Extra I/O: Yes (SD/TF, Biometrics, BT)

👎 No, no. That’s not the intented purpose for that column. Again, from the hardware power demand perspective, all of them have similar (not exactly the same) SD card readers, Wi-Fi and Bluetooth devices as standard. The Extra I/O: column is meant to refer to the capability of configurable/changeable extra I/O ports such as USB3, USB4, Ethernet LAN or others not available on the standard default I/O ports configuration. Some mini PCs allow you to add new I/O ports modules or change them later and provide the necessary hardware and power connections/support for it. Point being that only those mini PCs offering multiple extra I/O ports configuration options and/or an internal PCIe 4.0 ×4 expansion slot require higher power rated PSUs (from 300W) to support them. All others use similar 230-240W power rated PSUs which are sufficient.

✋ Anyway, let’s stop this. We are wasting too much time on it and you seem to keep trying to make points which don’t seem to be relevant to this discussion. And you don’t seem to want to understand it: regardless of your estimations/calculations…

👉 My point was: if this 230W power rated PSU was to be an issue - then it would also be an issue for all other similarly specced mini PCs using similarly power rated PSUs. And it isn’t. It’s on par with all other similarly specced mini PCs. And… The Intel NIC crashes also happen on idle or light usage - meaning during very low power demand/usage. Which indicates that it isn’t power demand/supply nor thermal issues related. Something else should be at play here.

« Previous Page Next Page »