NUC8i5BEK - Thunderbolt eGPU Useless after BIOS Update

James · ‎09-29-2019

Hi,

I have a NUC8i5BEK that's been previously running absolutely flawlessly with a Radeon VII housed inside Razer Core X Chroma eGPU enclosure.

Today I decided to do some of the Intel updates, and as soon as the BIOS update was applied, Windows would no longer fully boot. It gets to the end of the booting stage with the spinning dots, and then hangs, before eventually powering off. It also seems as though the NUC is creating a video output on the HDMI port, even though nothing is attached, rather than disabling it which I believe it used to previously do when it detected the eGPU at boot.

I originally had BIOS version 0041 before the upgrade (yes, the original BIOS), and now after many hours wasted have realised I can not downgrade to this version because the management engine has been updated as part of the 0074 BIOS now installed. I did manage to downgrade to both 0073 and 0071, however to no avail for the issue described.

I have tried every combination of relevant BIOS settings to get it to work, including trying all the different security settings for Thunderbolt to no avail. I've also done a complete fresh reinstall of Windows 10 Pro, and the NUC functions absolutely fine except if it is booted up with the eGPU attached. I have disabled fast boot, made sure all drivers are up to date, etc. However the root cause of this fault was seemingly the BIOS upgrade. I have also checked with the Thunderbolt utility within Windows that the Core X devices are set as Approved devices.

After much faffing about, If I attach the eGPU once Windows is booted, it functions in the desktop environment, but Windows crashes with a Kernel error when many 3D applications that previously worked are launched, e.g. certain games game. Even if it worked by attaching after OS boot properly, this is an undesirable step backwards compared to the previous configuration, as it means I need to have something connected to the HDMI port of the Intel NUC to first get into the OS. Before all of this, I could turn the NUC on with the eGPU connected, and everything would work correctly.

I am incredibly frustrated that a perfectly working setup has suddenly become unusable for its intended purpose because of a routine BIOS update, and I have wasted an entire day trying to get it to work to no avail. The eGPU still works flawlessly on other systems, whether it's connected at boot up or after the host OS is loaded.

Please advise as to how this matter may be rectified, and what further information might be required.

Thanks,

James

AlHill · ‎09-29-2019

What version of Windows 10 are you using? 1903?

Have you installed the TB3 DCH driver?

https://downloadcenter.intel.com/download/28735?product=126147

Doc

James · ‎09-30-2019

Hi Al,

Thanks for your response -- Yes it's version 1903. And I have indeed installed the latest TB3 DCH driver.

Further pointing to the BIOS update being the culprit, is the fact that previously I could setup the entire NUC from start to finish with a clean Windows install, without any issues. However now, using the exact same image and installation method, Windows will hang at the same point during the first boot after installation. Nothing else has changed.

Is there a way to get a response from Intel directly, or do I have to wait for one of their representatives to notice the thread?

AlHill · ‎09-30-2019

One of the Intel support engineers will be along to assist you.

Doc

Esteban_C_Intel · ‎10-01-2019

Hello James,

Thank you for posting on the Intel® communities.

To better assist you, could you please tell me if you sequentially update the BIOS or run from version 0041 to 0074? Also, have you tried a BIOS recovery? In case you have not please find the instructions at the following link (perform the BIOS Recovery by the Security Jumper method and to the version 0074),

If the issue persists after that access to the BIOS and go to Devices>Video>IGD Primary Video Port and change it for Thunderbolt. Also, make sure that the IGD Aperture Size is lower than 2048MB.

Let me know the outcome.

Esteban Ch.

Intel Customer Support Technician

A Contingent Worker at Intel

James · ‎10-02-2019

Hi Esteban,

I went from 0041 to 0074 directly. Afterwards when I realised it was no longe functioning correctly, I reset all BIOS settings to defaults and tried various configurations to see if I could get it to work again. I have indeed tried a BIOS recovery for numerous versions, however as mentioned the lowest I could downgrade to was 0071. Anything lower than that, the system would complain about "Incompatible BIOS Version".

I have since tried the method using the security jumper to re-flash 0074 to no avail. I have also tried changing the Primary Video Port to Thunderbolt, again to no avail. The IGD Aperture Size is set well below 2048MB.

Could someone at Intel please try and re-create this issue? I'm genuinely convinced something has been introduced during the BIOS updates between 0041 and 0071 that have caused this.

Esteban_C_Intel · ‎10-02-2019

Hello James,

Thank you for your response,

Please allow me to research a little bit, and as soon as we have an update we will let you know.

Esteban Ch.

Intel Customer Support Technician

A Contingent Worker at Intel

Ronny_G_Intel · ‎10-22-2019

Hi James,

Unfortunately, we have not been able to replicate this issue, Alex who I know has been working with you directly has tried a couple of e-GPUs different brands and models with no issues so far.

Also, from what I understand in your previous note to Alex, if you install a NVIDIA GPU with the same NUC you dont have any issues, only with the Radeon VII GPU. This is, of course, an indication of a compatibility issue and usually, it takes a long time to resolve.

Having said that I have a couple of recommendations:

1- Reach out to Radeon VII support, they could have a fix or workaround for this issue.

2- Try these settings in BIOS:

3- If none of these works, please give us a call to explore other possible options, here is the contact us information https://www.intel.com/content/www/us/en/support/contact-support.html#@6 (make sure that you select the correct region) and make reference to case 4363355.

Regards,

Ronny G

Ronny_G_Intel · ‎10-29-2019

Hello James,

I see that you have been working directly with Alex on this issue but I would recommend continuing this conversation via the public community so that other community members can also contribute or benefit from it.

I had a conversation with Alex about this issue, he has been doing a lot of testing with the e-GPUs we have available but unfortunately we haven't been able to replicate the issue, we have this report pending internally but we dont know if we are going to be able to come out with a fix for it and if we do, this may take a long time.

If you have already tried the BIOS settings I provided before and the issue persists, please reach out to our technical support service so that we can explore other options, make reference to case 4363355.

Regards,

Ronny G

James · ‎11-05-2019

Hi Ronny,

I'm almost certain at this point the issue pertains to the larger 16GB memory of the Radeon VII and some change in the BIOS between versions on the NUC8i5BEK. I managed to acquire a NUC7i5BNK, in which I installed my SSD and everything worked flawlessly without crashes (even on the latest BIOS version for that model). I then managed to borrow a NUC8i7 with the original BIOS version, and lo and behold it worked on that up until I updated the BIOS.

I've tried the settings you suggested, however it made no difference at all. I also tried reinstalling the OS after making all those changes and still crashes after the Radeon VII driver is installed (followed by the Code 12 Error).

All other eGPU setups where there's a Code 12 error point to something to do with memory mapping. Is there any chance the Intel Engineers could just review what changes were made between versions with this issue in mind, particularly around an eGPU with a larger sized memory, in case its obvious?

Thanks,

James

Esteban_C_Intel · ‎11-11-2019

Hello James,

We recommend you to reach out to our technical support service so that we can explore other options, make reference to case 4363355.

Esteban Ch.

Intel Customer Support Technician

A Contingent Worker at Intel