I have really annoying issue with my new intel NUC8i5BEH regarding the ethernet connection. There is massive packet loss and high RTT times on some quite weird situations. I'll try to clarify the issue as good as possible.
First when I received NUC I performed bios upgrade (BECFL357.86A.0056.2018.1128.1717) and disabled all unneeded fetures from bios before starting to install Ubuntu 18.04 LTS release to test some KVM virtualization in nuc and noticed that installation will fail because NUC is not able to download packages. Ethernet connection gets ip address but there is so huge packet loss that apt cannot donwload packages. This is easy to notify by pinging nuc or pinging gateway from nuc.
I tried all kind of debugging and performed bios reset etc. and after a while tried ESXi in nuc and suprisingly network worked. I also noticed that network was working with ubuntu live disk if I download drivers from intel and build e1000e module from intel source code. My first assumption was that there is buggy kernel driver in ubuntu's kernel. But...
Looks like I was gone little confused during those debugging-days and propably I had once again performed bios reset which made esxi to work temporarily because one day I noticed that network is no longer working in esxi either. That's after I have done my "final" configurations in bios (and disabled unneeded features).
I tried to find a specific setting in bios that might cause the issue to appear in network connection and managed to find it! And here is my conclusion.
If I go to bios and disable SD card reader (yes, you read correct) and start ubuntu 18.04 live linux, the connection is not working perfectly. Packet loss starts usually immediately but always after I start browser and go to some web page. Then ping starts to increase from < 1 ms to 100, 500, 5000 and 10 000 ms and also some packets are lost totally. After disconnecting the network cable or rmmod e1000e && modprobe e1000e it might be that the connection is working properly couple of seconds but then the issue is there again.
If I go to bios and set SD card reader again to read/write mode, everything seems to work quite ok. So what?? Disabling SD card reader breaks ethernet adapter? It might also be possible to fix this issue by compiling drivers from intel source code and modprobe e1000e (new module), but thats still really weird issue and I would like to hear some clarification from intel of some comments from other users. Ubuntu's kernel should support this version of adapter (Ethernet Connection (6) I219-V), because the device id is in place in kernel code and it will recognize adapter.
So I will be really glad if someone could try to reproduce this issue with same generation NUC. It's easy by flashing ubuntu desktop 18.04 to usb stick and booting from that after disabling sd-card reader from bios. Then just go to some web page with browser (maybe thats needed because that generates some traffic) and concurrently ping your network gateway. How does the ping delay behave?
Also some comments from intel would be really appreciated. There might be some really weird issue with hardware & bios & driver or then my NUC is broken. And I would like to get some solution because if nobody else cannot reproduce this, then I will return the product to reseller.
I tried also with older bios and the behaviour is exactly same.
I know that NUC is not verified to run linux, but also you might know at Intel the fact that quite many users are using linux on NUC and usually it works fine. This issue occurs in my NUC also with esxi and with Debian, so it is not Ubuntu specific issue.
It would be really nice, if someone could try to reproduce this with same generation NUC. It nobody can reproduce this during folliwing couple fo days, I do not have any other option than return this device because I assume there is a hardware issue.
Loading bios default settings does not help. I just need to reset bios to default and disable sd card reader to get this issue pop up.
Yes, of course, but it would be much easier to decide my belief about defectivity of the device, if someone could try to reproduce this and give me some input. Does Intel have debug devices available?
I am sorry that maybe I am too late to reply back to this community. I just found about it and I have to apologize for that.
I did the following test: loaded Ubuntu, latest I have available Ubuntu 18.04.1 LTS (I know that there is a more recent kernel version out there but I didnt try it), my system is a NUC8i5BE with 8GB of RAM and a regular M.2 Storage.
After several minutes of pinging:
and browsing and downloading packages for Linux, the pining time never exceeded 100ms, it kept within the range of 80 to 100ms.
If you had the system fully updated with the latest firmware and software and you were still having this issue perhaps replacing the unit under warranty was the right way to fix this issue.
I apologize again and please let me know if there is anything else I can do to help,
I'm having a NUC8i7BEH (BIOS version 0056) with 16GB RAM and M.2 storage and having the exact same problem in Ubuntu 18.04.1 LTS. If I change the SD card reader to read/write mode, it's working fine. If i disable the SD card in BIOS, the network will stop working within seconds after startup.
I'n not able to reproduce it when running Windows 10 as OS.
Extra info, apart from above and previous posts:
If I change the link speed of the port from 1000M to 100M (in the router/switch - not in the NUC) I'm not able to reproduce the problem.
Did you have a change to look into my reply?
Can you provide me with the exact steps to replicate this issue?
I did some testing as you can see on my previous reply and I had no issues at all.
Yes, I read your reply submitted on 11 February, that's why I added my response.
I have the NUC (setup with Crucial DDR4 CT16G4SFD824A RAM, Intel 760P SSDPEKKW256G8XT M2 drive) connected to a 1Gbit LAN, USB keyboard and mouse connected to yellow front USB port and a Samsung monitor HMDI-DVI input.
I updated to BIOS rev BECFL357.86A. Rebooted. Entered BIOS and resetted the bios settings. Rebooted and booted into Ubuntu. Results in normal network speed (970Mbit/s).
Reboot and enter BIOS. Change to disable the SD card reader. Save settings and reboot into Ubuntu. Results in 0.00bits/s
Network speed has been tested with iperf3 to a computer on the same network, but also something so simple as pinging www.google.com results in packet loss ang ping response times of 1sec to 5secs with 64bytes packets.
Sorry for not providing you with the BIOS version. I downloaded the latest available as of yeasterday.
The full name of the bios is (as written in the visual BIOS): BECFL357.86A.0066.2019.0225.1641
Underneith mu NUC it has a product code: BOXNUC8i7BEH2
I did some testing but couldnt replicate the issue.
I did a fresh installation of Ubuntu and went to the terminal and did a ping command to www.google.com and had not issues so I went to the BIOS and disabled the SDCard and ran the same ping command to www.google.com and had the same results.
See the screenshots below, I am previously running a lspci command so that you can see the SD card is enabled or disabled.
With the SD Card enabled in BIOS:
With the SD Card disabled in BIOS:
If you are having this issue, I would recommend that you contact Intel Customer Support and open a support ticket.
Here is the contact us information web site: https://www.intel.com/content/www/us/en/support/contact-support.html
We have been experiencing the same exact issue on 2 distinct nuc8i3beh devices. Disabling the SD Card causes a lot of network rx_errors and rx_crc_errors. We are running Debian Stretch (kernel 4.19.0-0bpo2-amd64) with same bios version as OP (BECFL357.86A.0056.2018.1128.1717). The e1000e module that ships whit it is version is 3.2.6-k while 18.104.22.168 is currently the latest (https://downloadcenter.intel.com/download/15817)
I didn't try upgrading the bios nor the device driver.
But what I find suspect is that the bio v0056 vanished from the download page (https://downloadcenter.intel.com/download/28638?v=t). @AmyC_Intel any reason for this?
We're about to launch a product with this version and discovering this makes me nervous.
@MK29 Did you find any other workaround yet other than re-enabling the SD card?