Marvell AQtion 10Gbit Network Adapter : Network link is lost.

D

dmsi154502d9

New member
Joined
Mar 19, 2021
Messages
29
Hi, every few hours (not predictably), I'm getting a 5-6 second network disconnection.

When it happens I see two events in the Windows Event Viewer from source aqnic650, first a Warning event: Marvell AQtion 10Gbit Network Adapter : Network link is lost.
Then when it restores, an Information event: Marvell AQtion 10Gbit Network Adapter : Network link has been established at 10Gbit/s Full Duplex.

This is connected over a short (<10ft) CAT6 cable to a TEG-7080ES, which has never given me any problems with other 10ge NICs I use (including an Aquantia card I bought a couple years ago).
I just switched the cable out with an even shorter CAT6 cable, although I would expect if it was a dodgy cable that the speed would degrade to a lower speed, not just cut out for 5 full seconds then return to full 10G with no other symptoms.

This issue feels more like a driver setting. I'm running v2.2.2.0, although looks like the newest version on MSI's support page is 2.2.1.0, so I would assume Windows updated the driver to something newer on its own. Maybe I should try down-grading? If this is a power management setting, I don't know which one it could be, because it happens both in the middle of the night when my machine is idling, as well as during online meetings while I'm working.

Anyone else seeing anything like this?
 

Attachments

  • aqnic650-events.png
    aqnic650-events.png
    61.2 KB · Views: 1,879
Sort by date Sort by votes
eqwebbie_with_PC said:
Please list your system specs, especially the board you are running.
Click to expand...

Oops, sorry, big oversight on my part!

I've got the MSI TRX40 Pro 10G, and am having the issue with the included 10G PCIe NIC that comes with the MB.

I have my GPU (nvidia 2070) plugged into PCI_E1, the included M.2 XPANDER-Z card plugged into PCI_E3, and the 10G NIC plugged into PCI_E5.
I also have an m.2 SSD plugged into M2_2 slot, and a SATA SSD plugged into one of the SATA ports, if it matters.
 
Upvote 0 Downvote
citay said:
Try disabling "Recv Segment Coalescing" (IPv4) and (IPv6) in Device Manager, if that does anything.
Click to expand...

I read about this issue, but my symptoms don't line up with the symptoms of those who found this change to be useful. I could still try it, of course, and probably will once I'm confident that swapping the cable didn't help.
 
Upvote 0 Downvote
I would check if Windows energy plan "PCIE power saving" is enabled, and disable it just for trying.
Installing latest Bios could be helpful, too.
Good luck!
 
Upvote 0 Downvote
eqwebbie_with_PC said:
I would check if Windows energy plan "PCIE power saving" is enabled, and disable it just for trying.
Click to expand...

Good thought, but it looks like that is already Off.

I'm hesitant to try a beta BIOS, but that is definitely something I should try.

This issue is really only inconvenient when it happens while I'm on a call. It doesn't seem to kill active SSH connections, which would make this issue far worse for me, personally.
 

Attachments

  • power-options.png
    power-options.png
    17.4 KB · Views: 1,989
Upvote 0 Downvote
Well, it kept happening no matter what I did, but I've been reading that this MSI card included with the MB is just poorly made, and sure enough, I swapped it with a Aquantia AQC-107 based card I bought a couple years ago, and so far the issue is resolved. It appears to be using the same driver in Windows, but so far no more drop outs!

UPDATE: nope, it wasn't the original card's fault, I've now had the same symptoms with this second card. Well, that's good, at least I don't have to go replace the one that came with the MB.
 
Last edited:
Upvote 0 Downvote
Just noticed the new 2.2.3 driver is available, and I still have 2.2.2 according to windows, so I installed it. We'll see if this helps. It is frustrating that it can go days without having the issue, and then happen multiple times in a matter of hours. I now keep an Event Viewer window open all the time, with a filtered view showing all the aqnic650 events from the past week.
 
Upvote 0 Downvote
Now I'm going to ask this because most people tend to forget the most basic solutions as years ago my electronics shop teacher pointed out to our class.

He told us that back when he was still in the field and most companies still used monstrous main frames, he was called out to help fix a problem with a mainframe. When he got there, there was three techs standing around arguing about what the problem must be. The first thing he did was to go over to the Mainframe power supply and sure enough the power cord was unplugged.

So what I want to ask you is if you have checked all your Network equipment and cables for any problems. - Not every problem is computer related.

Kinked Cables - Replace
Over Heating Network Devices - Clean out the dust from them, possibly add cooling fan/s - Heat failure due to dust or poor air circulation is the enemy.
Possible Network Devices Needing updating - Check for new Firmware, update if necessary
Check for any other possible problems.

Network cables can be upgraded to Cat 8 which are out now and will give you better performance for your network.

If the problem is not the network, but still persists, it could be your network connection which means you might need to contact your ISP to check your connection and the modem.

Good Luck
 
Upvote 0 Downvote
Thanks for your post! I agree it is very easy to overlook the simple solutions. I've been focusing on the new PC hardware/drivers since I never had this issue with my previous 10G capable Windows PC on this same network, but it doesn't hurt to make sure I haven't missed anything.

* Kinked cables: I've checked this previously, and there were no kinks, and the issue has happened while using different cables, each of which looked fine and had been used previously with no problems.
* Over Heating Network Devices: My 10G switch is well spaced from other devices, is warm but not hot to the touch, and the fans seem to be working fine. I've been using this switch for more than a year, and none of my other devices have experienced a similar issue, including the previous Windows 10 build that this PC replaced.
* Possible Network Devices Needing updating: As I said, I'd been focusing on the one new variable that had changed when this issue began, namely the new PC I built, but there was a firmware update available for my 10G switch, so I just now applied it. This also ensured the switch power cycled, which it probably hadn't done in a long time. The firmware release notes only mentioned improving performance for 2.5G and 5G connections (all my connected devices connect at either 1G or 10G). Otherwise, as I recently posted, I'm now on the new 2.2.3 Windows drivers for these cards, and I actually think this may have resolved the issue, as I haven't seen the issue happen for 3 days, which is the longest run I've gone without the issue being logged in Windows Event Viewer. I am planning on a writing a follow up post if I can go another 4 days without hitting the issue.
* Check for any other possible problems: My networking and PC are each plugged into their own UPS devices of appropriate specs for the respective plugged in hardware. And as I recently posted, I even tried swapping in a different Aquantia-based 10G NIC, to see if it was the MSI 10G Pro NIC (and it was not).

I will admit that I am using CAT6 cabling for my short 10G connections, but, again, I never had a problem with them on my previous Windows PC, nor do I have a problem with them on my current 10G linux PCs. They are all Monoprice cables, and kept stored in plastic bins that prevent the chance of kinking when not in use.

Thanks for your post, if you can think of anything else I didn't cover above, let me know, although like I said, I'm hopeful that it was a driver issue with the 2.2.1 and 2.2.2 driver release. I'll post again either way next Monday!
 
Upvote 0 Downvote
This system is about 3 months old, and has always had this issue, as far as I know. It took me a while to figure out what was going on, and once I found the Event Viewer entries, that made it much easier to track how often it was happening.

Granted, it works great something like 99.99994% of the time, it is just the unpredictability of the 5 second disconnects that are annoying.

And the issue happened again today, so I decided to buy a 10ft Monoprice CAT8 S/FTP cable, which arrives in 2 days. That should be entirely overkill for my needs, but if it solves the problem, I guess it'd be worth it. And in the mean time, I switched which port I'm using on the 10G switch.
 
Upvote 0 Downvote
dmsi154502d9 said:
This system is about 3 months old, and has always had this issue, as far as I know. It took me a while to figure out what was going on, and once I found the Event Viewer entries, that made it much easier to track how often it was happening.

Granted, it works great something like 99.99994% of the time, it is just the unpredictability of the 5 second disconnects that are annoying.

And the issue happened again today, so I decided to buy a 10ft Monoprice CAT8 S/FTP cable, which arrives in 2 days. That should be entirely overkill for my needs, but if it solves the problem, I guess it'd be worth it. And in the mean time, I switched which port I'm using on the 10G switch.
Click to expand...
I don't think it's the cable, the card is designed for cat6. I noticed it's a pcie3 card, I wonder if that's causing some issues. Have you tried running gen 3 only? Updated chipset drivers are a good idea as well. Personally I'd use the Intel NICs for latency sensitive communication.
 
Upvote 0 Downvote
Ha, so the whole reason I posted about this problem here was because this is the 10G NIC that is included with the MSI TRX40 10G Pro motherboard! If it wasn't supposed to be used with a PCIe 4 motherboard, then why did they package it in with one?
 
Upvote 0 Downvote
dmsi154502d9 said:
Ha, so the whole reason I posted about this problem here was because this is the 10G NIC that is included with the MSI TRX40 10G Pro motherboard! If it wasn't supposed to be used with a PCIe 4 motherboard, then why did they package it in with one?
Click to expand...
To sell you the motherboard. Also it's meant to work, but AMD are responsible for the code required. And they are hit and miss, bios updates might fix it. But threadripper doesn't get much support. To be fair it might be a driver issue. I'm just saying to try it.
 
Upvote 0 Downvote
The Ethernet Port on the card is RJ-45, not Cat-6 as RJ-45 is the type of connector used for the Ethernet Cable.

As for Cat-8 this can be purchased with RJ-45 Connectors or Non RJ-45 connectors.

The reason I posted the links earlier is so users can check out the difference in the cables from Cat-6 to Cat-8 especially seeing as the Op here appears to be running a high speed network which most home users typically don't have unless they are in an area where they can get Verizon FIOS or possibly Google Internet. Only other one I've seen is a T-1 connection which can be expensive the last I knew.
 
Upvote 0 Downvote
sybrwzar14c602c7 said:
The Ethernet Port on the card is RJ-45, not Cat-6 as RJ-45 is the type of connector used for the Ethernet Cable.

As for Cat-8 this can be purchased with RJ-45 Connectors or Non RJ-45 connectors.

The reason I posted the links earlier is so users can check out the difference in the cables from Cat-6 to Cat-8 especially seeing as the Op here appears to be running a high speed network which most home users typically don't have unless they are in an area where they can get Verizon FIOS or possibly Google Internet. Only other one I've seen is a T-1 connection which can be expensive the last I knew.
Click to expand...
I was referring to the chip the card uses, not the rj45. Maybe the cat8 works, but for such a short run I'm not betting on it.
 
Upvote 0 Downvote
sybrwzar14c602c7 said:
The Ethernet Port on the card is RJ-45, not Cat-6 as RJ-45 is the type of connector used for the Ethernet Cable.
Click to expand...

Yeah, but what @laurence1211 was saying is that the card is built to send 10G over CAT6. The distance is reduced from 6e or higher, but it is built to do it.

sybrwzar14c602c7 said:
Op here appears to be running a high speed network which most home users typically don't have unless they are in an area where they can get Verizon FIOS or possibly Google Internet
Click to expand...

Oh, no, my 10G network is entirely local, and was built for moving large files between local machines. Before I upgraded to 10G, the biggest bottleneck in my setup was gigabit ethernet. Now the limit is the spinning disk RAID in my NAS.

sybrwzar14c602c7 said:
Only other one I've seen is a T-1 connection which can be expensive
Click to expand...

I think T-1 is actually an older standard that was considered fast in the 90's, but only moves at something like 1.544mbps.
 
Upvote 0 Downvote
Back
Top