Question ASUS XG-C100C 10g adapter slow to connect and regularly disconnects

Tanyac

Reputable
If you've used Windows with mapped drives you'll be familiar with the "Could not reconnect all network drives" error that has been around since Windows 7. There are no absolute fixes, but several workarounds. One such is to run a batch script at log on which reconnects the drives so you don't get that dreaded red X and a bunch of disconnected drives.

Connected to a 1GB motherboard NIC the process is essentially instantaneous. I have a task set up which runs at log on and runs a vbs script to reconnect all drives.

I recently purchased a bunch of ASUS XG-C100C 10gb NICs. The PCs they are installed in connect via Cat6A to a Netgear MS510TX switch, in my PCs case, to the 10g port.

When I log on the vbs script fails because the XG-C100C has not completed initialization, or is not ready (however you want to phrase it). originally the script had no delay, but I'm now up to 6 seconds and it's still failing.

As there are scripts and start up programs trying to access my server, the XG-C100C then logs a "Link lost" error in the event log (ID: 14). I'm told this is different to the "Connection Lost" error that the latest ASUS driver allegedly addresses (Which is the same event log error with the same ID).

The Network notification icon shows the network as connected as soon as the PC logs in, but it's actually not up.

Additionally, loading web pages from my web server are substantially slower with this card.

Has any one experienced this, know of any fixes, or are there any tweaks to settings that this card needs to run optimally?
 

Tanyac

Reputable
All cabling is new CAT6A, LSZH (https://cplonline.com.au/alogic-c6a...le-10g-shielded-cat6a-lszh-network-cable.html)
They are in PCIe3 x16 slots. I don't have any motherboard with x4 slots (only x1 and x16). According to ASUS this is the recommended installation.
Switches are Netgear MS510TX and I've worked with Netgear to ensure they are fine.
ASUS have asked me to return the cards under warranty, though the likelihood of three DOA cards is remote. My feeling is that this is a driver or firmware issue, something that has been echoed by other networking buffs on other forums. So I've sent them one card back for testing. ASUS Australia are abhorrently slow at warranty issues, so I expect not to hear back until July 2019 (About 10 - 12 weeks).
 
I will assume you only have a single card in each machine and have not attached the 1gb ethernet to the same switch. Many switch/cards support bonded connections and this is negotiated it can delay the start of a port. You may want to disable spanning tree on these port if you can on your switch since port delay passing data to avoid loops. You of course have to be very careful that you only have devices connected with a single nic card to these ports.

Can you tell if it is a hardware or software type of delay. If you are using DHCP sometime this gets delayed. Setting the ip static will eliminate that issue if it exists.

Speeds wise it is strange it is slower. You can try a program like iperf on 2 of your machines and see how fast data can be transferred between them. This program is simplistic and should not be affected by other things in your machine like disk delays.
 

Tanyac

Reputable
Thanks @bill001g. Looks like good advice.
I have only the 1 PCIe card in each machine, but all systems have dual Intel NICs and Wifi on the motherboard. Those NICs are all disabled.
The advice I got direct from Netgear re the switch is that having cables connected from the same PC to the same switch should not cause an issue on these switches. But I am going to give your suggestion a try, even though the NICs are disabled.
Yes, everything here is DHCP on the PC. The router (RT-AC88U), has reserved IPs assigned to every MAC address. So I'll also try static IPs, though I suspect that one isn't going to improve anything, but its worth a try.
My gut feeling is that this is a hardware delay caused by the XG-C100Cs. The latest driver allegedly fixed a "link connection lost" issue, but it doesn't really seem to have fixed. it.
The data transfer speeds are erratic (more so than on the 1G ports, but that could be a limitation of the server disks to which I am copying data). The slowness that I have issues with is the length of time it takes the XG-C100C to detect the link. It's so long that the auto-VPN connection fails, as does the mapping of drives.
STP is configured, well actually, RSTP. I've disabled it on the 10G port.
I'll have to try each of these things one at a time. That'll take me a day or two.
 

Tanyac

Reputable
  1. Turning off STP at a port level disables the port. I turned off STP globally on both switches - Did not improve things. I did some reading on STP and it seems like something I would want enabled.
  2. Removing all redundant cables to a single PC (leaving only the 10G port connected), did not improve things. I'm glad for that as I want to keep one of my 1G links for my Windows 7 boot and 1 for Windows 10 as a backup.
  3. I'm not using link aggregation (LACP Bonding). It's not configured in the switch, no NIC teaming config has been set up in Windows and the 1G ports are disabled.
  4. I remove the fiber link and SFP+ transcivers and ran the copper CAT6A cable from port 9 on switch #1 to port 9 on Switch #2, then connected both PCs to the 5G ports on the switches. No improvement
  5. I have an ASRock Z390 Phantom Gaming 9 motherboard here with an Aquantia 2.5G port. It experiences no delays at all (It connects to the 2.5G port on Switch #1, but I also tried the 5G and 10G ports. No delays).
I have a ticket raised with Netgear but the ticket system has been down for the last 4 days.
 

TRENDING THREADS