What's new

10GB NIC only getting 1GB speeds

  • SNBForums Code of Conduct

    SNBForums is a community for everyone, no matter what their level of experience.

    Please be tolerant and patient of others, especially newcomers. We are all here to share and learn!

    The rules are simple: Be patient, be nice, be helpful or be gone!

skylinekiller

Occasional Visitor
I have two Win 10 Pro PCs (both fully updated as of today) networked using a ASUS XG-C100C and the other is using my MSI X570 Creation builtin NIC Aquantia AQtion. The mother board to the ASUS is a Rampage IV Gene with a i7-3970X CPU, and the other is my X570 with a 3950X CPU. This is connected via a CAT6e cable directly with the ips configured.


I have Jumbo Packets set ti 16348 Bytes and also changed Link Speed to 10G on both NICS. I get the same speeds with a CAT 5 or CAT6 cable. I used LAN Speed Test using a 2GB

If someone knows of another tester I would like to hear about it.
 

Attachments

  • Test Results.pdf
    193.8 KB · Views: 341
I'm sorry, but I do not understand "iperf with multiple threads to try to saturate the link ?" can you dummy it down for me?
 
i tried to figure it out. I donwloaded iPerf3.exe and I don't know what to do next. I tried running a speed test, but it failed. Both PCs are running Windows 10 and have two network cards. Internet is conencted to the 1GB and the PCs are connected to each other through a CAT6 cable.
10GB NIC settings
PC#1 IPV 4 (TCP/IPV4) settings are ip address 192.###.#.78 with gateway of 193.###.#.77
PC#2 IPV 4 (TCP/IPV4) settings are ip address 192.###.#.77 with gateway of 193.###.#.78
 
I have two Win 10 Pro PCs (both fully updated as of today) networked using a ASUS XG-C100C and the other is using my MSI X570 Creation builtin NIC Aquantia AQtion. The mother board to the ASUS is a Rampage IV Gene with a i7-3970X CPU, and the other is my X570 with a 3950X CPU. This is connected via a CAT6e cable directly with the ips configured.

I have Jumbo Packets set ti 16348 Bytes and also changed Link Speed to 10G on both NICS. I get the same speeds with a CAT 5 or CAT6 cable. I used LAN Speed Test using a 2GB
Sometimes you don't always measure what you think you're measuring. For example your "read" speeds are triple your "write" speeds. That discrepancy suggests your data resides on disk; which is fine if you're speed testing your disk drives or, maybe, "fast (100 Mbps) Ethernet" but the disk drive is likely to be slower than Gig and 10G Ethernet.

It's likely you can reconfigure your speed test such that all the data is RAM resident.

Another "pinch point" (besides disk speed) it the computer itself. Probably not in your case but it was on one of mine. Had an old laptop with "Fast (100 Mbps) Ethernet" (connected to router with 400 x 20 Mbps service). Internet bench showed 100 Mbps. Upgraded to Gigabit Ethernet. Speed test only went up to 120 Mbps. Conclusion; that's all that old machine could handle.
 
Collin, thank you about the Gateway, I deleted those, though it did not impact transfer speeds.

Klueless, I am not sure how the app LAN Speed Test works, but my hard drive is an AORUS NVMe Gen4 SSD, and my other is a Samsung SSD
 
I have two Win 10 Pro PCs (both fully updated as of today) networked using a ASUS XG-C100C and the other is using my MSI X570 Creation builtin NIC Aquantia AQtion. The mother board to the ASUS is a Rampage IV Gene with a i7-3970X CPU, and the other is my X570 with a 3950X CPU. This is connected via a CAT6e cable directly with the ips configured.


I have Jumbo Packets set ti 16348 Bytes and also changed Link Speed to 10G on both NICS. I get the same speeds with a CAT 5 or CAT6 cable. I used LAN Speed Test using a 2GB

If someone knows of another tester I would like to hear about it.

Try this.
1. 3rd party Firewall: uninstall.
2. 3rd party Antivirus: uninstall.
3. Jumbo frame: disable.
 
I don't have any 3rd party firewall/Antivirus installed yet. I didn't have jumbo frame, but i had Jumbo Packet, and I disabled both, but still had the same exact read/write speeds
 
LAN Speed Test works by reading and writing a file to a folder. Typically that is a shared folder on a network drive. This is not a good network test because the bottleneck is usually the read/write speed of the drive. But even taking that into account the speeds you're seeing on your speed test are diabolically bad. That says to me that there is something fundamentally wrong with your network.
 
It was in the PCIe 4x and I moved it to the PCIe 16x, but still same exact speeds. My PCIE_X16 Link Speeds are set to GEN3 in the bios too.
 
Last edited:
I'm doing this from memories of 10G introduction time frames and I'll also add that I've now been retired for almost 4 years so memories fade. Forgive me if I've missed and/or mis-spoken below. Just trying to give you ideas.

iPerf was the spot check tool of choice for my techs when customers reported such "out of whack" performance. They always used dedicated Linux images but I believe it's available on Windows as well. Google "10G testing with iPerf" and you'll get a LOT of help. Also, I'm referring to around the Sandy Bridge time frame and before for my comments below. ALWAYS used 10G on XEON platforms, never desktop. Not saying that's an issue, just putting it out there.

The biggest problems I recall they found were
  1. The networking hardware just wasn't up to the task - even at base link speed testing most of the available hardware could only deliver 50-70% link rate.
  2. Misconfiguration - as I noted above, they often found the cards in the wrong slots and/or misconfigured slots. Also make sure you're using slots connected to the main CPU and not a secondary chip. I'm guessing those are the "red slots" as highlighted in the manual below. You should be able to find a block diagram for the MB
  3. Bad cable/connections - Try different cables and make sure it's a simple direct connect. My experience is with optical only, no copper. According to this, cat 6E is not necessarily appropriate but i didn't do an exhaustive search. https://www.cablesys.com/updates/cat6-cat6e-cat6a-differences/
  4. Machine just not up to the task - Try a different machine

https://dlcdnets.asus.com/pub/ASUS/mb/LGA2011/Rampage_IV_GENE/E7091_Rampage_IV_Gene.zip
upload_2019-12-28_10-53-7.png


EDIT - Don't know if this applies or not
https://rog.asus.com/forum/showthre...etting-on-the-chip-thru-device-manager!/page2
 
Last edited:
Iwas trying to get into the rog.asus forum but I am having issues logging in. I did update the driver and no change. Look forward to seeing that forum topic.
 
first verify that the links between are all 10Gb/s. I use SFP+ direct because its cheaper and ensured.

When testing, use iperf, it eliminates some memory bottleneck and doesnt inolve drives. SATA 3 controller is only 6Gb/s but individual drive speeds vary.

Nothing to do with jumbo frame. 1GB/s = 8Gb/s.
 
I am using iperf 3.1.3 for Windows 10. I typed the following in CMD as administrator

iperf3.exe -c 192.168.6.76


PC-B ASUS 10GB NIC IPv4 address is 192.168.6.76
PC-A ASUS 10GB NIC IPv4 address is 192.168.6.77

It appears 3.1.3 is not compatiable with the latest Win 10 updates and the other one has been removed from the Windows store
https://iperf.fr/iperf-download.php
 
Last edited:
I am using iperf 3.1.3 for Windows 10. I typed the following in CMD as administrator

iperf3.exe -c 192.168.6.76

Were you also running the iperf3 server on the target machine? You'll also need to turn off the firewall on the target machine while you do the test.
 
Being that it is working a 1 gig levels I think you are being limited by layer 3. I would guess that some how the data is being routed by 1 gig connection.
 
I was finally able to do an iPerf test. These are the results I am getting. I guess this is what it should be?
 

Attachments

  • iperf teste results.jpg
    iperf teste results.jpg
    93.2 KB · Views: 925

Latest threads

Sign Up For SNBForums Daily Digest

Get an update of what's new every day delivered to your mailbox. Sign up here!
Top