10G/2.5G LAN Speed Issues

[Traveler2100]

A have a GT AXE-16000 running Merlin firmware 3004.388.3_2. I have a desktop A connected with a 10G card and a desktop B connected to a 2.5G switch which is connected to the other router 10G port reporting a 2.5G connection. iPerf3 speed to/from desktop A to the router is ~3G and desktop B to/from the router is about 2G. The issue is iPerf3 speed from Desktop B to Desktop A is about 380 Mbits/s and vice versa. Given the results, cables and cards are apparently not the issue since I can get into the router’s 10G port at 2G and out the other 10G port at 3G but I apparently cannot get between the two ports at greater than 350/400 Mb. QOS is off. AIProtection is on but I don’t expect it to affect the local network. Jumbo frames are off, Spanning Tree is on. Am I missing something?

 

[ColinTaylor]

Log into the router's webUI and look at the CPU usage while the iPerf transfer is happening. What do you see?

Do you see the same throughput if you do a normal file transfer (e.g. SMB or FTP) between the two PCs rather than using iPerf?

 

[TheLostSwede]

Try a factory reset, I had some weird issues with the 2.5 Gbps LAN port on my GT-AX6000 being stuck at sub Gigabit speed. Factory reset solved it.

 

[Traveler2100]

Thanks all, I did do a factory reset but I did a backup first and then a restore so it isn't completely clean. CPU usage seems normal, one core near 50% the others lower. I have more surprising results. on a current model iPad Pro connected to WIFI 6, I pull about 850/900 Mbits/s on an internet speed test but only about 300 Mbits iPerf3 to the router and 250 to computer A. Must be something with the iPerf implementation, why/how could I route 900 to the internet but only 300 to the router? and 250 on a 10G connection? Computer A is a 16 core AMD with 64 G Ram running Win10. Network settings on the computer and router indicate a 10G connection. Something isn't measuring correctly it seems to me. I guess I'll need to do a hard reset and rebuild by hand next.

 

[ColinTaylor]

iPerf isn't a reliable test of throughput which is why I asked you to test more representative methods, like SMB or FTP.

Likewise, testing speeds to or from the router itself is also unrepresentative because the test becomes limited by the router's CPU.

 

[Traveler2100]

Hmm, maybe it is something with the router. Computer C is 1G wired directly to the router. iperf3 computer C to the router is ~940. Computer C to the 10G computer A is ~840Mbits. so C to A is 2 to 3 times faster than B to A or iPad wireless 6 to A. So the 1G connection acts pretty much as expected but higher speed 2.5G or wifi 6 connections to compter A do not.

 

[Traveler2100]

Ok Colin, SMB test include hard disk speeds and file sizes so I havent a test I was happy with, but I'll get a good test. I thought iPerf was better since it is all in memory. As for the router CPU, I can't imagine a transfer to the router is more CPU intensive than a transfer thru the router to the internet, etc. and the 16000 router has a lot of horsepower and is not showing a heavy load.

 

[TheLostSwede]

No issues with the iPerf3 implementation on my GT-AX6000, so I doubt it would be something weird with your router.

[ ID] Interval           Transfer     Bitrate
[  5]   0.00-10.01  sec  2.75 GBytes  2.36 Gbits/sec                  sender
[  5]   0.00-10.01  sec  2.75 GBytes  2.36 Gbits/sec                  receiver

 

[TheLostSwede]

iPerf isn't a reliable test of throughput which is why I asked you to test more representative methods, like SMB or FTP.

Likewise, testing speeds to or from the router itself is also unrepresentative because the test becomes limited by the router's CPU.

iPerf works just fine, why are you spreading fud about it?

 

[ColinTaylor]

iPerf works just fine, why are you spreading fud about it?

I'm not spreading FUD. This is from personal experience over many years on multiple platforms. The results vary dependant on the platform, version and the parameters being used.

I've found that the default iPerf options are sometimes not very good, particularly on Windows which is what the OP is using. To get accurate results I often need to specify multiple parallel streams and increase the buffer size. I've also seen significant differences between 3.1.3 and 3.8.1. But YMMV which is why a suggested a second testing method to validate the results of first.

I know you're aware of this article but have chosen to dismiss it.

 

[TheLostSwede]

I'm not spreading FUD. This is from personal experience over many years on multiple platforms. The results vary dependant on the platform, version and the parameters being used.

I've found that the default iPerf options are sometimes not very good, particularly on Windows which is what the OP is using. To get accurate results I often need to specify multiple parallel streams and increase the buffer size. I've also seen significant differences between 3.1.3 and 3.8.1. But YMMV which is why a suggested a second testing method to validate the results of first.

I know you're aware of this article but have chosen to dismiss it.

You sure love M$ propaganda.
iPerf3 is perfectly fine as a quick throughput test, but yes, it's not an actual file transfer test, two different things.
What iPerf3 is great at, is detecting potential issues, which is just what I did with my own router last week.
I never said it was a good file transfer test and I would never suggest that to anyone.