What's new

Download speed is consistently lower than upload speed on one PC (Lumos Networks fiber Internet)

  • 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!

rmiller1959

Regular Contributor
I have had great luck with the smart people on this forum before, and maybe you can help me figure this out. We just had a Lumos Networks 2 Gbps fiber Internet service installed, and the speeds are incredible - on one computer:

Screenshot 2023-04-05 162203.jpg


This desktop system is connected via Ethernet to a Zyxel 2 Gbps unmanaged switch, which is, in turn, connected to the 2 Gbps LAN port on a Netgear Orbi RBRE960 router. So far, so good.

However, the speeds on the other computer, which is also hardwired, are pretty different. While the upload speeds are impressive, the download speeds are considerably lower:

Screenshot 2023-04-05 155018.png


This computer is linked to another Zyxel 2 Gbps unmanaged switch connected to the same switch as the computer above. The cable that links the two switches has a coupler linking two Ethernet cables. What do you make of this? If the upload speeds are the same or better, shouldn't the download speeds also be?
 
I have had great luck with the smart people on this forum before, and maybe you can help me figure this out. We just had a Lumos Networks 2 Gbps fiber Internet service installed, and the speeds are incredible - on one computer:



This desktop system is connected via Ethernet to a Zyxel 2 Gbps unmanaged switch, which is, in turn, connected to the 2 Gbps LAN port on a Netgear Orbi RBRE960 router. So far, so good.

However, the speeds on the other computer, which is also hardwired, are pretty different. While the upload speeds are impressive, the download speeds are considerably lower:



This computer is linked to another Zyxel 2 Gbps unmanaged switch connected to the same switch as the computer above. The cable that links the two switches has a coupler linking two Ethernet cables. What do you make of this? If the upload speeds are the same or better, shouldn't the download speeds also be?

Edited for clarity.
First, I'm assuming you're using all 2.5G connections (not 2G which would be bonded 1G in LAG which works differently)?

Try disabling firewall/antivirus on the PC while doing the speed test. If your results are significantly better, your PC just isn't up to the task of 2G throughput. I suspect that is the issue since the download is suffering more than upload (it scans the download much more thoroughly than upload). If that is the issue, you can either buy a new powerful PC, or just print out the speed test result, frame it, and be happy, since you likely do not need anywhere near 2G to a single PC.

If that doesn't help, watch the CPU in the PC during the speedtest, is it hitting 100% even with firewall disabled? If so, again, new PC or just be happy that you have more than enough speed, and your connection can support multiple computers all with a ton of speed at the same time.

You could try a better NIC in the PC, something from Intel (preferably server class) with its own dedicated processor on it. That may alleviate some of the CPU usage for the actual network traffic, but won't help with CPU usage caused by firewall/antivirus. You can see if there is a different firewall program that is more efficient but I doubt you'll find one that is twice as efficient.

If neither of the above, start checking wiring. Especially that RJ45 coupler, if it is some generic one, those are notoriously crappy. You can pick up good true Cat5e or better ones at home depot or from a good brand name off amazon. But also check your other wiring. If you have a laptop, test it at each point, before the coupler, after the coupler, at the final end near the PC, etc. Make sure the cable is not running along side power wires, or get a shielded cable if you have no choice. Distance needs to be about 330 feet or less. Or if no laptop available and you don't feel like moving the desktop around, pick up a long ethernet (CAT5e or better, probably better to go with a good quality 6 or 6E one) and temporarily run it down the hall directly to the main router, then to the main switch, then to where the coupler is, etc.

Daisy chaining switches should have very little impact so that shouldn't be the problem, unless that second switch has an issue or the cable between it and the first switch is no good.

But at these speeds, it is likely you are hitting hardware limitations. They sold you a 300MPH car but in reality there is nowhere you can drive that fast. 2G internet could support a business of 1000 people. So unless you're running a major hosting operation or porn site out of your house, kinda not needed, especially when most of the sites/servers you connect to will throttle you to much lower speeds anyway (except speed test sites).

I'd start with disabling firewall and antivirus for the test, I suspect you'll see significant increase in speed. Then you need to decide if it is worth the cost of a new computer just to see higher numbers.

Of course their speed test server may just be overloaded if there are lots of new customers coming on and testing. I see your first test has a slower upload than your second so it seems somewhat inconsistent. But if both PCs get the same result repeatedly over time then that probably isn't the problem.

If your router supports any sort of port counters to show errors that could help, but I suspect it doesn't. And the unmanaged switches won't either. That could help narrow it down if it is a wiring issue but the steps above will too.
 
Last edited:
move the second, slower result PC to the location of the first PC and use the same cable as the first. That will indicate if it is a cable, coupler, 2nd switch issue or the capabilities of the second PC.
 
Edited for clarity.
First, I'm assuming you're using all 2.5G connections (not 2G which would be bonded 1G in LAG which works differently)?

Try disabling firewall/antivirus on the PC while doing the speed test. If your results are significantly better, your PC just isn't up to the task of 2G throughput. I suspect that is the issue since the download is suffering more than upload (it scans the download much more thoroughly than upload). If that is the issue, you can either buy a new powerful PC, or just print out the speed test result, frame it, and be happy, since you likely do not need anywhere near 2G to a single PC.

If that doesn't help, watch the CPU in the PC during the speedtest, is it hitting 100% even with firewall disabled? If so, again, new PC or just be happy that you have more than enough speed, and your connection can support multiple computers all with a ton of speed at the same time.

You could try a better NIC in the PC, something from Intel (preferably server class) with its own dedicated processor on it. That may alleviate some of the CPU usage for the actual network traffic, but won't help with CPU usage caused by firewall/antivirus. You can see if there is a different firewall program that is more efficient but I doubt you'll find one that is twice as efficient.

If neither of the above, start checking wiring. Especially that RJ45 coupler, if it is some generic one, those are notoriously crappy. You can pick up good true Cat5e or better ones at home depot or from a good brand name off amazon. But also check your other wiring. If you have a laptop, test it at each point, before the coupler, after the coupler, at the final end near the PC, etc. Make sure the cable is not running along side power wires, or get a shielded cable if you have no choice. Distance needs to be about 330 feet or less. Or if no laptop available and you don't feel like moving the desktop around, pick up a long ethernet (CAT5e or better, probably better to go with a good quality 6 or 6E one) and temporarily run it down the hall directly to the main router, then to the main switch, then to where the coupler is, etc.

Daisy chaining switches should have very little impact so that shouldn't be the problem, unless that second switch has an issue or the cable between it and the first switch is no good.

But at these speeds, it is likely you are hitting hardware limitations. They sold you a 300MPH car but in reality there is nowhere you can drive that fast. 2G internet could support a business of 1000 people. So unless you're running a major hosting operation or porn site out of your house, kinda not needed, especially when most of the sites/servers you connect to will throttle you to much lower speeds anyway (except speed test sites).

I'd start with disabling firewall and antivirus for the test, I suspect you'll see significant increase in speed. Then you need to decide if it is worth the cost of a new computer just to see higher numbers.

Of course their speed test server may just be overloaded if there are lots of new customers coming on and testing. I see your first test has a slower upload than your second so it seems somewhat inconsistent. But if both PCs get the same result repeatedly over time then that probably isn't the problem.

If your router supports any sort of port counters to show errors that could help, but I suspect it doesn't. And the unmanaged switches won't either. That could help narrow it down if it is a wiring issue but the steps above will too.
Thank you for the comprehensive response! I will take the time to review it and determine how to attack the problem. However, since my initial post, I have done additional troubleshooting that may narrow down the problem. Note that the speed readings are from two components limited to 1 Gbps by the Ethernet ports (the PC with the upload/download speed conundrum has a 2.5 Gbps Internet port).

First, I use a mini-PC as a media PC connected to the same switch as the 2.5 Gbps PC. As you can see, the speed readings are roughly symmetrical and appropriate for the connection:
Screenshot 2023-04-06 075621.png

I then connected a laptop with a 1 Gbps Ethernet port to the cable leading to the 2.5 Gbps PC. Again, the readings are symmetrical and appropriate for the port speed:
Screenshot 2023-04-06 080124.png

These results suggest that the problem is with the 2.5 Gbps PC. It is also a mini-PC (ASUS PN63-S1) but significantly upgraded:

  • Intel Core i7-11370H Mobile Processor
  • Intel® Iris® Xe Graphics
  • G.Skill RipJaws Series 64GB (2 x 32GB) 260-Pin SO-DIMM PC4-25600 DDR4 3200 Dual Channel Memory
  • WD SN850x M.2 NVMe 1 TB Solid State Drive
  • WD SN770 NVMe 1 TB Solid State Drive
  • WD SA510 SATA 2.5-inch 1 TB Solid State Drive
  • Intel I225V 2.5G LAN Port
 
Try a 2.5 USB adapter. Assuming the other PC has a different nic inside. There have been issues with Intel 2.5 nics to the extent they're on rev 6 at this point.

If the USB nic does the same thing then it's something in running on the PC or some other issue in the hw or sw causing the bottleneck.
 
I have had great luck with the smart people on this forum before, and maybe you can help me figure this out. We just had a Lumos Networks 2 Gbps fiber Internet service installed, and the speeds are incredible - on one computer:

View attachment 49082

This desktop system is connected via Ethernet to a Zyxel 2 Gbps unmanaged switch, which is, in turn, connected to the 2 Gbps LAN port on a Netgear Orbi RBRE960 router. So far, so good.

However, the speeds on the other computer, which is also hardwired, are pretty different. While the upload speeds are impressive, the download speeds are considerably lower:

View attachment 49083

This computer is linked to another Zyxel 2 Gbps unmanaged switch connected to the same switch as the computer above. The cable that links the two switches has a coupler linking two Ethernet cables. What do you make of this? If the upload speeds are the same or better, shouldn't the download speeds also be?

The cable that links the two switches has a coupler linking two Ethernet cable
a switch ------ a cable ------ a coupler ------- a cable-----a switch
Is that right? Maybe a coupler issue? Otherwise your NIC issue or NIC option issue. You have to check it out one by one.
 
Thank you for the comprehensive response! I will take the time to review it and determine how to attack the problem. However, since my initial post, I have done additional troubleshooting that may narrow down the problem. Note that the speed readings are from two components limited to 1 Gbps by the Ethernet ports (the PC with the upload/download speed conundrum has a 2.5 Gbps Internet port).

First, I use a mini-PC as a media PC connected to the same switch as the 2.5 Gbps PC. As you can see, the speed readings are roughly symmetrical and appropriate for the connection:

I then connected a laptop with a 1 Gbps Ethernet port to the cable leading to the 2.5 Gbps PC. Again, the readings are symmetrical and appropriate for the port speed:

These results suggest that the problem is with the 2.5 Gbps PC. It is also a mini-PC (ASUS PN63-S1) but significantly upgraded:

  • Intel Core i7-11370H Mobile Processor
  • Intel® Iris® Xe Graphics
  • G.Skill RipJaws Series 64GB (2 x 32GB) 260-Pin SO-DIMM PC4-25600 DDR4 3200 Dual Channel Memory
  • WD SN850x M.2 NVMe 1 TB Solid State Drive
  • WD SN770 NVMe 1 TB Solid State Drive
  • WD SA510 SATA 2.5-inch 1 TB Solid State Drive
  • Intel I225V 2.5G LAN Port

That test doesn't tell you much of anything. If you hardcode the NIC in the "problem" PC to 1 gig you'll probably get about the same result.

I would start by just disabling AV and firewall on the PC for the speed test, see what that gets you. Will help determine if it is a PC CPU limitation or something else.
 
That brings to mind a quick test idea and boot to Linux from USB. It's quick and provides raw performance for testing things better than with windows.

Yeah figured they probably didn't have an ubuntu stick laying around. Sometimes the inbuilt drivers aren't optimized either. Disabling AV/Firewall should be a quick and easy test, even if it doesn't give the full 2G, if there is a significant improvement, points to CPU limitations.
 
Try a 2.5 USB adapter. Assuming the other PC has a different nic inside. There have been issues with Intel 2.5 nics to the extent they're on rev 6 at this point.

Thinking the same thing - NIC or driver issue...

One thing OP could perhaps try - boot up an Ubuntu 22.10 Live USB (don't install it, unless you want to), and see what happens - the linux drivers for the 225 are in decent shape - so if performance still doesn't improve... it's safe to consider the interface card...

Doesn't hurt to replace the cable, but even CAT5e should be ok for runs under 30 meters on NBase-T
 
The cable that links the two switches has a coupler linking two Ethernet cables. What do you make of this?

There's the suspect problem - replace that with a single cable run between the two switches - that "coupler" might work fine for gigabit, but NBase-T has more strict requirement, and you might be out of spec.
 
or use a another 2.5 Gbit switch ( overkill i know, but still...) in place of the coupler as quick test.
 
There's the suspect problem - replace that with a single cable run between the two switches - that "coupler" might work fine for gigabit, but NBase-T has more strict requirement, and you might be out of spec.

2.5G has the same requirements as 1G. Cat5e standard. But yes couplers are always suspect, suggested they replace it with a good branded one assuming it is a cheap/no name one which may very well be Cat3 or less. May have managed 1G (probably with some errors), but can't push more sensitive 2.5. The rare occasions I have to use one I use a punchdown keystone jack - punch one side and plug the other. Unless both cables are stranded, then a keystone coupler with RJ45 on both ends is the way to go.

My vote is that it is either the horsepower of the PC (AV/Firewall can't keep up) or the coupler, but guess we'll see if they ever reply.
 
or use a another 2.5 Gbit switch ( overkill i know, but still...) in place of the coupler as quick test.

Reminds me of my college, their network (10/100 hubs at the time) was deployed in the mid 90s, and by 2000 it was basically useless. I was interning as a network engineer at a big company, college IT called me in to look at it, they showed me a diagram, was just like um, there's your problem. The entire college (school and dorms) was like 3 huge daisy chains of large (probably 250 port or so) enterprise hubs. I believe the longest was about 15 in a row end to end, supporting dorms with about 4000 students. Glad I wasn't in the dorm on the end one. Managed to convince them to add a bunch of fiber home runs (they were already upgrading to 10/100 switches). I think it got down to 2 or 3 max on a daisy chain, they wouldn't pay for additional fiber to home run them all. At least at that point they went gig on the home runs/interlinks.

When I got there in '97 they were on a single T1 for the entire campus and everyone got public IPs (I ran a web server out of my dorm room, just picked a random static IP since not many had computers at that point). By 2000 they were using "business class" comcast cable modem at like 30 megs or something and had moved to NAT. Amazing how the internet use exploded in 2-3 years.
 
I appreciate all the suggestions! I'm trying several of them and will get back to you with what I learn!
 
OK. Test #1. My Bitdefender Total Security software doesn't have a disabling feature, so I uninstalled it completely using Revo Uninstaller Pro. I restarted the system and ran SpeedTest, with the following result:

1680840815559.png

This is the first time since my fiber service was installed that the download speed has passed the 1 Gb barrier. Subsequent tests yielded even better results:

1680841043045.png

1680841172050.png

The CPU peaked at 37% and averaged around 25% during the SpeedTest run, with five other apps open (Microsoft Edge, Microsoft Teams, Task Manager, Snipping Tool, and Windows Security). What do these test results tell you?

I ordered a 2.5 Gbps to USB 3.0 adapter that should arrive tomorrow, and I'll test it to compare it to the Intel I225-V.

I use a Y.D.F RJ45 in-Line Cat 7/Cat 6/Cat 5e Ethernet Cable Extender Coupler to link two Cat 6A Ethernet cables together. One of those cables runs from the first 2.5 Gbps switch, and the other cable goes from the coupler to the second 2.5 Gbps switch, which is where the test PC is also connected. The coupler rates 4.7 out of 5 stars with 1,622 reviews on Amazon, so I assume it is up to par.

That's enough for tonight - I look forward to your observations later!
 
OK. Test #1. My Bitdefender Total Security software doesn't have a disabling feature, so I uninstalled it completely using Revo Uninstaller Pro. I restarted the system and ran SpeedTest, with the following result:

View attachment 49126
This is the first time since my fiber service was installed that the download speed has passed the 1 Gb barrier. Subsequent tests yielded even better results:



The CPU peaked at 37% and averaged around 25% during the SpeedTest run, with five other apps open (Microsoft Edge, Microsoft Teams, Task Manager, Snipping Tool, and Windows Security). What do these test results tell you?

I ordered a 2.5 Gbps to USB 3.0 adapter that should arrive tomorrow, and I'll test it to compare it to the Intel I225-V.

I use a Y.D.F RJ45 in-Line Cat 7/Cat 6/Cat 5e Ethernet Cable Extender Coupler to link two Cat 6A Ethernet cables together. One of those cables runs from the first 2.5 Gbps switch, and the other cable goes from the coupler to the second 2.5 Gbps switch, which is where the test PC is also connected. The coupler rates 4.7 out of 5 stars with 1,622 reviews on Amazon, so I assume it is up to par.

That's enough for tonight - I look forward to your observations later!

When you uninstalled bitdefender, windows firewall/defender automatically re-enabled itself. So now you need to disable that and re-test.

Doesn't look like anyone has talked about using that coupler with 2.5G in the Q&A or reviews. You could unplug it, plug it inline with your other PC using a good patch cable, and see if it impacts that other PC. While it does look decently built from the pictures, who knows.

Try updating the NIC drivers to the latest from Intel if you haven't already done so too.

The fact that you can max out the upload appears to imply your cabling is good, however upload does use different wires than download, so not definite.

The speed test seems to vary a lot, are you sure this isn't just a matter of their server not being consistent? Have you tried other speed test servers in your area (of course they may not support those speeds consistently either) What speed to you get if you transfer a large file between the two PCs? You should be able to get around 300MBytes/sec, probably a bit less due to the overhead of windows file sharing and the way bytes are counted on a hard drive vs. on the network. That of course assumes both PCs can handle 300MB/sec, would need decent SSDs in both and disable any AV/Firewalls for the test.

You can also use software called Iperf to test between the two PCs which is not reliant on storage speed. However it is run from the command line so takes a bit of learning (not too bad).
 
Last edited:
I disabled Windows Defender and retested, with the following results:

1680873677293.png


1680873705582.png


The Intel NIC has the most current drivers (Version 2.1.3.3, 2/9/2023). I will see if I can get enough access to test the cable leading into the coupler with my laptop. In the meantime, I took to heart your statement about not finding references to 2.5 Gbps in the Q&A and reviews for the coupler. I purchased another one to try it out. The UGREEN RJ45 Coupler 10Gbps Ethernet Extender touts its support for 10 Gbps speed and up to 328 feet of Ethernet cabling. My cable runs are not nearly that long! The worst-case scenario is that it slows my connection, and I send it back and put the old one back in place.

When you mention the server not being consistent, that could explain the variation between tests. Lumos is a brand-new service in our area, and they may just be getting their lines and equipment calibrated.

I will report back soon with additional information!
 
speed tests across the internet have too many variables in traffic congestion and routing to be accurate. They are a rough indication.
If you want to figure out if any of your LAN is having issues, you need to use something like iperf between two devices. Even then, one of the devices serving or receiving may be the limitation.
 
I disabled Windows Defender and retested, with the following results:

View attachment 49136

View attachment 49137

The Intel NIC has the most current drivers (Version 2.1.3.3, 2/9/2023). I will see if I can get enough access to test the cable leading into the coupler with my laptop. In the meantime, I took to heart your statement about not finding references to 2.5 Gbps in the Q&A and reviews for the coupler. I purchased another one to try it out. The UGREEN RJ45 Coupler 10Gbps Ethernet Extender touts its support for 10 Gbps speed and up to 328 feet of Ethernet cabling. My cable runs are not nearly that long! The worst-case scenario is that it slows my connection, and I send it back and put the old one back in place.

When you mention the server not being consistent, that could explain the variation between tests. Lumos is a brand-new service in our area, and they may just be getting their lines and equipment calibrated.

I will report back soon with additional information!

I wonder where your browser is storing the temporary speed test file. If it is on the SATA 2.5 drive it is possible you're just hitting the limit of that (or maybe your storage subsystem has some sort of bottleneck). The fact that upload is full speed and download is 75% still points to a bottleneck somewhere. Less likely it is cabling (not impossible though, something could be affecting the RX pairs and not TX).

I don't think ugreen is going to be any different than what you have now. Probably same adapter made in the same place.

Maybe the orbi can't do 2 gigs, download would be passing through the stateful firewall and thus more limited than upload
 

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