[SOLVED] Old router vs Speedtest.net

Mar 6, 2020
6
0
10
Hello everyone

I am currently using an aging Linksys E3200 router to connect to the internet (fiber). My ISP has steadily been increasing our speed, and we are now up to 250/250 Mbps, and if I connect my PC directly to the plug in the wall, this is what I get (and then some). I've tested using a range of sites (speedcheck.org, Fast.com, Speedof.Me, nperf.com, Speedtest.net), and also straight up download, eg. Steam.

When connected through the router, my speed drops to ~110/110 (170 with CTF disabled). This is "fine", either the router just can't handle the speed, or it is bugged. It's probably time to upgrade anyway.

However, there is one curious thing: Even when using the router, Speedtest.net shows a speed of 250/250+. All other sites, and normal download, rates the speed at ~110.

Anyone have an idea as to how this is possible?

Thanks in advance for any input!
 
Solution
I think you are right, but an oddly specific bug.

I've tried your idea with older firmware. I started this endeavor on 1.0.04, upgraded to 1.0.05, and now down to 1.0.03. I am unable to finder older versions. The results are the same on all available versions.

For good measure, I also tried dslreports.com. It's another tick in the ~110/170-box.

I guess I'll have to give it up, and start looking for a replacement. I have an R6220 set up as an AP right now, maybe I'll try that, and then look for a new router/AP for the wireless part.

Thanks for your replies, folks!
Yes, but I've seen these things before. Back in the day when I had 3 cable modems at home and an rv016 to multi-wan all of them, I started having problems with a...
Mar 6, 2020
6
0
10
Welcome to the forums, newcomer!

Are you on the latest firmware for that router? According to the press info, here, your router is capable of higher bandwidth. Might want to see if the issue persists after a firmware update(if pending) and then connecting off the Ethernet port.

Thank you for the quick reply! Yes, I've updated to the most recent firmware (which isn't very recent at all - 2014), which made no difference.
I'm not sure about the last part - "connecting off the Ethernet port"?
 

kanewolf

Titan
Moderator
Thank you for the quick reply! Yes, I've updated to the most recent firmware (which isn't very recent at all - 2014), which made no difference.
I'm not sure about the last part - "connecting off the Ethernet port"?
In addition to latest firmware, when questioning router performance, it is always best to start with a factory reset of the router. Then set just the admin password, SSID/password, and any ISP required info. That way you ensure you didn't accidentally change something that impacts performance.
 
  • Like
Reactions: StewCannon
Mar 6, 2020
6
0
10
In addition to latest firmware, when questioning router performance, it is always best to start with a factory reset of the router. Then set just the admin password, SSID/password, and any ISP required info. That way you ensure you didn't accidentally change something that impacts performance.

I did do a factory reset after the firmware update. And I am running my tests with only my PC connected to the router, and I have tried all 4 ports.

Interesting fact (at least to me):
With CTF enabled, Speedtest shows 250+ Mbps, all other tests show ~110 Mbps
With CTF disabled, all tests, including Speedtest, show ~170 Mbps

And that is actually the main mystery to me - how can Speedtest - and ONLY Speedtest - get a result that matches the expected speed?
 

kanewolf

Titan
Moderator
I did do a factory reset after the firmware update. And I am running my tests with only my PC connected to the router, and I have tried all 4 ports.

Interesting fact (at least to me):
With CTF enabled, Speedtest shows 250+ Mbps, all other tests show ~110 Mbps
With CTF disabled, all tests, including Speedtest, show ~170 Mbps

And that is actually the main mystery to me - how can Speedtest - and ONLY Speedtest - get a result that matches the expected speed?
That is not unexpected. CTF (or hardware offload) allows the packets to traverse the router with minimal CPU interaction. Turning it off, requires the CPU to touch every packet. That is generally slower.

What are you using for "all other tests show" ??? This could be a limitation of your ISP and not under your control.
 
  • Like
Reactions: StewCannon
Mar 6, 2020
6
0
10
That is not unexpected. CTF (or hardware offload) allows the packets to traverse the router with minimal CPU interaction. Turning it off, requires the CPU to touch every packet. That is generally slower.

What are you using for "all other tests show" ??? This could be a limitation of your ISP and not under your control.

I agree, enabling CTF should speed things up. But only Speedtest.net seems to benefit.

CTF on
Microsoft Network Speed Test, bredbandskollen.se, Fast.com, Speedof.Me, nperf.com, speedcheck.org and simply downloading via Steam and others:
~110 Mbps

Speedtest.net:
270-320 Mbps

CTF off
bredbandskollen.se, Fast.com, etc. etc.:
160-170 Mbps

Speedtest.net:
160-170 Mbps

So, turning CTF on speeds up Speedtest.net, but ONLY Speedtest.net. I get the best, real life speed by turning it off. And then I'm still missing about 100 Mpbs.

So the mystery still remains: Why am I only getting the 250+ speeds with Speedtest.net (CTF on), and literally no other scenario?
 
Last edited:
This is very interesting. I would probably explain it as a bug in the firmware, and you can confirm this by loading different versions of firmware on the unit and repeating the test. I'm suspecting you'll run across different results on a much earlier firmware that was before the bug was introduced.

I've actually always had good luck with dslreports.com/speedtest. It is the only 100% accurate test I've found when testing with any isp in the USA.
 
  • Like
Reactions: StewCannon
Mar 6, 2020
6
0
10
This is very interesting. I would probably explain it as a bug in the firmware, and you can confirm this by loading different versions of firmware on the unit and repeating the test. I'm suspecting you'll run across different results on a much earlier firmware that was before the bug was introduced.

I've actually always had good luck with dslreports.com/speedtest. It is the only 100% accurate test I've found when testing with any isp in the USA.
I think you are right, but an oddly specific bug.

I've tried your idea with older firmware. I started this endeavor on 1.0.04, upgraded to 1.0.05, and now down to 1.0.03. I am unable to finder older versions. The results are the same on all available versions.

For good measure, I also tried dslreports.com. It's another tick in the ~110/170-box.

I guess I'll have to give it up, and start looking for a replacement. I have an R6220 set up as an AP right now, maybe I'll try that, and then look for a new router/AP for the wireless part.

Thanks for your replies, folks!
 
I think you are right, but an oddly specific bug.

I've tried your idea with older firmware. I started this endeavor on 1.0.04, upgraded to 1.0.05, and now down to 1.0.03. I am unable to finder older versions. The results are the same on all available versions.

For good measure, I also tried dslreports.com. It's another tick in the ~110/170-box.

I guess I'll have to give it up, and start looking for a replacement. I have an R6220 set up as an AP right now, maybe I'll try that, and then look for a new router/AP for the wireless part.

Thanks for your replies, folks!
Yes, but I've seen these things before. Back in the day when I had 3 cable modems at home and an rv016 to multi-wan all of them, I started having problems with a wan connection being removed from the multi-wan. I found out that the isp changed their carrier end equipment and somehow that new equipment was sending 100packets/sec of ack/nak packets to my rv016 which took it as an attack and would remove the wan connection. After months of sending logs to both Cisco and the isp, I got fed up with it and put 2 other routers in front of my rv016 on 2 of the 3 wans and the problems disappeared since the packet floods were being filters by the other routers. Prior to finding out this solution, I actually had purchased another rv016 figuring that 4 years 24x7 and it had given up the ghost. I ended up with a spare rv016, and I still have both units today somewhere. Bottom line was that there was nothing wrong on my end at all--it was an isp issue.

Sometimes it can be a firmware issue or odd interaction between firmware and the isp. The Netgear FVS318N was notorious for this, having issues with WAN to LAN throughput because of this, and typically never reaching the 250Mbps that it was rated for. Well, after a lot of experimenting someone found a firmware version that worked, and today that firmware version is hitting 400Mbps--much higher than what the unit was even spec'd for. And even stranger--on another site where I have this unit installed and was preparing to swap it out for one with the 'right' firmware so it could exceed the 50Mbps cap, I did a speedtest first and noticed it hits the full 100Mbps without an issue whereas it would only hit 50Mbps on the older 65Mbps plan. Again, nothing changed on my end, but changed on the isp side.

Bottom line is that there may be absolutely nothing wrong with your router except that it has some sort of funk in relation to your isp. Of course that means that you need to get a replacement, but could quite possibly also mean that you have a perfectly mint condition unit that will work great for someone else on another isp. Who knows.

Interesting to see dslreports test end up in the same bin. I know their tests use a lot more threading than other tests which helps on the accuracy.

Since you're about to give up, I had bit of an oddball idea to try--boot up a linux live cd and see what results you get that way--just for kicks. I don't expect it to be different, but if it is it may lead us down another path of tests. :)
 
Solution
Mar 6, 2020
6
0
10
Since you're about to give up, I had bit of an oddball idea to try--boot up a linux live cd and see what results you get that way--just for kicks. I don't expect it to be different, but if it is it may lead us down another path of tests. :)

If I can find a USB key somewhere, I'll give a try :)

Thanks for taking the time to help me!