What's new

Expected differences in Speedtest results?

  • 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!

ADFHogan

Regular Contributor
Nominally, my ISP provides me with a 1000/50 connection, but this isn't guaranteed. The uplink, no matter how I test it, is within 4Mbps to this 50.
However, their download is only suggested to be an average of 600Mbps, and the current plans they're offering on their site suggest closer to 270Mbps.
That aside, trying to get a handle on what I'm functionally getting these days has gotten a little more complex.

I have a Samknows box attached to my network.. and that shows a monthly daily average of 897/46 when I connect to their portal.
If I run speedtest.net in my browser, in my router's UI (through Asus UI), or via CLI/script in spdmerlin I see something closer to 300-400Mbps.
If I run speedtest CLI on my linux desktop, I see 480-928 which seems to match with the Samknows more (targeting the same servers UI did).

What's more likely the realistic figure? Is it expected that speed tests from router would differ by a deficit of couple to a few hundred Mbps?

I do get about 300Mbps down if I target a US server (I'm in AU) from CLI version on desktop.

RT-AC5300 showing its age?
Running 386.11 with just routing no WiFi as I have UI switches + APs (running controller on another linux box, no cloudkey/dream machine, the Asus is the router/FW).
 
On Windows your antivirus will slow things down. Your browser may also have URL/security checking running.

Packet inspection , DNS over TLS can be an issue.

Antivirus needs to be off for testing (Kaspersky for example destroys speed tests ) , use an icongnito browser session

Asus router , are you running AiProtection?

Try https://www.nperf.com/en/ for speed tests .

In my experience Linux will always show a more reliable result.
 
Is this wifi or cable, did you test the speedtest app on your router aswell. With my ac86u iam also unable to see the speeds go to 1gb / 1gb. Although it is more then able to get it. Is your hw flow enabled?
 
Desktop machine is Linux, not Windows. Wired with ethernet, no
Is this wifi or cable, did you test the speedtest app on your router aswell. With my ac86u iam also unable to see the speeds go to 1gb / 1gb. Although it is more then able to get it. Is your hw flow enabled?
Via wired ethernet.
HW flow? Maybe? Is that the NAT Acceleration option (which is on)? Or do you mean QoS which is off right now.
 
Trying to run a speedtest of the RT-AC5300 itself (GUI, CLI or addon) will give you a false result because the router's CPU will be limiting you to about 300Mbps. The only valid test would be from an ethernet connected PC to a reasonably close server.
 
On Windows your antivirus will slow things down. Your browser may also have URL/security checking running.

Packet inspection , DNS over TLS can be an issue.

Antivirus needs to be off for testing (Kaspersky for example destroys speed tests ) , use an icongnito browser session

Asus router , are you running AiProtection?

Try https://www.nperf.com/en/ for speed tests .

In my experience Linux will always show a more reliable result.

DOT won't change your speed test result. The DNS lookup happens before the test starts.
 
Desktop machine is Linux, not Windows. Wired with ethernet, no

Via wired ethernet.
HW flow? Maybe? Is that the NAT Acceleration option (which is on)? Or do you mean QoS which is off right now.

Wired desktop with av/firewall disabled, a decent network card, and fairly modern CPU is the only way to test. Testing from the router is no good.

You are on a shared medium (cable) which is pushed to its limits already, if there are a lot of people in your area on the same node, variation is expected.
 
Wired desktop with av/firewall disabled, a decent network card, and fairly modern CPU is the only way to test. Testing from the router is no good.

You are on a shared medium (cable) which is pushed to its limits already, if there are a lot of people in your area on the same node, variation is expected.
So from what you say, the results I get from my router are meaningless?

Speed Test (from router).png
 
I know that, but I was asking in the general sense.
 
I know that, but I was asking in the general sense.
Accuracy would vary by router model and the speed of the person's internet connection, so it's not possible to make a general statement. If you want to discuss your particular router model I suggest you start a new thread.
 
No, it's OK - I don't need any help at the moment. I only asked the question out of curiosity.
 
No, it's OK - I don't need any help at the moment. I only asked the question out of curiosity.

I would never take the router result as accurate. The test you show may have been, then when the router is under load from something else you'll get a totally different result. Always use a wired PC with a CPU and NIC capable of your internet speed as it will be far more reliable and consistent.
 
No, it's OK - I don't need any help at the moment. I only asked the question out of curiosity.
Nah, on my ax86u pro I have more then a enough horsepower to speedtest on the router. It maxes out my connection perfectly, also the wifi is more then cable to max out the 1gb fibre. On my ac86u I cannot, it does work the speed but only on a PC on a wire. Different gen, out dated, but perfect for me as a mesh unit.
 
Ahh.. ~300Mbps expected for "on device" on RT-AC5300.. got it.. thanks!

As for my internet, it's NBN FttN (Passive optical fibre network managed by government last mile carrier, NBNCo, with my ISP connecting to the other end at one of the NBN points of interconnect)... so it's not HFC. I have an NBN optical network terminal device into which I plug my RT-AC5300 directly. Speed level provided by NBNCo on my plan is 1000/50, but that's only my NTD to the PoI. ISP will then allocate a varying amount of backhaul, and speed varies depending on if one's accessing a peered CDN or an international site with no peering. Apparently the fibre coming to my ONT is shared between me and 31 others due to NBN PON configuration.

The speed tests I've been doing have been a mix of on-router (browser CLI, GUI), and on my Linux desktop (browser, CLI). Turns out I get the fastest results when I use the Linux CLI client on my desktop. Desktop is using wired ethernet through a gigabit switch to the RT-AC5300.

So basically, putting this all together, using the CLI client on Linux desktop, wired into network as opposed to wireless, with minimal other apps or network activity should give me the most reliable result?

With that in mind, besides speedtest.net, what do folks use?

PS. Am curious - someone asked previously about if I had HW acceleration enabled. Which portion of the router config is that? Is it the NAT acceleration, or is it not applicable to the RT-AC5300?
 
Ahh.. ~300Mbps expected for "on device" on RT-AC5300.. got it.. thanks!

As for my internet, it's NBN FttN (Passive optical fibre network managed by government last mile carrier, NBNCo, with my ISP connecting to the other end at one of the NBN points of interconnect)... so it's not HFC. I have an NBN optical network terminal device into which I plug my RT-AC5300 directly. Speed level provided by NBNCo on my plan is 1000/50, but that's only my NTD to the PoI. ISP will then allocate a varying amount of backhaul, and speed varies depending on if one's accessing a peered CDN or an international site with no peering. Apparently the fibre coming to my ONT is shared between me and 31 others due to NBN PON configuration.

The speed tests I've been doing have been a mix of on-router (browser CLI, GUI), and on my Linux desktop (browser, CLI). Turns out I get the fastest results when I use the Linux CLI client on my desktop. Desktop is using wired ethernet through a gigabit switch to the RT-AC5300.

So basically, putting this all together, using the CLI client on Linux desktop, wired into network as opposed to wireless, with minimal other apps or network activity should give me the most reliable result?

With that in mind, besides speedtest.net, what do folks use?

PS. Am curious - someone asked previously about if I had HW acceleration enabled. Which portion of the router config is that? Is it the NAT acceleration, or is it not applicable to the RT-AC5300?

Speedtest.net is the best as not only does it have a good selection of high speed servers, but it runs multiple parallel streams which is often necessary to totally saturate your connection.

The GUI does take more CPU but most modern PCs can handle that fine. Depends on the PC you're running it on, if one CPU core maxes out during the GUI (Browser) test then the CLI may be better.

Always test from a wired PC, don't bother with the built in one on the router.

As far as other apps running again it depends on your PC and CPU, but generally speaking, yes, minimal apps running, disable AV/Firewall (temporarily) etc.

There are two flavors of hardware acceleration. You can turn it on or off but whether you get one or both depends on the features you have enabled. CTF is usually enabled, but if you enable trend micro, QOS, parental controls, traffic statistics, etc you will lose some of the other hardware acceleration benefits. For example my old AC1900 I can get just shy of 500 with traffic stats and trend micro enabled, and probably double that or more with them disabled.
 
Trying to run a speedtest of the RT-AC5300 itself (GUI, CLI or addon) will give you a false result because the router's CPU will be limiting you to about 300Mbps. The only valid test would be from an ethernet connected PC to a reasonably close server.
Well… my rt-ax88u pro are very accurate with speed test. Though I have not found its limit. I’ve measured up to 2 Gbps

I’ve confirmed this with my isp senior tech who came to my house with their own Speedtest hardware.
 
It depends highly/entirely on what options/features and scripts you have enabled and running, as to how accurate it is.
 
Well… my rt-ax88u pro are very accurate with speed test. Though I have not found its limit. I’ve measured up to 2 Gbps

I’ve confirmed this with my isp senior tech who came to my house with their own Speedtest hardware.

We're talking about the RT-AC5300.
Accuracy would vary by router model and the speed of the person's internet connection, so it's not possible to make a general statement. If you want to discuss your particular router model I suggest you start a new thread.
 
You are testing the speed between your home and a server someplace on the internet. Your ISP may be providing the contracted bandwidth yet someplace upstream of your ISP may be congested and beyond there control. When you get a speed test that is what you are paying for that's great. When you don't, there are too many possible reasons to have meaning. I hope you enjoy your nice fast 600 Mb or faster link.
 

Similar threads

Latest threads

Support SNBForums w/ Amazon

If you'd like to support SNBForums, just use this link and buy anything on Amazon. Thanks!

Sign Up For SNBForums Daily Digest

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