Question Download speed is really low after 1 hour of using new internet ?

Kvahuest01

Prominent
Aug 20, 2020
69
2
535
0
So my download speed tanked down all of a sudden.

we got new internet today, stated as 800 Mbps, did a test and got that, wonderful.

Started to download a game that was 60gbs, download speed was perfect on that, and got what I tested prior.

left it to download then my download speed just dropped to around 80 Mbps, I'm connected via ethernet, and waited till people got off so it was just me using and I still have the same issue.

My phone is getting a higher download speed than my pc, i do get the 800 mbps every now and then but i only get it for a short ammount of time before i lose it again.

tried to prioritize my PC but my router doesn't have QoS

Heres a picture of all the speed tests -

View: https://imgur.com/a/iDwahGt
 
That is a very interesting speed test. When you say it dropped to 80mbps is very different the low 90mbps rates you show.

Numbers in the low 90's generally mean some ethernet connection is only running at 100mbps. I would first check your PC but if there are switches inbetween it could be others.

It is almost always a bad cable. There is a massive amount of fake cables being sold now days, mostly because copper metal is so expensive. You want a cable that is pure copper (no CCA) and has wire size 22-24 AWG (none of that flat of thin cable).
Still even quality cables can go bad. It takes just 1 wire to get slightly loose inside the plug. It can work fine and as it warms up the metal moves a bit it no longer does.

Luckly this is a very cheap fix. Be sure to look for details on the metal and wire size. Repuitble vendors know about the fakes and make it a point to show that theirs is not, and why it might cost 20-30% more.
 
Reactions: CompuGuy71

Kvahuest01

Prominent
Aug 20, 2020
69
2
535
0
That is a very interesting speed test. When you say it dropped to 80mbps is very different the low 90mbps rates you show.

Numbers in the low 90's generally mean some ethernet connection is only running at 100mbps. I would first check your PC but if there are switches inbetween it could be others.

It is almost always a bad cable. There is a massive amount of fake cables being sold now days, mostly because copper metal is so expensive. You want a cable that is pure copper (no CCA) and has wire size 22-24 AWG (none of that flat of thin cable).
Still even quality cables can go bad. It takes just 1 wire to get slightly loose inside the plug. It can work fine and as it warms up the metal moves a bit it no longer does.

Luckly this is a very cheap fix. Be sure to look for details on the metal and wire size. Repuitble vendors know about the fakes and make it a point to show that theirs is not, and why it might cost 20-30% more.
thats what it could be then, did buy this cable years ago and its a flat one, could you recommend any? i don't know what to look out for.
 
This is hard thing from what I can tell most cable on amazon and ebay are fakes, even cable sold directly by amazon......but they don't allow fake purses or watches :)

I know many people like the cable sold by monoprice but even they sell the stuff with the thin wires so you have to be careful.

This is one example that seems ok
https://www.amazon.com/Monoprice-Cat6-Ethernet-Patch-Cable/dp/B002RBECAE/ref=psdc_306629011_t1_B01BGV2KSQ

You do not need better than cat5e but if cat6 is cheaper that is fine even though it will not perform any differently. Key here is to look for ads like that make it point to tell you the wire is pure copper and the wire size. Most are AWG 23 but this one is 24 which is fine. The ones that are bad have numbers in the awg 30-33
 

Kvahuest01

Prominent
Aug 20, 2020
69
2
535
0
This is hard thing from what I can tell most cable on amazon and ebay are fakes, even cable sold directly by amazon......but they don't allow fake purses or watches :)

I know many people like the cable sold by monoprice but even they sell the stuff with the thin wires so you have to be careful.

This is one example that seems ok
https://www.amazon.com/Monoprice-Cat6-Ethernet-Patch-Cable/dp/B002RBECAE/ref=psdc_306629011_t1_B01BGV2KSQ

You do not need better than cat5e but if cat6 is cheaper that is fine even though it will not perform any differently. Key here is to look for ads like that make it point to tell you the wire is pure copper and the wire size. Most are AWG 23 but this one is 24 which is fine. The ones that are bad have numbers in the awg 30-33
i do need a longer cable aswell so like 15-20m, i dont mind paying like £10+ for a good cable, from the reviews on that it seems on the bad side.
 
Can't say for sure you have to be careful about reviews just like good ones most people giving the reviews are idiots. These are the same idiots that give good reviews to the flat cable because they can hide it better.

It was purely a example of what to look for. I bought a lot of cable a few years ago so it is hard to recommend a particular company now. Many times when you buy just 1 cable because of the shipping costs it will about the same price as for cable you can get from your local home improvement store.
You still need to read the packaging or look at what it printed on the side of the cable. Most cable will say on the side if it CU copper or CCA (fake aluminum stuff) and the AWG size of the wire. Local vendors tend to not sell as much fake stuff since they take returns but you have to be careful, I saw walmart selling that flat cable.
 

Kvahuest01

Prominent
Aug 20, 2020
69
2
535
0
Can't say for sure you have to be careful about reviews just like good ones most people giving the reviews are idiots. These are the same idiots that give good reviews to the flat cable because they can hide it better.

It was purely a example of what to look for. I bought a lot of cable a few years ago so it is hard to recommend a particular company now. Many times when you buy just 1 cable because of the shipping costs it will about the same price as for cable you can get from your local home improvement store.
You still need to read the packaging or look at what it printed on the side of the cable. Most cable will say on the side if it CU copper or CCA (fake aluminum stuff) and the AWG size of the wire. Local vendors tend to not sell as much fake stuff since they take returns but you have to be careful, I saw walmart selling that flat cable.
ill have to have a look arround my area, if not ill try and get a good one off amazon or something
 
Jul 23, 2022
19
4
15
0
I would recommend doing a speed test on a system with FreeBSD or NetBSD installed. These two systems are known for their high IOPS on the hard disk, and for their network stack which is faster than Linux/windows/macOS. So you may well get higher results with these systems. In some specific cases, FreeBSD quadruples the network performance of Windows 10/11.
 

Ralston18

Titan
Moderator
I would recommend doing a speed test on a system with FreeBSD or NetBSD installed. These two systems are known for their high IOPS on the hard disk, and for their network stack which is faster than Linux/windows/macOS. So you may well get higher results with these systems. In some specific cases, FreeBSD quadruples the network performance of Windows 10/11.
References?

The posted problem is the drop in performance after one hour. So the system/network does work and I do not understand how you can conclude that FreeBSD or NetBSD will resolve that problem.

= = = =

@Kvahuest01

I recommend using Task Manager and Resource Monitor (use both but only one at a time) to observe system performance. Objective being to discover what changes when the performance drops.

You can also use Reliability History and Even t Viewer to look for some error code, warning, or even an informational event that occurs around that 1 hour mark. Reliability History uses a time line format that could prove very revealing.

Lastly, look in Task Scheduler. There could be some scheduled trigger involved. May be some update, backup, or other action launching - may or may not be game related.
 
Jul 23, 2022
19
4
15
0
References?
The operating system's network stack can have an impact on your computer's download performance.
https://openbenchmarking.org/embed.php?i=1612062-TA-NETWORKLA06&sha=89445e1&p=2
https://openbenchmarking.org/embed.php?i=1612062-TA-NETWORKLA06&sha=982006d&p=2

Windows is also probably the system with the most bugs and security problems of all systems that exist. So there are many factors that can have an impact. Also the server. There are almost zero servers that will forward your data at 800 mbps, that's just too high a speed that most servers can't deliver to thousands of people at the same time.

You can simply test it like this:

1. Close all apps and all tabs in your browser, then do the following test in Chrome: https://www.nperf.com/en/

2. Install FreeBSD on fast hardware that you don't currently use and connect it to the same cable (ethernet) connection, then do the same test (in Chromium instead of Chrome).

Do you see similar results? Then there is no problem with the performance of your Windows system.
But if there is a big difference in the values, then there is definitely something wrong with your windows PC.
 

Ralston18

Titan
Moderator
None of the three links include a Windows OS comparison.

Were all three tests run with the default installation values? In other words, could the end results have been biased by end user configuration options?

Noted that the target IP 198.162.1.111 is a private IP address. The original post was (provided that I correctly understood the post) involved servers on the internet.

All tests must target the same IP address. For example, if the linked tests had targeted Braintree then the OS comparisons with or without Windows OS included) then the tests would be fairer.

Also this:

"2. Install FreeBSD on fast hardware that you don't currently use and connect it to the same cable (ethernet) connection, then do the same test (in Chromium instead of Chrome).

Such a "fast hardware" test (in my mind) immediately invalidates the test. Likewise you can cannot change the browser.

The tests and all tests must be done in exactly the same physical hardware, network, and software environment with only one change (OS for example) being made at a time.

All things must be consistent except for what is being tested. One variable only.
 
Jul 23, 2022
19
4
15
0
You're right, if you want to do a completely equivalent test, you're going to have to use the same hardware and browser. Chromium and Chrome usually deliver similar performance, so that shouldn't have much of an impact.

Normally the speed of the provider will be the biggest bottleneck in this nperf test. That's why these criteria probably shouldn't be exactly the same and I knew that when I gave my previous answer, because all the things you list are usually not the main bottleneck.

windows server usually performs similar to the slower Linux distros:
https://openbenchmarking.org/embed.php?i=1901259-SP-NETWORK1072&sha=71133d6&p=2
https://openbenchmarking.org/embed.php?i=1803140-FO-NETWORKS460&sha=997248d&p=2

But what I would like to know, what is the result of your windows pc in Nperf? Can you take a screenshot and upload it?
 
Jul 23, 2022
19
4
15
0
Here you can see the performance I get on a more than 10 years old (budget) dual core and 4GB RAM @ 1600Mhz: https://i.ibb.co/nPHsrwD/Screenshot-2021-03-14-16-06-05.png

This was almost a year ago that I had taken the test, with networking has gotten a bit faster now.

This confirms my first post, that you should do a test with FreeBSD or NetBSD and see if you see a difference in Nperf. Normally it should be a similar result with your windows PC, unless windows PC has major performance issues.

I personally know countless IT professionals, but I don't know anyone with the same general IT knowledge, so you can trust my posts..
 

fball922

Distinguished
Feb 23, 2008
174
23
18,695
3
It definitely looks like you are connected at 100Base-T. What do you see under your connection properties? You would see 100/100 if the connection to your switch/router is degraded.
 

Kvahuest01

Prominent
Aug 20, 2020
69
2
535
0
I would recommend doing a speed test on a system with FreeBSD or NetBSD installed. These two systems are known for their high IOPS on the hard disk, and for their network stack which is faster than Linux/windows/macOS. So you may well get higher results with these systems. In some specific cases, FreeBSD quadruples the network performance of Windows 10/11.
i just want to sort my download speed out.. not install other O/S....
 
Reactions: fball922

Ralston18

Titan
Moderator
Key is to discover what is happening or stops happening when performance drops.

Take a closer look at what you system is doing per my/Post #9.

Start by observing with the system just sitting and stable. Leave the observation window open and drag aside so you can watch.

Then begin the download. Hopefully the download will begin at or near the expected speed.

And you will see what changes when the speed drops.

It make take a bit of time and effort get the process down and spot a potential problem.

Just be as methodical and as consistent as possible. Take your time no need to rush it.

Another tool is Process Explorer (Microsoft, free).

FYI:

https://docs.microsoft.com/en-us/sysinternals/downloads/process-explorer

If you can eliminate your computer as being the source of the problem then that is one less potential culprit.
 

Kvahuest01

Prominent
Aug 20, 2020
69
2
535
0
Key is to discover what is happening or stops happening when performance drops.

Take a closer look at what you system is doing per my/Post #9.

Start by observing with the system just sitting and stable. Leave the observation window open and drag aside so you can watch.

Then begin the download. Hopefully the download will begin at or near the expected speed.

And you will see what changes when the speed drops.

It make take a bit of time and effort get the process down and spot a potential problem.

Just be as methodical and as consistent as possible. Take your time no need to rush it.

Another tool is Process Explorer (Microsoft, free).

FYI:

https://docs.microsoft.com/en-us/sysinternals/downloads/process-explorer

If you can eliminate your computer as being the source of the problem then that is one less potential culprit.
trouble is even on first bootup some times its below 100, seems to be luck of the draw to get full speed or not, just bought a new eth cable since the one i use now was old and one of those flat ones.

the new one hopefully fixes it, did say it was full copper but theres so many fakes out there now.
 
Jul 23, 2022
19
4
15
0
i just want to sort my download speed out.. not install other O/S....
I only recommended it as a test. I thought I also literally said that you shouldn't replace your current OS.

What I find especially strange is that you say that your smartphone downloads faster than your PC. Could you please run the Nperf test on your desktop and your smartphone and send the detailed results of both tests?
 
I only recommended it as a test. I thought I also literally said that you shouldn't replace your current OS.
You are new here and you have to remember most people asking questions have very little technical knowledge. You many times have to point them to youtube videos to do even very very basic things. In this case the OP knows some of the stuff.
I agree that sometimes testing with linux is useful. It mostly proves the problem is software in windows "someplace". The normal method I recommend it to run from a USB image that does not actually install and run completely from the USB stick. There are many of these testing images available. It of course can be limiting but network problem and testing tend to not be impacted much by the slower USB.

Still I suspect a bad cable someplace. He many times can test at full speed but then it drops to speeds around say 94mbits. This is very different than if it dropped to say 50mbps or 120mbps or pretty much any other number. When you see speeds in the low 90mbps range this tends to be a connection that is running at 100mbps but it does not account for the overhead in the packet the mac addresses take up. It can be a number of things that drop a port to 100mbps but it tends to be cables most the time. Finding it when it is random can be hard sometimes.

If it had been pretty much any other speed I would be more suspect of things like the browser or maybe QoS software or something but when you see this numbers in this range that is pretty much a normal 100mbps connection but why it is running at 100mbps is the key question.
 
Reactions: ComplexNumbers

ASK THE COMMUNITY