Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
bc6d1946-0d82-41d7-a08b-b5467b9873b4.jpeg


3rd world country FTW. $25 per month.
 
  • Like
  • Wow
Reactions: dmccombs and EugW
Don't believe Ookla. They make tests with dedicated servers, not used in production, which are designed to fool the users. https://packetstats.com/ is ways more reliable to measure your rel network performance with independent servers.
am I missing something or is this simply a ping test tool? Ookla tests bandwidth and throughput compared to this.
 
am I missing something or is this simply a ping test tool? Ookla tests bandwidth and throughput compared to this.
This is a repeated test with independent worldwide Internet servers using random packets. Not ping.

Ookla tests bandwidth and throughput toward botched servers from your ISP designed to give you perfect immediate results for the very last mile. But internet is more than that. You need DNS servers and a lot of relays on the way before reaching your final destination. (Traceroute) There are the real performance killers located and Ookla does not test that.
 
"Internet speed" is also affected by the servers providing the data as well as chokepoints between the last mile and the servers. It's a bit like a car good for 200 mph in an area where the roads are only good for -say- 50 mph.

The ISP that I used between 1995 to 2003 had something like a 6Mbps connection with the internet backbone when it started offering DSL connectivity. It would only take a few DSL users maxing out on DSL bandwidth to saturate the ISP's internet connection. My current ISP has been mum of what their connection with the internet can handle or how much content distribution is done localy.

Another amusing development is the likes of Roku's "Live TV" where programs are distributed as real-time broadcasts. This presumably results in a very large reduction in bandwidth required and likely reduces the royalties paid to the content owners with respect to on-demand streaming services.
 
I pay for a gig from Verizon fios. My last speed test wireless was 591 download and 295 upload
 
I still remember those days when using a dial up modem with a limited amount of hours per month. My parents (I was a teen back then) would ask me to disconnect since they wanted to make or were waiting for a phone call !

I used to play counter strike (started at v1.1) and when an update was released I'd make sure I bought a video game magazine which offered along with the usual demos the necessary update which would have been cumbersome to download.
 
I love this thread especially when I see a self-claimed 3rd world country getting 2GBits down!

Here I am in the middle of one of the nice bits of London in the UK and I can get 80Mbits down and 20Mbits up. The reason being that there is a bloody tree in the way of where they would have to run the fibre. Three engineer call outs so far and they worked out they'd have to dig up the entire side road and run the thing from the other side. But of course they can't get a works order from the local council due to some "unspecified problems". They have however drilled a hole through the front of the house and left it. Has been 2 years now.

My neighbour, in the same physical building has 1gbit up and down. But it's a crap ISP so I don't want it.
 
This is a repeated test with independent worldwide Internet servers using random packets. Not ping.

Ookla tests bandwidth and throughput toward botched servers from your ISP designed to give you perfect immediate results for the very last mile. But internet is more than that. You need DNS servers and a lot of relays on the way before reaching your final destination. (Traceroute) There are the real performance killers located and Ookla does not test that.
Yet the webpage describes itself as

This simple ping stability testing tool...​

 
Don't believe Ookla. They make tests with dedicated servers, not used in production, which are designed to fool the users. https://packetstats.com/ is ways more reliable to measure your rel network performance with independent servers.
This is just a ping & packet test, not so much as a speed (megabits per second) tool. Ookla isn't a great test, but it'll show max potential throughput in MBps. Google's speed test often gives me a different result than Ookla, so likely the truth lies somewhere between.
 
Ookla isn't a great test, but it'll show max potential throughput in MBps.
But it just tests the last mile. Crappy ISP servers behind are today the major hurdle on Internet communication.
Even Google is a well known test, so the cheating servers frequently mock it up.
Did you wonder why your Internet feels sluggish while you get fantastic Ookla results?
 
But it just tests the last mile. Crappy ISP servers behind are today the major hurdle on Internet communication.
Even Google is a well known test, so the cheating servers frequently mock it up.
Did you wonder why your Internet feels sluggish while you get fantastic Ookla results?
Everything you said is true. But there’s no way to test speeds from every site you might hit, so if last-mile testing is the best that can be done, so be it.
But no, my “internet” never feels sluggish. I’m well aware that my last-mile service is faster than the hosting servers I’m connecting to.
 
Last edited:
  • Like
Reactions: IvyKing
I still remember those days when using a dial up modem with a limited amount of hours per month. My parents (I was a teen back then) would ask me to disconnect since they wanted to make or were waiting for a phone call !

I used to play counter strike (started at v1.1) and when an update was released I'd make sure I bought a video game magazine which offered along with the usual demos the necessary update which would have been cumbersome to download.
Those were the days. I started around beta 3 when it was still very clearly a mod for Half Life and stuck around until 1.3 or so. I'd occasionally play with a few friends over a 33.6 kbps modem connection, but it was expensive and we mostly ended up taking turns playing against bots (that weren't part of the game and had to be downloaded separately). I stopped play CS when Steam came around and wouldn't accept my (legitimate) Half Life CD key, telling me that another user had already used the key. I wasn't until Valve released Half Life and CS for Mac OS X in 2013 that I gave it another try.
 
  • Like
Reactions: Lyrca
Everything you said is true. But there’s no way to test speeds from every site you might hit, so if last-mile testing is the best that can be done, so be it.
But no, my “internet” never feels sluggish. I’m well aware that my last-mile service is faster than the hosting servers I’m connecting to.
The very problem is not really the one distant server you are just using. But more or less every website is crippled with hundreds of "business partner's servers" that must be DNS'ed before being reached and the result is frequently a pity.
Today's internet is badly ill.
 
The very problem is not really the one distant server you are just using. But more or less every website is crippled with hundreds of "business partner's servers" that must be DNS'ed before being reached and the result is frequently a pity.
Today's internet is badly ill.

I run AdGuard home on a RaspberryPi to handle and filter DNS requests on my local network. It definitely speeds up browsing.
 
  • Like
Reactions: splifingate
I run AdGuard home on a RaspberryPi to handle and filter DNS requests on my local network. It definitely speeds up browsing.

Yes.

I run two, independent pi-hole instances as my DNS resolvers (one on a rPi5, and one in Prox) on the Daily.

Use a 1Gbps/symm. fibre pipe, so non-filtering doesn't skew my experience; but it's much, much better when I filter-out all the cruft.
 
  • Like
Reactions: Lyrca
Don't believe Ookla. They make tests with dedicated servers, not used in production, which are designed to fool the users. https://packetstats.com/ is ways more reliable to measure your rel network performance with independent servers.
Sure, your internet speed is determined by the server to which you are connecting, and everything between. If you're connecting to a slow server, or your data is routed over a slower path, you're going to have a slow connection.

But the purpose of ISP performance tests like Ookla isn't to test the speed of the internet as a whole, it's only to test the speed of the portion for which your ISP is responsible.

So let me ask you this: Does Ookla omit any portion for which the ISP is responsible? If it doesn't, then it's a fair test. If it does, then it's not.
 
A lot depends on whether the Ookla servers are directly connected to the ISP or are connected via the general internet. The chokepoint for internet traffic could be the ISP's connection with the internet as a whole. My first ISP (dial-up, then DSL) had, IIRC, about 6Mbps connectivity to be shared by all of its customers.
 
Does Ookla omit any portion for which the ISP is responsible?
Ookla tests the connection to their local specialized servers, optimised to deliver good results.
They test only the last mile, not the server chain behind, which is used operatively for the regular internet.
Most of the lag you may experience in daily internet usage, comes from there; when these servers are overloaded.
 
Ookla tests the connection to their local specialized servers, optimised to deliver good results.
They test only the last mile, not the server chain behind, which is used operatively for the regular internet.
Most of the lag you may experience in daily internet usage, comes from there; when these servers are overloaded.
Right, but that doesn't directly answer my question:

Is there any part of the connection that the ISP is responsible for, but that is omitted from the Ookla test?

If it captures everything the ISP is responsible for, and only omits parts that are outside the ISP's control, then it is a perfectly fair test of the performance of your ISP, and is thus doing exactly what you'd want it to, which is to provide information on that specific component.

To make the value dependent on stuff outside the ISP's control is to introduce conflating variables. If you really want to know why you have slow internet, you need to isolate each component. Only then do you get clarity on whether the issue is your ISP, or other parts of the connection between you and the server you are accessing.
 
Last edited:
If it captures everything the ISP is responsible for, and only omits parts that are outside the ISP's control,
That's obviously not possible.
But you have really bad ISP's using completely overloaded DNS servers and overloaded trunks.
That is the part for which they are responsible and do not deliver to promise.
Their Ookla testes shine and that's enough for them.
And these bad ISPs are not the smallest ones.
This is revealed by every traceroute. You just need to dig further and search who is the owner of the laggy servers.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.