Real International speed tests

[xiaofan]

Testing from Singapore Free Amazon VPS (one year free tier).

ubuntu@aws:~$ speedtest

Speedtest by Ookla

Server: MyRepublic - Singapore (id = 5935)
ISP: Amazon.com
Latency: 1.65 ms (0.05 ms jitter)
Download: 965.87 Mbps (data used: 483.9 MB)
Upload: 968.23 Mbps (data used: 484.2 MB)
Packet Loss: 0.0%
Result URL: https://www.speedtest.net/result/c/17ac7def-7c2a-46be-82dd-b6c0da1c660b

ubuntu@aws:~$ speedtest -s 24333

Speedtest by Ookla

Server: Rakuten Mobile , Inc - Tokyo (id = 24333)
ISP: Amazon.com
Latency: 85.50 ms (0.21 ms jitter)
Download: 439.81 Mbps (data used: 608.4 MB)
Upload: 218.90 Mbps (data used: 232.5 MB)
Packet Loss: 0.0%
Result URL: https://www.speedtest.net/result/c/371bc8a5-4a48-4c30-a98b-6c153370e836

ubuntu@aws:~$ speedtest -s 30932

Speedtest by Ookla

Server: Optus - Melbourne (id = 30932)
ISP: Amazon.com
Latency: 223.30 ms (0.05 ms jitter)
Download: 205.97 Mbps (data used: 270.5 MB)
Upload: 175.56 Mbps (data used: 259.8 MB)
Packet Loss: 0.0%
Result URL: https://www.speedtest.net/result/c/b0749739-cc5f-45ea-906a-49776807454d

ubuntu@aws:~$ speedtest -s 1783

Speedtest by Ookla

Server: Comcast - San Francisco, CA (id = 1783)
ISP: Amazon.com
Latency: 170.72 ms (0.06 ms jitter)
Download: 79.84 Mbps (data used: 109.4 MB)
Upload: 181.28 Mbps (data used: 283.2 MB)
Packet Loss: Not available.
Result URL: https://www.speedtest.net/result/c/73c03da0-7a09-4d04-b2f3-c0da095d8819

ubuntu@aws:~$ speedtest -s 16089

Speedtest by Ookla

Server: University of Texas, Austin - Austin, TX (id = 16089)
ISP: Amazon.com
Latency: 243.02 ms (4.42 ms jitter)
Download: 7.79 Mbps (data used: 9.9 MB)
Upload: 63.10 Mbps (data used: 94.5 MB)
Packet Loss: 0.0%
Result URL: https://www.speedtest.net/result/c/9b04ee89-a8ab-463b-b829-4bd1fc08386c

 

[firesong]

Was trawling the internet for speedtest servers.

speedtest -s 13623
speedtest -s 3147
speedtest -s 11557
speedtest -s 30932
speedtest -s 1783
speedtest -s 24333
speedtest -s 17402
speedtest -s 8648

Haven't tested it yet. For some reason, apt update fails on my VIM3L - getting Error message "File has unexpected size} from ookla.bintray.com.

Here's the server list repo I pulled some of the data from. Couldn't get China Mobile, so I went with "China Broadband Comms" (8648).
https://williamyaps.github.io/wlmjavascript/servercli.html

Was hoping to put it up on a cron job and map via Grafana, but using the official speedtest client rather than the OSS speedtest-cli tool... Oh well.

 

[cscs3]

Here is my test. My place happen to have starhub and m1. Both 1G plan.
09OCT test
M1 with Linksys EA9500 down=534 up=446
SH with linksys wrt3200 down=355 up=514
09NOV test(switch router around)
M1 with linksys wrt3200 down=447 up=495
SH with linksys EA9500 down=232 up=485

Test is conduct in such a way M1 to M1 and SH to SH.

My take.
1. Router has some factor (but time/day has greater impact). Linksys clearly is a wrong choice as ONT to device I can get 900+ on ethernet, so confirm telco did not short change.
2. M1 still consistently faster on upload section (which is quite useless for general user), but latency is higher then SH.
3. All test on 5G channel.

Added.
Forgot to mention this.
Device is mobile phone, connecting to both router at 866Mbps speed.
Will find sometime to teat with PC with LAN connection.

Edit. Just notice in my table, upload and download is reverse. Corrected.

 

[firesong]

Here is my test. My place happen to have starhub and m1. Both 1G plan.
09OCT test
M1 with Linksys EA9500 up=534 down=446
SH with linksys wrt3200 up=355 down=514
09NOV test(switch router around)
M1 with linksys wrt3200 up=447 down=495
SH with linksys EA9500 up=232 down=485

Test is conduct in such a way M1 to M1 and SH to SH.

My take.
1. Router has some factor (but time/day has greater impact). Linksys clearly is a wrong choice as ONT to device I can get 900+ on ethernet, so confirm telco did not short change.
2. M1 still consistently faster on upload section (which is quite useless for general user), but latency is higher then SH.
3. All test on 5G channel.

Added.
Forgot to mention this.
Device is mobile phone, connecting to both router at 866Mbps speed.
Will find sometime to teat with PC with LAN connection.

Thanks for your contribution. Do note that the intention was for international traffic tests too. A sample of international servers are provided.

Local speeds should always be near telco-rated numbers, else that's already indicative of a problem somewhere. Overseas speeds, the real traffic, might not be the case. This bottleneck is encountered when there is no local mirror for the content you want, so it is sent directly from overseas. For example, new OS updates, or new shows, or very old data that is no longer mirrored locally, or whatever low priority content not found locally like Linux ISOs.

In which case, in most cases wireless speeds are fine, you just need to indicate it as such - since we take that anything 500-600Mbps means you've essentially saturated the theoretical 2x2 867Mbps 802.11ac that is typical for most current machines and phones. We can assume also that you would be on 1Gbps plans, but indicate plan speed upfront to conform.

 

[xiaofan]

Was trawling the internet for speedtest servers.

speedtest -s 13623
speedtest -s 3147
speedtest -s 11557
speedtest -s 30932
speedtest -s 1783
speedtest -s 24333
speedtest -s 17402
speedtest -s 8648

Haven't tested it yet. For some reason, apt update fails on my VIM3L - getting Error message "File has unexpected size} from ookla.bintray.com.

Here's the server list repo I pulled some of the data from. Couldn't get China Mobile, so I went with "China Broadband Comms" (8648).
https://williamyaps.github.io/wlmjavascript/servercli.html

Was hoping to put it up on a cron job and map via Grafana, but using the official speedtest client rather than the OSS speedtest-cli tool... Oh well.

Yeah, unfortunately the OSS speedtest-cli does not work really well in this case, it can only test the servers near to the locations. Only the official clients can do the tests against test servers at farther away locations.

As for the test server IDs, I have also come across the list you mentioned above. But they contain many outdated info. The way I find out is to use VPN to fool the official client or official website to list the test server IDs around the VPN server locations.

Website
https://c.speedtest.net/speedtest-servers-static.php

The above website will list different severs around the VPN server location.

China Mobile HK SpeedTest Server ID is 32155.

 

[cscs3]

Thanks for your contribution. Do note that the intention was for international traffic tests too. A sample of international servers are provided.

Local speeds should always be near telco-rated numbers, else that's already indicative of a problem somewhere. Overseas speeds, the real traffic, might not be the case. This bottleneck is encountered when there is no local mirror for the content you want, so it is sent directly from overseas. For example, new OS updates, or new shows, or very old data that is no longer mirrored locally, or whatever low priority content not found locally like Linux ISOs.

In which case, in most cases wireless speeds are fine, you just need to indicate it as such - since we take that anything 500-600Mbps means you've essentially saturated the theoretical 2x2 867Mbps 802.11ac that is typical for most current machines and phones. We can assume also that you would be on 1Gbps plans, but indicate plan speed upfront to conform.

Just notice I mix up download and uplpad column. Corrected.
Testing with Kdatacenter (korea)
SH - down=413 up=306
M1 - down=383 up=503

 

[miloaisdino]

Yeah, unfortunately the OSS speedtest-cli does not work really well in this case, it can only test the servers near to the locations. Only the official clients can do the tests against test servers at farther away locations.

As for the test server IDs, I have also come across the list you mentioned above. But they contain many outdated info. The way I find out is to use VPN to fool the official client or official website to list the test server IDs around the VPN server locations.

Website
https://c.speedtest.net/speedtest-servers-static.php

The above website will list different severs around the VPN server location.

China Mobile HK SpeedTest Server ID is 32155.

try --server for the open source client instead of option s...

 

[xiaofan]

try --server for the open source client instead of option s...

It does not work for further away servers. I have tried it.

 

[irbadboyz]

I think we have to understand one crucial aspect, Singapore is literally a physical internet hub. Sea cables between countries connect through us, since it is easier to lay internet cables in the sea as opposed to land.

Most of the lines you see coming from countries overseas, which connect through Singapore are charged by Singapore. Which is why if you go to NZ, they have fiber but 200Mbps is considered fast over there. And it costs a lot too.

Whereas here in Singapore internet is dirt cheap because many of the underground cable are being leased by local companies like Opennet for example.

Google underground sea cable and you will see what I mean.





IT company Singapore

 

[miloaisdino]

I think we have to understand one crucial aspect, Singapore is literally a physical internet hub. Sea cables between countries connect through us, since it is easier to lay internet cables in the sea as opposed to land.

Most of the lines you see coming from countries overseas, which connect through Singapore are charged by Singapore. Which is why if you go to NZ, they have fiber but 200Mbps is considered fast over there. And it costs a lot too.

Whereas here in Singapore internet is dirt cheap because many of the underground cable are being leased by local companies like Opennet for example.

Google underground sea cable and you will see what I mean.





IT company Singapore

yes, but the question is "do we get to use these cables or not"? many of them are for international/corporate customers only or just too expensive for residential use, so in reality residential users can only make use of a small subset of them depending what isp you are on..

 

[Mach3.2]

I think we have to understand one crucial aspect, Singapore is literally a physical internet hub. Sea cables between countries connect through us, since it is easier to lay internet cables in the sea as opposed to land.

Most of the lines you see coming from countries overseas, which connect through Singapore are charged by Singapore. Which is why if you go to NZ, they have fiber but 200Mbps is considered fast over there. And it costs a lot too.

Whereas here in Singapore internet is dirt cheap because many of the underground cable are being leased by local companies like Opennet for example.

Google underground sea cable and you will see what I mean.





IT company Singapore

Just as an example, SingTel has been shown time and time again to have poor routes, there are plenty of posts covering that in this sub-forum. Not all ISPs are created equal, you be your own judge.

 

[dudley02]

M1 500 Mbps, on wifi (but wifi speeds are not limiting)
12 Nov 2020, 9pm

Comcast, San Francisco

Rakuten, Tokyo

Bell Mobility, Vancouver, BC

Jump Networks Ltd, London

China Mobile, Hong Kong

 

[dudley02]

(continued)

AIS Fibre, Bangkok

U Mobile, Kuala Lumpur

Optus, Melbourne

kdatacentre, Seoul

 

[firesong]

Thanks. Looks like the $29.90 M1 500Mbps plan is more than capable enough for most users.

Over a 2y contract, it would appear that paying $240 extra yields no major benefits.

 

[dudley02]

Yes you're right. Just renewed my M1 500 Mbps for another 2 years since there weren't any attractive offers for 11.11. Considered, but eventually decided against upgrading to 1 Gbps (I'm on wifi most of the time and 500 Mbps is fast enough for my use case).

 

[miloaisdino]

M1's latency is so much better than singtels.. quality over quantity!

 

[alvino]

hey all,
hope my ST tests help to contribute

1. ST
2. 1gbps
3. 13/11/2020 8.30am

ST>ONT>xiaomi ax3600>cat6>desktop

Singtel, SG

Comcast, San fran

Rakuten Japan

Bell Mobility Vancouver

Jump Network, London

China Mobile, HK

AIS, BKK

 

[alvino]

U Mobile, KL

Optus, Melbourne

kdatacentre, Seoul

on sidenote my bittorrent on ubuntu64 20.04 iso hit peak of 60+MB/s. If i spam 4 different variant of iso i could hit peak over 100MB/s

 

[xiaofan]

Bumping up this thread and hopefully more people will contribute.

 

[firesong]

Modifying this project a little,

Does anyone have any recommendations for smokeping servers across this set of countries?

I am not sure if we can ping speedtest servers directly, or if we should.