Ya, testing, but.
lol, SG 18s(18000ms, yes) ping is blocking me to test US server.

You means 100% becoming 99% packet loss?

Unsure, I never faced Google SMS TAC, as they just push to any of my Android phone...
But I can confirm, Shopee, Taobao/Aliyun, WeChat can only reaching via phone call TAC, with regular Malaysian phone number, lol.

----
For the internet, 21s to SG and MY.

Roger, and sadly I don't found tools for testing cloudflare over UDP, are there any third party speedtest website that host at Cloud Flare?
Just like rethink DNS.

This post has been edited by BenYeeHua: Aug 26 2025, 10:21 PM

Just saying, if you wanna speed test for DNS, it is best time with worst connection now.
https://dnsspeedtest.online/

But, please be aware that, it is over HTTPS, which having different IP address over TLS for some.
Like CloudFlare, it is test over 2606:4700::6810:f9f9

This post has been edited by BenYeeHua: Aug 26 2025, 10:32 PM

Thanks, confirmed H/3.
Much better, no packet loss, no slow start, but observed the hang issues, as high as 9s, even timeout.

colo=KUL
sliver=none
http=http/3
https://h3.speed.cloudflare.com/cdn-cgi/trace



So it is just like old Aliyun, the server keep hanging with newer http.



---
After 30s timeout, it fallback to h2.

colo=KUL
sliver=none
http=http/2

This post has been edited by BenYeeHua: Aug 26 2025, 10:59 PM

This is why your guys reporting lagging over website, but speedtest nothing.
ISP just hate UDP.




Ya, like me, keep getting those DDoS/burst force usage IPv4, yuck.

Easy, buy old CCTV that no server relay, and required public IP.

Seem like cloudflare issues, when it is sliver 005-c-cloudflare or 001, it works.

colo=KUL
sliver=005-c-cloudflare
http=http/3



When sliver none or 010 or 001-tier3 or 005-tier3, it break.

colo=KUL
sliver=none
http=http/3


---
But yes, for Firefox, network.http.http3.enable.

This seem more unlikely http3 issues, but certain silver server having issues, disable h3 might not helping, it need to be fix by CloudFlare.

As above, certain sliver got issues, when I connect to JHB, I also got the similar hang issues.



This post has been edited by BenYeeHua: Aug 26 2025, 11:25 PM

You means the slliver thing?
Reopen browser, lol.

I just open incognito windows then open the h3 link, then reload again to kick it into h3, and perform testing.

https://h3.speed.cloudflare.com/cdn-cgi/trace

1.1.1.1 app, Zerotrust = KUL, wireguard = SG, MASQUW = JHB.

Yes, but the sliver thing proof that, the routing inside CloudFlare is having issues.

And when it is none, sometimes it is working fine with 80ms 50Mbps.

Ya, more likely his IPv6 is broken, sometimes certain area will get v6 burst while v4 still running.

GN630V is not that bad as host/single, just WPA3 is broken sometimes.(30 days uptime until my parent reboot it because CCTV hang, lol.)
and yes, client got strange high latency when I using it to convert lan to WiFi(big WiFi adapter la, lol)

Secondly, there is no disconnect/connect button, making hard to get new IP.

This post has been edited by BenYeeHua: Aug 27 2025, 12:01 AM

Look like the sliver thing back to normal as most user sleep, retest next night then.

Ya, those IP address special for WARP only, so you know la.
Few years ago can get HKG and MRS etc, which also solved strange Asia network issues, sadly no such bug anymore, lol.
(Like Huawei can't backup at all Asia, but fine with MRS etc.)

Maybe Netflix think your IP address jumping strange, can be shared account, so temporary ban.

lol, I found the WiFi reboot by saving WiFi, never think about WAN.

Yup, and VPN burn battery faster, still, it depend on how easy you can change the IP address result la.
Feel lazy, so just go with WARP, lol.

This post has been edited by BenYeeHua: Aug 27 2025, 12:41 AM

yes, mostly it is other website got issues(even midnight slow), just use simple warp solved it.
Just CloudFlare going to KUL these few weeks, so the issues shown up, but as it is CloudFlare loadbalance side, should be easy fix than TM side, lol.

Ya, the problem is always the receive side.
My PC need to buy a 2m extension + 8dbi(should be fake, just 3/5dbi) antenna, then can reach the same speed as my mate 20 or poco F5 at same location.

Should be the MTK RZ616/MT7922 louzy la, like now it is cheating showing me 2402Mbps, because it is >75% signal, it only show the real speed when it drop to 75%...
There is also 1 driver version, which it broken the IPv6, I only found out after I reboot the router.



----
Anyways, fixing the WiFi as bottleneck is important, else your latency will sky high to 100ms.
Just like that night, my fiber also sky high to 100ms, I wonder did 100ms is the answer to every bottleneck.

And yes, if your PC's CPU running games, when you speed test over just nicely 200Mbps WiFi speed, it will drop to <100Mbps, then rise to 50ms latency.
By rising the PCI-E MSI-X to 4 threads and high priority, it some kind of improve bit more la.

Well, MTK one don't have such configuration, lol.

Just having quick check on NDIS version, this MTK even using 6.70 while the Realtek LAN driver is using latest 6.89, which is released for 24H2, lol.
So I guess MTK don't really put his mind on optimizing the performance of this chipset then, and my best choice is pull a cable if upgraded to 500M, or Intel BE200 etc.

For the WiFi >75%, just ignore the receive speed as it is power saving now.


So as you can see, when it is 76%, you will get fake speed, so I don't think MTK really like to doing good thing on this WiFi Chipset...

What, me?

There is a lot of 175.x.x.x, some already get corrupted by those DDoS/burst force kids, so had to jump ship liao.
Else, my family keep blame me why take 30 mins to reconnect the internet, I had such bad luck keep kicking/sharing the CGNAT IP with those script kiddie.

There is some early period of CGNAT issues, like during opening lowyat, it will reset connection as out of port etc, guess they solved it by reducing the sharing of same IP address, or because it is IPv6 generation.

This post has been edited by BenYeeHua: Aug 28 2025, 10:26 AM