Achieving the last 10%

[#1]

Hi folks,

So, my 500-550/75 service went live yesterday and, generally, I'm really pleased with it. There is one minor frustration (and I've researched several other threads with a similar theme), and that is that I never quite see it achieve the full expected 500-550mbps download bandwidth unless I use the thinkbroadband speed test. For many years, I've always used the Ookla site because I've found their servers to be very effective at maxing out the available bandwidth and yielding the best results, but this new connection won't ever go any higher than about 470mbps via speedtest.net. All my hard-wired hosts are performing similarly.  I've also tried running iperf to various public servers directly from the WAN interface of my firewall (it's a J3160 based mini PC running Pfsense) and that's the same story - I can squeeze a tiny bit more (480mbps) because it's straight off the WAN, but I never get any higher than that.

Now, here's the evidence to the contrary (a TBB speed test):

And here's the typical result via speedtest.net or any other type of server I can think to try:

https://www.speedtest.net/result/17594961266 

Finally, a typical test using iperf direct from the edge device:

The CPU on the pfsense only reaches about 60% during the above, and it routes traffic between NICs & VLANs at full gigabit no problem at all.

Do you think there could be some sort of throttling going on that TBB is able to slightly defeat where others can't? I can't think of any other explanation and it's bugging me, because I really don't think it's down to equipment or test method in my case. I did try to check with a direct PPPoE connection from my laptop using a USB NIC, but that didn't work out and all I could get with that approach was about 150mbps. Very strange, because the same laptop & nic combination will  do 450mbps no problem when it's connected conventionally via the switch & pfsense!

Would be pleased to hear your thoughts, keeping in mind that this is a brand new connection with only 30h uptime so far. I would have given it a few more days first but my understanding from reading other posts is that there is no 'training' period. It makes little difference and I'm probably fussing other nothing, but can't shake the feeling something isn't quite 100%.

Thanks,

Stuart

[#2]

There's no throttling, I can get > 900mbps on my connection on speedtest.net.

[#3]

Thanks @dvorak, I didn't really expect there to be throttling, just wanted to understand the shortfall vs expected bandwidth. I was playing with iperf again earlier this evening (running direct from pfsense again) and I discovered that if I run a test using 2x parallel streams rather than just a single one, the aggregate throughput can indeed reach 500mbps and beyond. So, despite my initial doubts, it seems that there might be an equipment issue concerning the way my pfsense firewall handles PPPoE. If I decide to pursue this further then the next logical step would be to unwrap the PN issued domestic router and compare, but I think I'm satisfied for now that I'm getting what I pay for.

[#4]

Mine's the other way round on my 300 line. TBB returns a bad line, TestmyNet returns better. Single thread tests seem to be terrible with this router/ line, multi thread, as expected, return almost full line speed. I find my phone connects fastest of all my devices, even wireless compared to wired.

[#5]

Well, speedtest is doing the business now and without having made any changes on my side:

https://www.speedtest.net/result/17603722342 

Not sure why I couldn't manage that on day one if there's no settling period but, never mind, it's all good! 😁