Sponsored

Less than ideal Zen FTTP speeds

Msh

Pro Member
Evening all,

I am after a bit of advice on what steps I should take, hopefully, you can help!

Today, I had an OR FTTP installation for Zen's 900 Mbps package. The installation went well (a bit of an issue with PPPoE not being set up on the router by default but no worries) and the internet is now live.

Latencies seem great and most importantly, stable. However, throughput is far less than ideal. It's ranging from 400-600 in most cases, the odd 650 and 350, but ultimately not a consistent or speedy experience. These tests are from speedtest.net using Zen's own server.

Initially, I had my own router running (RB5009) with correctly tweaked MTU, after some time, I tried the provided router with double NAT, the provided router on its own, and finally the ONT directly attached to the laptop with a PPPoE client. None of which made the slightest difference.

I have spoken with Zen on live chat, as well as a phone call, and finally a call with their fault team. An Openreach engineer has been scheduled for tomorrow. I also learnt from the fault team guy that my exchange is "on-net" (ie, not on BTW backhaul). The thing that has worried me about all three of these encounters with Zen staff is they have all quoted a 450 Mbps guarantee.

The engineer will arrive tomorrow afternoon, I am not overly optimistic, however. What are my options here? I am not okay with this speed and will cancel if it's not resolved in the contract grace period (even BT guarantee 700). Is there anything I should try? Is there anything I can do to assist?

I realise this is a pretty vague/open-ended topic, but I do want to make this work out with Zen so any recommendations would be very greatly appreciated.

Thanks!
Msh
 

Pheasant

ULTIMATE Member
What machine/age/spec/build are you testing with? Using the speedtest.net app/CLI or browser baaed test?

Have you replicated the results by testing with another machine (wired)?

Guarantee threshold is normal.
 

Msh

Pro Member
Modern Ryzen Zen 2 (4800U) and 3 (5600x, 5900x) computers, all hardwired. App, CLI, and web. The devices and tests all exceed gbit on Virgin Media (except laptop which is limited to GbE).

I respect there needs to be some minimum, however this minimum seems a tad too low for me to accept (and some tests are falling short of this even). BT offer 700 guarantee so maybe a move there’s the right thing to do? Just wonder if it’s futile if it’s not an issue with Zen themselves. If this isn’t a Zen issue, what could it even be?

I noticed this evening I am getting ~800, I’ll spend a bit more time data gathering.
 

NTTP

Casual Member
I am hopefully several weeks away from ordering FTTC and I was considering Zen over BT but this post is a bit concerning as I planned for the 900mb speed.
Have you tried https://speed.cloudflare.com or https://fast.com to see how they compare?
I tend to distrust speedtest.net as it was poor when I was on Hyperoptic even though Hyperoptic did get me full 1GB when I actually did a large file test with JDownloader (real multi-thread connection with 16 transfers)

JDownloader may be worth a go and using some of the large file test sites to see how it goes with concurrent connections/chunks set to 16.



 

Pheasant

ULTIMATE Member
Just wonder if it’s futile if it’s not an issue with Zen themselves. If this isn’t a Zen issue, what could it even be?

I noticed this evening I am getting ~800, I’ll spend a bit more time data gathering.
There’s a decent probability that Zen are actively traffic shaping your connection or there is a backhaul limitation. I would suspect this before any sort of notions around a squeeze on PON capacity.

Openreach can now do PON only speed testing on behalf of their CP customers to prove there isn’t excess congestion /contention on the PON. There was a trial announced here.

In any case continue to monitor at various times over the next few days.
 

Msh

Pro Member
Have you tried https://speed.cloudflare.com or https://fast.com to see how they compare?
I tend to distrust speedtest.net as it was poor when I was on Hyperoptic even though Hyperoptic did get me full 1GB when I actually did a large file test with JDownloader (real multi-thread connection with 16 transfers)
That's interesting, as I tend to distrust fast (gives me faster than theoretical speeds on my VM link).

That said, it's a bit all over the place, I am getting far closer to gigabit speeds on Cloudflare's speed test with 869, 850, and 864. Fast is 740, 680, and 650. Speed test, right this moment, is giving 883, 433, and 633. As for chunked downloads, I just opted to try on Steam which maxed out at 643 Mbps.

As much as I loathe Virgin Media, I am able to get 1100 fairly consistently everywhere I test.

What's a bit interesting, is that I left some speedtest-cli tests running through the night (using SpeedFlux). They are all far closer to gigabit than I have been able to get with web or desktop apps.

Gjpy7VA.png

I also noticed there was some loss throughout the night on the same device doing the tests.
V4b1vIX.png

I still don't quite know how significant this data is and I have never heard of speedtest-cli performing better than the other apps.

There’s a decent probability that Zen are actively traffic shaping your connection or there is a backhaul limitation. I would suspect this before any sort of notions around a squeeze on PON capacity.
I am not questioning the PON capacity (at least not right now). I don't really believe enough of my neighbours would have ordered as quickly as myself.

Openreach can now do PON only speed testing on behalf of their CP customers to prove there isn’t excess congestion /contention on the PON. There was a trial announced here.
Would be good to see if they can do that.

My current thinking is bailing on Zen :(. If what Zen said on the phone was true about not using BTW backhaul, I guess I should be safe trying BT?
 

baby_frogmella

Top Member
Its worth remembering that not all speed testers can cope with Gigabit connections. On speedtest.net, only the Xilo server gives me 800+ Mbps consistently on my BT Business 900 service, others are anywhere between 300-700 Mbps.
 

Msh

Pro Member
Its worth remembering that not all speed testers can cope with Gigabit connections. On speedtest.net, only the Xilo server gives me 800+ Mbps consistently on my BT Business 900 service, others are anywhere between 300-700 Mbps.
As I said, I can consistently get greater than Gigabit elsewhere so I think there's a bit more to this story than the other end of the speedtest.

That said, I have written a script to constantly download gigabyte test files from a variety of places and I have noticed the throughput on my router getting into the 900s

KMoLO64.png


I'll let the engineer come this afternoon still but this is surely Zen backhaul at this point. I can get gigabit-ish speeds from the ONT.
 

Pheasant

ULTIMATE Member
That's interesting, as I tend to distrust fast (gives me faster than theoretical speeds on my VM link).

That said, it's a bit all over the place, I am getting far closer to gigabit speeds on Cloudflare's speed test with 869, 850, and 864. Fast is 740, 680, and 650. Speed test, right this moment, is giving 883, 433, and 633. As for chunked downloads, I just opted to try on Steam which maxed out at 643 Mbps.

As much as I loathe Virgin Media, I am able to get 1100 fairly consistently everywhere I test.

What's a bit interesting, is that I left some speedtest-cli tests running through the night (using SpeedFlux). They are all far closer to gigabit than I have been able to get with web or desktop apps.

Gjpy7VA.png

I also noticed there was some loss throughout the night on the same device doing the tests.
V4b1vIX.png

I still don't quite know how significant this data is and I have never heard of speedtest-cli performing better than the other apps.


I am not questioning the PON capacity (at least not right now). I don't really believe enough of my neighbours would have ordered as quickly as myself.


Would be good to see if they can do that.

My current thinking is bailing on Zen :(. If what Zen said on the phone was true about not using BTW backhaul, I guess I should be safe trying BT?
Interesting to see the "step change" in DL speed on that SpeedFlux graphic at around 1:15am. Almost like a profiling change...

Overall I wouldn't say the connection is performing especially badly ~ I think you would have a hard time arguing that the connection is not performing as expected / as a basis for rejection.
 

Msh

Pro Member
Interesting to see the "step change" in DL speed on that SpeedFlux graphic at around 1:15am. Almost like a profiling change...
Yes, it's quite interesting, I wonder if stuff is still going on. It's also a bit strange as these results don't seem to reflect reality, or at least, I can't reproduce that kinda speed on desktop speed tests.

Overall I wouldn't say the connection is performing especially badly ~ I think you would have a hard time arguing that the connection is not performing as expected / as a basis for rejection.
Unfortunately, I am inclined to agree. I do expect better, however.

I wonder if I am being too picky, but at the same time, if I don't get anywhere near the advertised speeds even on Steam, what's the point in this package? If it's possible to get much closer to GbE speeds then I will try my best to get that.

Right at this moment in time, I feel like it's justified to point the finger at Zen.
 

kommando828

ULTIMATE Member
Well someone with ADSL on Zen got upgraded to FTTP but Zen left them on the ADSL profile internally in Zen and still they got ADSL speeds on a FTTP 900mbs connection.

So maybe Zen got you on the wrong profile and the uplift is when they changed to the correct profile.
 

Pheasant

ULTIMATE Member
I am hopefully several weeks away from ordering FTTC and I was considering Zen over BT but this post is a bit concerning as I planned for the 900mb speed.
Have you tried https://speed.cloudflare.com or https://fast.com to see how they compare?
I tend to distrust speedtest.net as it was poor when I was on Hyperoptic even though Hyperoptic did get me full 1GB when I actually did a large file test with JDownloader (real multi-thread connection with 16 transfers)

JDownloader may be worth a go and using some of the large file test sites to see how it goes with concurrent connections/chunks set to 16.



Fast should come with a health warning: it is patent fakery - the 'speeds' they report are extremely overestimated

In terms of downloading v.large test files, a UK based server will be more repeatable and accurate than an offshore setup, unless they using a CDN - large file transfers from the continent or the US will be negatively impacted by other wider factors outside the control of the serving ISP as they traverse the wider internet. Peering, busy routes and other factors could sway the results.
 

NTTP

Casual Member
Fast should come with a health warning: it is patent fakery - the 'speeds' they report are extremely overestimated

In terms of downloading v.large test files, a UK based server will be more repeatable and accurate than an offshore setup, unless they using a CDN - large file transfers from the continent or the US will be negatively impacted by other wider factors outside the control of the serving ISP as they traverse the wider internet. Peering, busy routes and other factors could sway the results.
100% about fast as they use peering boxes in isp networks but it, like everything is an indicator of the topper speed. I just find speedtest.net useless as it could never get close when I was on Hyperoptic.
If I recall the large file site I used a lot was actually http://speedtest.tele2.net/ which did max out my 1GB Hyperoptic when I had it.

I do find large file (5GB +) with high chunks gives a better indication of your limits over normal testers but there is a certain degree of IT awareness to use them unlike other testers.
 

Msh

Pro Member
I have played around with the laptop directly to the ONT and via the provider's Fritzbox today with my better results (a little inconsistent, but some results there which I will accept). Once I am on my own router I will leave some longer-term testing on to make sure it's all good.

After I noticed these numbers the OR engineer arrived and respliced the fibre outside my house. He said the link quality was the same before and after he respliced, so in reality, nothing changed (as expected).

Now, however, I am unable to get the speeds through my own router, be that double NAT via the fritzbox, or it PPPoEing itself... either way this is a problem for me to look into now. I am not sure what's going on there. I have no leads at the moment, especially since the same router is dealing with VM just fine.

Maybe it was just a profile error by Zen yesterday after all... but let's wait and see
 

Pheasant

ULTIMATE Member
After I noticed these numbers the OR engineer arrived and respliced the fibre outside my house. He said the link quality was the same before and after he respliced, so in reality, nothing changed (as expected).
I guess it makes them look as through they did something…have something to report back to Zen. But when the connection is working (perfectly) that is there is sufficient light - then resplicing fibre is utterly pointless. They would have been better off spending the resources doing a PON based speed test. Old dogs and new tricks.
 

candlerb

Top Member
I still don't quite know how significant this data is and I have never heard of speedtest-cli performing better than the other apps.
It makes sense if you think about it. A graphical speedtest, especially browser-based, burns up CPU with its graphics and animation. A CLI tool doesn't have that overhead. On my 2015 i7 dual-core Macbook Pro, with Activity Monitor open, I can see a browser-based speedtest saturates both CPUs, and that's only on a 300M line.

Windows machines often suffer from two problems: antivirus, and/or bad NIC drivers. So another thing you can try is to boot your machine from an Ubuntu Live USB stick, and run your speedtest again. Although I do take on board that you were able to get more consistent results via Virgin, differences in TCP stack can come into play too. You may be surprised what a real operating system is capable of :)
 

Msh

Pro Member
It makes sense if you think about it. A graphical speedtest, especially browser-based, burns up CPU with its graphics and animation. A CLI tool doesn't have that overhead. On my 2015 i7 dual-core Macbook Pro, with Activity Monitor open, I can see a browser-based speedtest saturates both CPUs, and that's only on a 300M line.
I'd maybe agree if we were talking about that dual core i7 from 7 years ago. We're not. These devices are not CPU bound. I make this comment because speedtest-cli has a terrible reputation with performance issues.

What sort of results are you seeing?
From what I'd expect (like 900+/100) to 400/80. It's all over the place.
 

Msh

Pro Member
H83jcBg.jpeg

You can see this morning I had one acceptable result then again it's chaotic (sorry for the photo of a display :D)

This is taken with the 4800U laptop directly attached to the ONT/PPPoE.

Zen have at least said this is a "prioritised" fault, so hopefully, I can speak with someone who can give a bit more insight some time in the next hours/days,

I get the horrible feeling that the exchange I am connected to is not set up for FTTP capacity, there's only a handful of addresses here that are OR FTTP serviceable. But this is just speculation of course.
 

baby_frogmella

Top Member
I'd maybe agree if we were talking about that dual core i7 from 7 years ago. We're not. These devices are not CPU bound. I make this comment because speedtest-cli has a terrible reputation with performance issues.


From what I'd expect (like 900+/100) to 400/80. It's all over the place.
What does using the Xilo server on speedtest.net (wired test) give you? Do the test 10 times , if you get 700+ Mbps more than 5 times I'd say absolutely nothing wrong with your line. Liker I said earlier, not all speedtests can handle a 1 Gig line and its perfectly normally to see results of 300-400 Mbps on many speedtest sites.

Also if using Windoze 10, I suggest using the Speedtest app rather than using a browser.
 
Top
Promotion
Cheapest Superfast ISPs
  • Hyperoptic £17.99
    Speed 30Mbps, Unlimited
    Gift: None
  • NOW £21.00
    Speed 36Mbps, Unlimited
    Gift: None
  • Shell Energy £21.99
    Speed 35Mbps, Unlimited
    Gift: None
  • Vodafone £22.00
    Speed 38Mbps, Unlimited
    Gift: None
  • Plusnet £22.99
    Speed 36Mbps, Unlimited
    Gift: £75 Reward Card
Large Availability | View All
Cheapest Ultrafast ISPs
  • Gigaclear £17.00
    Speed: 200Mbps, Unlimited
    Gift: None
  • Community Fibre £20.00
    Speed: 150Mbps, Unlimited
    Gift: None
  • Virgin Media £25.00
    Speed: 108Mbps, Unlimited
    Gift: None
  • Vodafone £25.00
    Speed: 100Mbps, Unlimited
    Gift: None
  • Hyperoptic £25.00
    Speed: 150Mbps, Unlimited
    Gift: None
Large Availability | View All
Helpful ISP Guides and Tips
»
»
»
»
»
»
»
»
»
»
»
»
»
»
»
»
»
»
»
»
»
»
»
»
»
»
»
»
»
Promotion
The Top 20 Category Tags
  1. FTTP (4015)
  2. BT (3131)
  3. Politics (2086)
  4. Building Digital UK (2007)
  5. Openreach (1950)
  6. FTTC (1917)
  7. Business (1803)
  8. Mobile Broadband (1588)
  9. Statistics (1487)
  10. FTTH (1369)
  11. 4G (1359)
  12. Virgin Media (1264)
  13. Ofcom Regulation (1229)
  14. Fibre Optic (1221)
  15. Wireless Internet (1220)
  16. Vodafone (920)
  17. EE (899)
  18. 5G (874)
  19. TalkTalk (815)
  20. Sky Broadband (782)
Sponsored

Copyright © 1999 to Present - ISPreview.co.uk - All Rights Reserved - Terms  ,  Privacy and Cookie Policy  ,  Links  ,  Website Rules