Different speeds from different modems

I had to move my FritzBox router for a bit, plugged in an old Openreach ECI and Edgerouter with PPPoE to keep a connection.

I get a pretty consistent 40/12 connection on FTTC via Zen (on an 80/20 subscription).



However the ECI/Edgerouter gave me a 'better' speed. According to the TBB tests I did I had over 15 up. Which for me is a Big Thing. The download didn't increase and may have dropped a little, but it's the upload I want. A direct Apple Mac PPPoE with the ECI gives the higher upload as well.



So the ECI modem can get me a better upload than the FritzBox?

Or maybe its because I'm testing via the actual Master socket rather than a 12 meter away extension. So I've brought the FritzBox to the Master socket and according the TBB the speed is much the same as when its on the extension.

Now that I'm back on the Fritz (in the Master socket) I see its statistics reporting

Attainable throughput - kbit/s - 38699 - 15772
Current throughput - kbit/s - 37157 - 15723

Meanwhile a TBB test reports the more usual 34.1 & 12.9.

So what's happening? The line appears capable of - and TBB reports - uploads of up to 15.9 kbps. But only when using the ECI. The FritzBox says its doing 15.7 kbps at the moment but TBB only reports 12.9.

Usually I'd put this down to 'that's the way it works'. But I seem to have a better result from the ECI.

In reply to a post by Woolwich:

The FritzBox says its doing 15.7 kbps at the moment but TBB only reports 12.9.

You're looking at a two different things.

The speed reported by the FritzBox is the line rate: the raw number of bits per second which are transferred from your modem via the copper cable to the DSLAM in the cabinet.

TBB is measuring the throughput of TCP/IP, and is measuring just the payload of the packets, not the IP and TCP headers and any other framing overhead on the link. (DSL used to use ATM encapsulation with an overhead of 5 bytes every 48, but I think VDSL is just 1 byte every 64)

You can expect your usable throughput as measured by a speedtest to be about 7% below the line rate.

Now, if you're not achieving that, there's a third thing which may come into play. Some routers intentionally buffer the outbound data and enforce that it's sent out at a speed *slightly below* the actual line speed. This is to ensure better latency on jitter-sensitive applications like VOIP, gaming or interactive logins. If you've got a large background upload going on, and it's buffered in this way, a small interactive packet can "jump the queue" and be sent out next, reducing its latency and jitter.

If the Fritzbox is doing this, then it's doing so to improve your overall Internet experience. You can turn this feature off, but if you try to play games or do voice calls at the same time as your PC is backing up to the cloud (say), then you may find it doesn't work as well.

So it's your call which you go for: squeezing out an extra 5-10% of upload throughput, or having the better performance for interactive applications.

EDIT: the other question is what line stats you're getting from the ECI modem. I think you need to upload custom firmware to it to get these.

Edited by candlerb (Sun 27-Feb-22 17:03:42)

In reply to a post by candlerb:

You're looking at a two different things.

Yes, I think I knew that. But I think I thought the two speeds on the FritzBox were trying to take this into account. There's "attainable throughput", I assume the maximum the line will tolerate, as you say the raw bits per second. Then the second - always lower - figure. But thinking, I don't think I see this number from the TBB speed result. Currently I have 38.8 "attainable" and 37.1 "current" throughput. Is that 7%?, no it should be about 36. Still higher than TBB's last test at 34.

If the FritzBox is lowering upload for interactive performance I'm not sure where that setting is. The only upload setting I can see is "Send direction - Intended signal-to-noise ratio" which I have (along with all the download options) set to "maximum performance". But yes, I'd rather squeeze 5 to 10% extra on the upload. There's no gaming on this line and little VoIP.

EDIT: the other question is what line stats you're getting from the ECI modem. I think you need to upload custom firmware to it to get these.

Involves a soldering iron as I recall. Not happening!

I'll do more tests next weekend but I wonder if its just the location (isn't the advice always have your router as close to the Master as possible!).

You'll find the "attainable" versus "actual" difference on other modem chipsets too.

DLM caps the throughput to the "actual" level, even though the modem chipset thinks it could go faster, in order to make the line more stable (adjusting the SNR target).

If you live very close to the cabinet you may see "attainable" rate of 100,000+ kbps in the downstream side, whereas "actual" is capped at 79,999 kbps.

The attainable/max attainable is just a theoretical figure calculated by the modem and is sometimes very wrong.
When Interleaving is enabled the attainable is particularly exaggerated.

Fritzbox and the ECI modems are both Lantiq chipsets which can see large swings in the attainable rate just between 2 firmware versions.
That's because newer Lantiq dsl firmware uses a more advanced method of calculating the attainable rate.

When comparing the performance of 2 individual modems the sync rate and your actual throughput are really what matter.

Any preference for chips? <insert salt 'n' sauce jibe here>

I'm starting to look for another router and so far DreyTek are at the top of the pile as they claim fast VPN throughput. That will be down to their choice of chip, but I have no idea who's or if it matters much.

More tests, cos some work was cancelled...

ECI Openreach modem connected to Edgerouter X verses FritzBox 7490 with WiFi and DECT off, network disconnected.

Two locations, first at the Openreach Master Socket, then at an extension about 12 metres of cable away.

The TL;DR is (surprising)....

The ECI gets a better overall up and down speed no matter the location. The Master location gets a better upload speed, the Extension is best for download.


FritzBox at Master


ECI at Master


FritzBox at Extension


ECI at Extension

How is this happening? I'm assuming there's bandwith being shared down and up and the ratio is different at the different sockets. Why would that be?

As I've said, upload is what I want so the ECI at the Master is the way to go.

But am I not losing a lot of download for a little on the up?

In reply to a post by Woolwich:

ECI Openreach modem connected to Edgerouter X verses FritzBox 7490

Are you comparing Edgerouter + ECI modem versus FritzBox with its built-in VDSL modem? Then you are changing two things at once.

If the FritzBox can be put into modem mode then you could try Edgerouter + Fritzbox as modem. You could also try FritzBox with ECI modem, using its ethernet WAN port.

But sadly, unless you can get line stats from the ECI modem, it's going to be hard to draw accurate conclusions. And if you keep changing this too much, you may nudge DLM into reducing your line rate as it sees all these disconnections.

In reply to a post by Woolwich:

How is this happening? I'm assuming there's bandwith being shared down and up

Yes: different frequency bands are used for download and upload. Within each band there are many different tones, whose data rate is adjusted based on the line conditions. Small changes in line length, capacitance, reflections etc will affect the achievable speed for each tone.

A good DSL modem will even list all the tones and the data rate achieved on each one.

But personally I'd say you're over-thinking this. How much difference in your day-to-day life does upload of 14.3Mbps versus 13.1Mbps actually make? Pick the solution which is the most reliable, stable, and easy to manage.

If the difference is that important and you're prepared to pay for it, then buy two VDSL lines and bond them.

In reply to a post by candlerb:

But personally I'd say you're over-thinking this.

It was just an observation really, out of having to move a router for a while. I didn't expect to see such a difference, I didn't expect the difference to be different.

Too many combinations to try and I wasn't planning on doing any more.

A bonded line, yes but I'm not going to get double my current 11 to 13 ish for more than double the cost. Upload will continue to take a Long Time and hard drives driven around.

For comparison, FTTC here gives 25-32M down and 4-5M up depending on the time of year. If you're getting 13-15M upload that's actually pretty good.

I went for FTTPoD, but at a time when the pricing was cheaper, and reduced further by a gigabit voucher (no longer available where I live).