The new HG659b is doing the same result.

Down stream can not go over 300mbps, up stream is always maxed out.

Not sure if there is a new version of HG659b now, since the two I have seems identical, except the old one come in a green box says "Business Gateway", the new one in a orange box says "Spark HG659b Modem". So I emailed back Spark and said there is no need to post me another modem.

Well... I received the new HG659b this morning(not sure if Spark received my email says the new router is not necessary). I plugged it in, did not change any settings in the router, with only 1 ethernet device on the network, the speedtest result remains the same. To make sure it is not caused by system/software, I found a USB drive and downloaded the latest Ubuntu to it then boot into Ubuntu. But I can't find speedtest client for Ubuntu on their website so I only did the test with firefox that came with Ubuntu, and speed it the same.

I'll email back my test results to Spark and see what they will suggest.

Until you try a machine proven to work on another gig connection you're chasing your tail.

Cheers - N

Hibino:

 

hio77:

 

What version of Surface? and what dock exactly? sounds likely a USB2.0 to Gbit interface...

 

 

It is the first gen surface book, and the surface dock is this one:

 

https://www.microsoft.com/en-nz/p/surface-dock/8QRH2NPZ0S0P/1187

 

 

 

I just received an email from Spark says they are going to post me a new HG659b for the testing. I actually still got a brand new HG659b, I'll try to find it and do the test again. After the test will let Spark know there is no need to waste money sending me another HG659b

 

the SB1 is quite a slugish cpu.

I'd be happy with 3-400mbit on a Speedtest there.

Talkiet:

 

Until you try a machine proven to work on another gig connection you're chasing your tail.

 

Cheers - N

 

I found the speedtest result using browser back 2015 when the fibre first installed on the same machine I am using now, so it should proven this machine works on gig connection right(I removed my social media tag on the image)?

 

the SB1 is quite a slugish cpu.

 

I'd be happy with 3-400mbit on a Speedtest there.

 

Emmm.... Yes it is old but this is first time I heard that i7-6600u can not handle a gigabit network... But the upload always over 400mbps so I should consider lucky since it is over 3-400mbps?

Anyway, I visited several neighbors but none of them got Spark gigabit fibre so can't test in their place. I am asking my friends see if anyone got a laptop that have gigabit ethernet installed. 

Hibino:

 

Emmm.... Yes it is old but this is first time I heard that i7-6600u can not handle a gigabit network... But the upload always over 400mbps so I should consider lucky since it is over 3-400mbps?

 

 

 

Anyway, I visited several neighbors but none of them got Spark gigabit fibre so can't test in their place. I am asking my friends see if anyone got a laptop that have gigabit ethernet installed. 

 

The thing to remember is Ookla also changed how their test works quite heavily.

For the downstream, there are far more downstream threads used than back in the day, This would mean for quite different CPU Utilization (Sometimes more isn't always better)

There is also a complete change of tech, from flash to HTML5.

Given in most of cases this should be an increase in performance not a decrease. but totally depends on your situation.

I think Geekzone need to crowdfund/ISP donate a little computer (like what 2degrees occasionally send out) with a Linux distro that is proven to work at gbit speeds.

This would allow it to be sent to anyone, on any ISP, for them to test to rule out their devices as the issue (provided they're using the ISP supplied router)

Chances are it'd get lost or stolen though -_-

I'm assuming that the surface book doesn't have a dedicated NIC on it, so the tests will be done over a usb to gbit nic? In which case it could be drivers or firmware update that have made it worse compared to how it was back then, or you're at the limit of usb -> gbit nic etc?

I'll just add in my 2 cents that maybe of use.
Everyone in this thread has raised excellent points, Speedtest has changed the way it runs tests, and without a "current" proven gigabit capable PC on hand, that is still a possible cause for slow speeds.
I will admit it seems odd that you had gigabit speeds using the same PC in the past.

I'm curious if you have tried other speedtest servers outside of the South Island?  Reason being that I tried running tests on my Spark Gigabit (940Mbps to Spark AKL) connection from Auckland to several South Island servers and get the same speeds you are seeing.

(Yes I know you should test to your local server and that routing traffic across the country will impact speeds)

Yet when I test to 2Degrees Sydney or Los Angeles I get 700-800Mbps, not the 300-400Mpbs to Spark/2Degrees Christchurch or Dunedin.

Have you tried any other tests?  Such as trying Nperf instead?  Or maybe some real world tests such as downloading a game from Steam or Origin.  I can often get 700-800Mbps from those.

Just some food for thought. :)

dfnt:

 

Chances are it'd get lost or stolen though -_-

 

I'm assuming that the surface book doesn't have a dedicated NIC on it, so the tests will be done over a usb to gbit nic? In which case it could be drivers or firmware update that have made it worse compared to how it was back then, or you're at the limit of usb -> gbit nic etc?

 

It's a miniDP port from memory... the weird one that looks like a USB B. (Like printers)

on a SB4 it can preform at about 600~800mbit.

Risks around the security of the device is a massive concern..

2D can run iperf from the fritz Just as they can send out those little PC's

Both of them have the risk though of the customer going "Well i dont believe you it's not doing that on MY computer"

I suspect this is why they use the miniPC's - Customer could see a real screenshot rather than commands etc.

Truth be told, Your looking at expensive devices to be able to actually Test gbit often...

All the providers get these sorts of complaints though, Even 2D.

We've already done that with an Odroid 2. They can reliably and consistently test to an iPerf server at full speed. With speedtest-cli it's a lot more variable because that's not an official client and it's performance can be variable in Python.

And yes, they would get immediately lost or stolen.

FYI, there was no special config needed with the odroid. Just install ubuntu, install iperf 3 (and THE SAME VERSION ON THE SERVER) and speedtest-cli and away you go.

Actually, I have just tried it again and it seems some of the recent ookla changes have introduced massive variability into the speedtest-cli tests, and the latency is wrong as well.

odroid@odroid:~$ speedtest-cli
Retrieving speedtest.net configuration...
Testing from Spark New Zealand (122.61.xxx.yyy)...
Retrieving speedtest.net server list...
Selecting best server based on ping...
Hosted by Spark New Zealand (Christchurch) [4.57 km]: 7.541 ms
Testing download speed................................................................................
Download: 828.74 Mbit/s
Testing upload speed......................................................................................................
Upload: 3.84 Mbit/s

Cheers - N

techocrazy:

 

Have you tried any other tests?  Such as trying Nperf instead?  Or maybe some real world tests such as downloading a game from Steam or Origin.  I can often get 700-800Mbps from those.

 

Long as it's not a steampipe compressed game, those are killing on IO and CPU ontop of the normal download.

It just needs to be a live bootable OS which can easily be wiped, discarded etc

In any case, I understand the logistical nightmare behind it and was more commenting that having something like that saves the constant back and forth of customer blaming RSP/not believing it's their equipment at fault

And many times that you do large downloads (depending on client) your SW may pre-allocate the file, which depending on disk subsystem, PCI config etc can hugely impact download speed.

If I am doing a download test, it's always to some sort of >NUL

Cheers -N

techocrazy:I'm curious if you have tried other speedtest servers outside of the South Island?  Reason being that I tried running tests on my Spark Gigabit (940Mbps to Spark AKL) connection from Auckland to several South Island servers and get the same speeds you are seeing.

 

I would agree with this. When I was doing some testing on Spark Gigatown a few months ago, I was getting better speeds to Spark WLG than Chch. 

WLG using desktop app gave me 900/500. 

CHC using desktop app gave about 500/500. 

Thanks for all the suggestions guys.

I did boot my desktop to Ubuntu through USB doing the tests but there is no change.

The surface dock is connected to surface book with it's special connector(I think all surface got this connector since gen 4), it has 1 gigabit ethernet + 2 displayport + 4 usb3 so I guess it should have enough bandwidth to handle a gigabit network.

I just tried different server locations as suggested:

Dunedin 2degrees: 124mpbs down, 100mbps up

Wellington Spark: 136mbps down, 201mbps up

Wellington Vodafone: 230mbps down, 114mpbs up

Auckland Spark: 126mpbs down, 183mbps up

Auckland Vodafone: 130mbps down, 131mpbs up

Auckland MyRepublic: 336mpbs down, 138mbps up(wait...)

Auckland 2degrees: 934mbps down, 151mpbs up(huh??????)

Sydney Vodafone: 238mbps down, 81mpbs up

Sydney 2degrees: 814mbps down, 76mpbs up(what???)

From the results I guess the fibre is gigabit although it only can achieve that on 2 of the 2degrees servers? So the problem is Spark's speedtest server is overloaded then...