Afrihost Weird Speed Test Results...

qdelpeche

Member
Joined
Sep 2, 2019
Messages
10
Reaction score
5
Hi Everyone,

I have Fibre through Frogfoot and Afrihost, in Sunningdale Cape Town. I have a 400 Down / 200 Up package and I am noticing something really weird.

I run a Speed Test every day at Midnight and over the last few months I have noticed the following:
  • Sat 22 Aug 2026
    • Download: 393.55 Mbps
    • Upload: 204.36 Mbps
  • Fri 21 Aug 2026
    • Download: 83.28 Mbps
    • Upload: 76.81 Mbps
  • Thu 20 Aug 2026
    • Download: 83.21 Mbps
    • Upload: 78.51 Mbps
  • Wed 19 Aug 2026
    • Download: 85.11 Mbps
    • Upload: 76.96 Mbps
  • Tue 18 Aug 2026
    • Download: 393.66
    • Upload: 204.59
  • Mon 17 Aug
    • Download: 393.81 Mbps
    • Upload: 204.56 Mbps
  • Sun 16 Aug 2026
    • Download: 83.91 Mbps
    • Upload: 78.90 Mbps
So this morning, I ran some tests directly. This is a Linux server plugged directly in to the router with no other devices connected inbetween. Something I have tried to explain to Afrihost support on numerous ocassions, but that is a story for another day.
Surely you should get what you pay for, or am I just being naive? I have tried raising this with Afrihost, sent them all my logs, but they insist there is nothing wrong when they test. Is there something going on here, are they charging us for something that they then throttle when no one is watching? Not trying to be conspiratorial or anything , but something weird is going on here.

Something just feels off here. Anybody have suggestions or ideas? Thank you for your time.

These tests are 1 after the other in a space of 5 minutes.

Speedtest by Ookla

Server: Webafrica Networks (Pty) Ltd - Cape Town (id: 69537)
ISP: AFRIHOST OTHER
Idle Latency: 3.33 ms (jitter: 0.94ms, low: 2.54ms, high: 4.05ms)
Download: 82.73 Mbps (data used: 37.4 MB)
14.17 ms (jitter: 1.39ms, low: 4.46ms, high: 25.42ms)
Upload: -73786976294838.20 Mbps [| ] 0% - latency: 14 Upload: 77.71 Mbps (data used: 36.2 MB)
36.24 ms (jitter: 1.63ms, low: 8.46ms, high: 40.09ms)
Packet Loss: Not available.


Speedtest by Ookla

Server: Tech5 Group Pty Ltd - Cape Town (id: 57315)
ISP: AFRIHOST OTHER
Idle Latency: 3.02 ms (jitter: 0.25ms, low: 2.91ms, high: 3.53ms)
Download: 393.68 Mbps (data used: 177.8 MB)
14.83 ms (jitter: 0.97ms, low: 3.30ms, high: 22.14ms)
Upload: 204.46 Mbps (data used: 92.1 MB)
25.22 ms (jitter: 1.54ms, low: 3.35ms, high: 38.71ms)
Packet Loss: 0.0%


Speedtest by Ookla

Server: Tech5 Group Pty Ltd - Cape Town (id: 57315)
ISP: AFRIHOST OTHER
Idle Latency: 3.33 ms (jitter: 1.04ms, low: 2.22ms, high: 3.89ms)
Download: 393.12 Mbps (data used: 177.7 MB)
20.01 ms (jitter: 1.02ms, low: 2.28ms, high: 22.21ms)
Upload: 204.51 Mbps (data used: 92.1 MB)
33.63 ms (jitter: 1.51ms, low: 3.17ms, high: 38.87ms)
Packet Loss: 0.0%


Speedtest by Ookla

Server: Webafrica Networks (Pty) Ltd - Cape Town (id: 69537)
ISP: AFRIHOST OTHER
Idle Latency: 4.00 ms (jitter: 0.25ms, low: 3.65ms, high: 4.10ms)
Download: 80.19 Mbps (data used: 85.0 MB)
14.15 ms (jitter: 1.43ms, low: 2.99ms, high: 25.30ms)
Upload: 75.41 Mbps (data used: 85.6 MB)
36.01 ms (jitter: 2.33ms, low: 6.44ms, high: 67.02ms)
Packet Loss: Not available.


Speedtest by Ookla

Server: Network Platforms (Pty) Ltd - Cape Town (id: 8453)
ISP: AFRIHOST OTHER
Idle Latency: 3.99 ms (jitter: 0.21ms, low: 3.81ms, high: 4.29ms)
Download: 78.26 Mbps (data used: 53.8 MB)
28.45 ms (jitter: 7.27ms, low: 5.51ms, high: 288.87ms)
Upload: 71.93 Mbps (data used: 37.0 MB)
30.99 ms (jitter: 9.21ms, low: 6.07ms, high: 295.34ms)
Packet Loss: 0.0%


Speedtest by Ookla

Server: Atomic Access - Cape Town (id: 48238)
ISP: AFRIHOST OTHER
Idle Latency: 3.87 ms (jitter: 0.21ms, low: 3.55ms, high: 3.98ms)
Download: 83.78 Mbps (data used: 40.0 MB)
21.05 ms (jitter: 1.28ms, low: 6.02ms, high: 29.51ms)
Upload: 71.44 Mbps (data used: 81.3 MB)
36.76 ms (jitter: 4.82ms, low: 5.99ms, high: 293.26ms)
Packet Loss: 0.0%


Speedtest by Ookla

Server: RSAWEB - Cape Town (id: 71176)
ISP: AFRIHOST OTHER
Idle Latency: 3.54 ms (jitter: 0.87ms, low: 2.44ms, high: 4.23ms)
Download: 79.20 Mbps (data used: 42.9 MB)
30.89 ms (jitter: 8.09ms, low: 11.71ms, high: 284.13ms)
Upload: -73786976294838.20 Mbps [\ ] 0% - latency: 30 Upload: 74.93 Mbps (data used: 33.9 MB)
38.06 ms (jitter: 16.89ms, low: 7.33ms, high: 314.06ms)
Packet Loss: 0.0%
 
Update, while going through the data, I noticed this:

Speedtest serverDownloadUpload
Webafrica82.7377.71
Tech5393.68204.46
Tech5393.12204.51
Webafrica80.1975.41
Network Platforms78.2671.93
Atomic Access83.7871.44
RSAWEB79.2074.93

The fact that Tech5 repeatedly gives me almost exactly the provisioned 400/200, while several completely different Cape Town Speedtest servers cluster remarkably closely around ~80/75, tells us something important.

Your Frogfoot fibre line is clearly capable of delivering 400/200. That makes things like my Ethernet connection, Linux server, router port negotiating at 100 Mbps, ONT provisioning at ~100 Mbps, etc. extremely unlikely explanations. If any of those were the bottleneck, Tech5 wouldn't suddenly deliver 393 Mbps.

Afrihost explicitly says its Pure Fibre service is uncapped, unshaped and unthrottled, with no usage thresholds.

Their Fibre terms nevertheless say speeds are supplied on a best-effort basis, depending on the last-mile provider's constraints and demand on Afrihost's network. They specifically state that varying fibre speeds don't automatically constitute a fault.

So legally/contractually, I know I am probably not entitled to 400.000 Mbps 24/7 to every destination on the Internet.

But there's an enormous difference between that and what I am seeing. I'd consider 370–395 Mbps perfectly reasonable for a 400 Mbps residential connection. I'd even accept significant occasional variation caused by congestion.

But ~80 Mbps repeatedly on a 400 Mbps connection, while another local endpoint simultaneously gives 393 Mbps, deserves a technical explanation.
 
This does look like an interface issue to be honest, do you have a 2nd device to run the test side by side?
 
Afrihost explicitly says its Pure Fibre service is uncapped, unshaped and unthrottled, with no usage thresholds.
If you believe that, you might as well believe the schyte talanum1 spews
 
If you believe that, you might as well believe the schyte talanum1 spews

This guy again... :cautious:

Afrihost doesn't even have any device that can restrict traffic or monitor usage anymore. They were retired from service and all FTTH was moved off them.
 
To be fair, any interface will show you the egress and ingress.
I doubt every subscriber has an interface on their core router and I doubt they own the infrastructure on the access side either (lets not even consider GPON shared interfaces)
 
Which leaves us with packet sniffing blah blah, mirrored interfaces. So instead of going through a traffic management device, all packets are duplicated to another device for sniffing and accounting. Pure speculation
 

You make a super good point, how is this data collected then? @cavedog @Afrigirl

There used to be Sandvine devices in between. These are the type of devices mobile networks use. It's how you can control allocations like after hours data or whatsapp data bundles and youtube only bundles etc.

Those were removed so DHCP FTTH now doesn't have the DPI device in between and so data usage is no longer recorded - for now - but the existing pppoe users are not affected as we use basic radius acounting data to populate into those usage graphs in clientzone.

DHCP is getting moved to PPPoE soonish and we are setting up authentications on the new bngs for the dhcp people so we will use radius accounting for those session so dhcp should start recording usage soonish.
 
There used to be Sandvine devices in between. These are the type of devices mobile networks use. It's how you can control allocations like after hours data or whatsapp data bundles and youtube only bundles etc.

Those were removed so DHCP FTTH now doesn't have the DPI device in between and so data usage is no longer recorded - for now - but the existing pppoe users are not affected as we use basic radius acounting data to populate into those usage graphs in clientzone.

DHCP is getting moved to PPPoE soonish and we are setting up authentications on the new bngs for the dhcp people so we will use radius accounting for those session so dhcp should start recording usage soonish.
When will this switch/change take place?

And would this be staged per region or FNO? Will both systems run concurrently or one big bang switch?

Really need to know beforehand, as this change is pretty big, and working from home, I'd like to ensure I have an sufficient backup (or potentially switching service providers for the time being)
 
When will this switch/change take place?

And would this be staged per region or FNO? Will both systems run concurrently or one big bang switch?

Really need to know beforehand, as this change is pretty big, and working from home, I'd like to ensure I have an sufficient backup (or potentially switching service providers for the time being)

I can't speak for how Afri. will handle it, but when Axxess moved us over in Aug. 2024 already (I referred back to an e-mail they sent me), it took a minute or two to go into my router settings via TP-Link app, nominate PPPoE, and put the username and password they sent me in appropriately.

As long as it authenticates you shouldn't have any downtime, apart from the time you spend putting the details in.
If it doesn't authenticate, I presume you go back to using DHCP temp., then contact them here to have a look as to why & sort out.

It shouldn't turn into a major event for you to navigate.

DHCP is getting moved to PPPoE
 
DHCP is getting moved to PPPoE soonish

Thanks for the info, another great insight and value of having AfriFolks around here.

Isn't PPPOE bad? I refer to people smarter than me

At the very least, your router needs to start doing encapsulation, right?
 
Which leaves us with packet sniffing blah blah, mirrored interfaces. So instead of going through a traffic management device, all packets are duplicated to another device for sniffing and accounting. Pure speculation

RICA requires every ISP by law to make sure that their networks are so called, 'interception ready'. So yes - they have the hardware to route duplicate streams, but do not use it (yeah right huh?) unless instructed by Law Enfocement.
 
RICA requires every ISP by law to make sure that their networks are so called, 'interception ready'. So yes - they have the hardware to route duplicate streams, but do not use it (yeah right huh?) unless instructed by Law Enfocement.
Being able to do this at switch level on a case by case basis is very different to having automated systems to do so.

As mentioned Sandvine is able to do so, but to have Sandvine do so for the whole fibre base would make it unfeasible as an offering.
 
Being able to do this at switch level on a case by case basis is very different to having automated systems to do so.

As mentioned Sandvine is able to do so, but to have Sandvine do so for the whole fibre base would make it unfeasible as an offering.

How many interception orders did you get from April 2025 till April 2026? Single digits, double, triple...more?
 
Top
Sign up to the MyBroadband newsletter
X