Tool AT-W10REV. 2026-A
Speed test: how fast is your internet, and is it enough?
The speed test measures ping, jitter, download and upload to our web server and then tells you whether that is enough for your daily use. As a rule of thumb: according to Microsoft, a video call needs about 4 Mbit/s download and 2.5 upload, and a 4K stream 15 to 20 Mbit/s. Add up what runs at the same time and allow 50 per cent as a buffer.
- free
- no sign-up
- stores no measurements
01 / MEASUREMENT
Measure first, then assess
Measurement · Assessment
What is measured is the path from your device to our web server. Wi-Fi, other devices on the network and the browser all affect the result. As evidence towards your provider, only the Federal Network Agency's broadband measurement (Breitbandmessung) is any use.
Step 1Measure
One run takes about 20 seconds and transfers test data: about 60 MB on a 50 Mbit/s connection, at most 200 MB on very fast connections. On mobile data it comes off your data allowance.
02 / MEASUREMENTS
What do download, upload, ping and jitter mean?
Download is what arrives at your end each second. It determines how fast pages load, streams start and updates finish. It is measured in megabits per second: 100 Mbit/s is 12.5 megabytes per second.
Upload is the opposite direction, that is, what you send out: your camera image in a video call, the cloud backup, the photo in an e-mail attachment. With DSL and cable, upload is a fraction of download. That is exactly where things get tight first when working from home.
Ping is the time a small request takes there and back, in milliseconds. It decides whether a conversation feels delayed. Microsoft Teams rates round trips under 200 ms as unremarkable; the statutory minimum service allows at most 150 ms. It also matters more than you would think when browsing: from about 16 Mbit/s, it is mainly the round-trip time that determines how fast a page loads, no longer the bandwidth (Sundaresan 2013).
Jitter says how much the ping varies from one measurement to the next. We calculate it as the mean difference between successive pings. If it is over 30 ms, the sound breaks up even when the bandwidth is enough. The typical cause is Wi-Fi struggling with neighbouring networks or thick walls: in a series of measurements at 47 access points, the wireless link alone cost more than 100 ms for one packet in ten (Pei 2016).
03 / GUIDELINE VALUES
How much bandwidth does what need?
The manufacturers say themselves what their services need. We have collected the figures here, each with its source. Where two manufacturers give different values, the tool uses the higher one. All figures apply per device and direction.
| Application | Download | Upload | Source |
|---|---|---|---|
| Teams meeting, minimum (up to 240p) | 0,2 | 0,15 | Microsoft |
| Teams meeting, recommended (up to 1080p) | 4,0 | 2,5 | Microsoft |
| Zoom group video 720p | 1,8 | 2,6 | Zoom |
| Zoom group video 1080p | 3,0 | 3,8 | Zoom |
| Teams audio, recommended | 0,058 | 0,058 | Microsoft |
| Phone call over the internet (FRITZ!Box) | 0,1–0,11 | 0,1–0,11 | AVM |
| HD video (720p) | 2,5–3 | — | YouTube, Netflix |
| Full HD video (1080p) | 5 | — | Netflix, YouTube |
| 4K video | 15–20 | — | Netflix 15, YouTube 20 |
Two worked examples
Family of three: one parent working from home with a video call, one child watching in Full HD, and a 4K film playing in the living room. That adds up to 32 Mbit/s download and 3.4 upload. With a 50 per cent buffer it is 48 and 5 Mbit/s. A 50 Mbit/s connection is therefore enough; 16 Mbit/s is not, as soon as everything runs at once.
Office with six workstations: three video calls at the same time, four phone calls over the internet. That comes to 18.4 Mbit/s download, but already 9.7 Mbit/s upload, rising to 14.6 with the buffer. A 50 Mbit/s VDSL connection often has only 10 Mbit/s upload. Download would be no issue; upload is.
For browsing, e-mail and office software in the browser there is no manufacturer figure. The tool assumes 1 Mbit/s download and 0.25 upload per person. That is our assumption, not a measured value. We add the 50 per cent buffer because updates, cloud syncing and page loads come in bursts and no connection delivers the same amount every minute. For video calls there is a further point: after a brief drop in the line, Zoom, Meet and Teams took up to 50 seconds in measurements before picture and sound ran stably again (MacMillan 2021).
04 / DEVIATIONS
Why does every speed test show something different?
If you run three tests in a row, you get three results. That is not a fault but down to what is being measured. The most common reasons:
- The remote end. Every test measures to its own server. Our website is delivered through our content delivery service Cloudflare; so what is measured is the path to its nearest data centre and on to our web server. Tests with servers in your provider's network can show more because the path is shorter.
- The Wi-Fi. Wireless shares its bandwidth with all devices on the same network, and every wall costs signal. A laptop upstairs measures your Wi-Fi, not your connection. In home networks above about 20 Mbit/s, Wi-Fi was more often the bottleneck than the line (Sundaresan 2016).
- Other devices. A Windows update, a phone pushing photos to the cloud, a stream in progress: everything that runs in the background is missing from the measurement.
- The time. Even within the same household, values vary considerably from day to day (Deng 2015). The time of day alone explains less than you might suspect (Paul 2022). On a cable network, the households on one segment share the capacity; there it is worth comparing with a measurement in the evening.
- Device and browser. An older computer, a 100 Mbit/s network card, a VPN or an antivirus scanner that checks every data stream will slow the measurement regardless of the line.
- The measuring method. Two widely used test services differed by up to a factor of two on the same tariff in US data because they measure differently (Paul 2022).
Our test opens 4 connections in parallel per direction, measures for about 8 seconds and discards the start-up phase, in which a connection is still picking up speed. What remains is the throughput in steady state. This is what measurement research has recommended for years (Feamster 2020).
05 / TROUBLESHOOTING
Wi-Fi or connection: how to find the bottleneck
“Our internet is too slow” often means, in practice, that the Wi-Fi in the back office is too slow. The sequence below separates the two without you having to buy anything.
- Measure by cable. Plug a computer directly into the router with a network cable, switch off Wi-Fi on the computer, start the test. That is the value of your connection.
- Compare with the contract. The contract summary lists three values: minimum, normally available and maximum speed. If the cable measurement is close to the normal one, the connection is fine.
- Then measure on Wi-Fi, where it stutters. If a lot arrives on the cable and little in the room, the Wi-Fi is the bottleneck. What helps then is in the guide Improving Wi-Fi at home; for workshops and warehouses, see the guide Wi-Fi in the workshop hall. How many access points a building needs is calculated by the Wi-Fi planner.
- If even the cable measurement is poor: take other devices off the network, restart the router, measure again. With DSL, the router's interface shows the synchronised data rate. If it is already below the tariff, the cause is the line and not your building.
- Measure at three times of day and copy the results. The button in the tool adds the date and time. With this list you report a fault to your provider, and they cannot fall back on “everything was fine a moment ago”.
If you would rather not go through this yourself: we check the connection, router and Wi-Fi at your place, at home through PC help for Wi-Fi and internet, and at your business through our network services.
06 / YOUR RIGHTS
What can you do if the provider delivers too little?
If a fixed-line connection delivers considerably, permanently or regularly less than agreed, consumers may reduce the fee or cancel without notice under § 57 Abs. 4 TKG. You have to prove this with the Federal Network Agency's measurement method, the desktop app on breitbandmessung.de. A screenshot of any speed test, ours included, is not enough for that.
How the measurement campaign works
- 30 measurements within 14 days, spread over 3 calendar days with 10 measurements each.
- Between two measurement days there is at least one calendar day without measurements.
- At least 5 minutes between two measurements, and at least 3 hours between the fifth and sixth of a day.
- Measurements are made on a computer connected directly to the router by network cable. Wi-Fi or powerline links in between are not permitted, and a laptop must be plugged into mains power.
When there is a deviation
One of the three conditions is enough, in download or in upload:
- 90 per cent of the agreed maximum speed is reached at least once on no more than one of the three measurement days.
- The normally available speed is not reached in 90 per cent of the measurements.
- The minimum speed is undercut at least once on each of at least two of the three measurement days.
The three reference values are in your provider's product information sheet or contract summary. The rules in detail are explained by the Federal Network Agency. The right to a reduction applies to consumers; for businesses, what counts is what the contract says. Even so, the measurement record is the strongest argument in negotiating with the provider.
If, in principle, nothing more is available at your address, the right to a basic supply applies: it guarantees 15 Mbit/s download, 5 upload and at most 150 ms latency, no more. Which connections are actually available in the Vogtland, from DSL through fibre to wireless, is in the guide Slow internet in the Vogtland.
Legal position and manufacturer figures last checked on 13.11.2026.
07 / RESEARCH
What research says about speed tests
How to measure internet speed properly is a field of research in its own right. Six findings have shaped this tool:
- One connection is not enough. Several parallel connections come much closer to the booked speed than a single one; research recommends at least four (Sundaresan 2011, Feamster 2020). Our test uses 4.
- The first seconds do not count. Every connection starts slowly. If you include this phase, you measure too little; reliable tests leave it out (Feamster 2020).
- Often you are measuring the Wi-Fi. From about 20 Mbit/s, the bottleneck in home networks lies more often in the wireless link than in the line (Sundaresan 2016). That is why every troubleshooting starts with the measurement on the cable.
- Full line, high ping. If the upload is fully used, data piles up in oversized buffers in the modem and router. The ping then rises many times over, in the worst case measured to over ten seconds (Sundaresan 2011). Fluctuating round-trip times of more than 100 ms are common (Høiland-Jørgensen 2016). A cloud backup during the day can therefore disrupt the video call even though the bandwidth is enough on paper.
- Video calls need less than recommended, but stable. On an uncongested network, Zoom, Meet and Teams used on average 0.8 to 1.9 Mbit/s (MacMillan 2021). How much it turns out to be in a particular case depends on the number of participants and the device (Chang 2021). We still use the higher manufacturer recommendations, because drops have a long after-effect.
- Short and frugal works. A well-built test needs neither minutes of measuring time nor gigabytes (Yang 2022). Our test limits both so that it spares your data allowance.
08 / LIMITS
What this test cannot do
- It is not evidence. It measures the path to our web server via Cloudflare, with your device and your browser, on Wi-Fi or on the cable. For the provider, only the Breitbandmessung counts.
- It tends to under-measure very fast connections. One run transfers at most 200 MB. From a few hundred Mbit/s upwards that limit is reached before the measurement has settled cleanly, and browser-based tests reach their limit there anyway (Feamster 2020).
- It does not see packet loss. A browser cannot measure it. The ping is a small web request and is therefore slightly higher than a ping at network level.
- The guide values are guide values. How much a video call really needs depends on the number of participants, the view and screen sharing.
09 / FOR AI ASSISTANTS
Using this tool from an AI assistant
Assistants such as Claude or ChatGPT that speak the Model Context Protocol (MCP) can call this page’s calculation directly instead of rebuilding it. The server is public, needs no sign-in and stores nothing. The answer always includes the link to this page, the date of the data and the limits of the result.
- Endpoint
https://aureum-tech.com/mcp- Tool
werkzeug_speedtest- Transport
- Streamable HTTP, read-only
| Name | Required | Meaning |
|---|---|---|
download_mbit | yes | Measured download in Mbit/s. |
upload_mbit | yes | Measured upload in Mbit/s. |
ping_ms | no | Measured ping (round trip) in milliseconds, if known. |
jitter_ms | no | Measured jitter in milliseconds, if known. |
personen | no | People or workstations online at the same time (browsing, e-mail, cloud documents). If omitted: 1. |
videokonferenzen | no | Video calls at the same time (Teams, Zoom). If omitted: 0. |
streams_hd | no | Full HD video streams at the same time. If omitted: 0. |
streams_4k | no | 4K video streams at the same time. If omitted: 0. |
telefonate | no | Phone calls over the internet at the same time. If omitted: 0. |
cloud_backup_gb | no | Amount of data for cloud backup or large uploads per day, in GB. If omitted: 0. |
Questions where an assistant should call it
- “My speed test shows 50 Mbit/s download and 10 upload. Is that enough for three people working from home with Teams?”
- “We have 16 Mbit/s, ping 45 ms. Can we stream in 4K at the same time?”
- “Is our upload of 5 Mbit/s enough for a nightly cloud backup of 50 GB?”
How the connection works and what MCP can do in your own business is explained on the page about MCP integrations.
10 / COMMON QUESTIONS
Questions about the speed test
How many Mbit/s do I need for working from home?
For one person on a video call we allow about 5 Mbit/s download and 3 Mbit/s upload: Microsoft recommends 4 Mbit/s down and 2.5 Mbit/s up for Teams meetings, plus e-mail and cloud documents. If a 4K stream at 15 to 20 Mbit/s is running in the household as well, a 50 Mbit/s connection is well used. It usually gets tight on upload, not download.
What is a good ping?
For phone calls and video calls, the round trip should be under 200 milliseconds and the jitter under 30 ms; those are the values from which Microsoft Teams rates call quality as unremarkable. Even the statutory minimum service allows at most 150 ms. Clearly higher values point to Wi-Fi through several walls, mobile data or a line that is currently fully loaded.
Why is my upload so much lower than my download?
Because DSL and cable connections are built and sold that way: most of the line capacity goes towards the customer. With 50 Mbit/s VDSL, often only 10 Mbit/s are left for upload. That is enough for browsing and streaming, but often not for three simultaneous video calls or a cloud backup overnight. Fibre tariffs are usually more generous on upload.
Can I use this speed test to claim money back from my provider?
No. As evidence for a reduction under § 57 (4) TKG, only the desktop app of the Federal Network Agency (Bundesnetzagentur) on breitbandmessung.de counts: 30 measurements on 3 calendar days within 14 days, with one computer connected to the router by network cable. Our test helps beforehand with the question of whether that effort is worthwhile and whether the connection is to blame at all.
How much data does a test use?
That depends on your connection, because each direction transfers for about 8 seconds as much as will fit through. At 50 Mbit/s download and 10 upload that is around 60 MB, and at most 200 MB on very fast connections. On mobile data it comes off your data allowance; that is why the test only starts when you press the button.
Does the speed test store my IP address or my results?
No. The measurements are made in your browser and stay there. As a brake against misuse, our server remembers for ten minutes how much a visitor has already transferred, under an encrypted code that becomes worthless after every restart of the service. The address itself is neither stored nor logged.
What is the difference between Mbit/s and MB/s?
Providers and speed tests count in megabits per second, while download windows in the browser usually show megabytes per second. A byte has eight bits, so 100 Mbit/s is about 12.5 MB/s. A 5 gigabyte update therefore takes just under seven minutes at 100 Mbit/s and about 42 minutes at 16 Mbit/s.
11 / STUDIES AND SOURCES
What the tool is based on
The guide values and notes above are based on these 11 scientific papers. Each has been checked through its DOI or the publisher's page (last checked on 13 November 2026): it exists, it has not been retracted, and its finding still holds for the question today.
Feamster, N.; Livingood, J. (2020): Measuring internet speed: current challenges and future recommendations. Communications of the ACM 63 (12), pp. 72–80
Finding A speed test always measures the weakest link on the path, often the Wi-Fi or the device. Good tests use several parallel connections, run long enough and leave the connection's start-up phase out of the calculation; browser-based tests barely get beyond a few hundred Mbit/s.
In the tool Four parallel streams, a discarded start-up phase and the note that very fast connections are under-measured in the browser all go directly back to this paper.
doi.org/10.1145/3372135 · Journal article
Sundaresan, S.; de Donato, W.; Feamster, N. et al. (2011): Broadband internet performance: a view from the gateway. Proceedings of ACM SIGCOMM 2011, pp. 134–145
Finding Measurements at around 4,000 home routers: several parallel TCP connections come closer to the booked speed than a single one. When the upload is fully used, the round-trip time rises many times over because of oversized buffers in the modem, in the worst case measured to over ten seconds.
In the tool This provides the reason for the parallel streams and for the warning that a running cloud backup drives up the ping and disrupts video calls.
doi.org/10.1145/2018436.2018452 · Conference paper
Sundaresan, S.; Feamster, N.; Teixeira, R. (2016): Home Network or Access Link? Locating Last-Mile Downstream Throughput Bottlenecks. Passive and Active Measurement (PAM 2016), LNCS 9631, pp. 111–123
Finding In home networks the Wi-Fi is more often the bottleneck than the connection, especially in households with more than about 20 Mbit/s download.
In the tool From 20 Mbit/s the tool expressly advises a comparison measurement on the cable before you suspect the connection.
doi.org/10.1007/978-3-319-30505-9_9 · Conference paper
Sundaresan, S.; Feamster, N.; Teixeira, R.; Magharei, N. (2013): Measuring and mitigating web performance bottlenecks in broadband access networks. Proceedings of the ACM Internet Measurement Conference (IMC 2013), pp. 213–226
Finding Measurements from over 5,000 connections: if the download is above about 16 Mbit/s, it is mainly latency that determines how fast web pages load, no longer bandwidth.
In the tool Explains on the page why, beyond a certain bandwidth, a low ping does more for browsing than a faster tariff.
doi.org/10.1145/2504730.2504741 · Conference paper
Paul, U.; Liu, J.; Gu, M. et al. (2022): The importance of contextualization of crowdsourced active speed test measurements. Proceedings of the ACM Internet Measurement Conference (IMC 2022), pp. 274–289
Finding Around 1.5 million speed tests from four US cities: many low values are explained by small tariffs, Wi-Fi instead of cable, radio band and devices. Two widely used test services differed by up to a factor of two on the same tariff; the time of day played only a small role.
In the tool That is why the tool asks for the context, compares with the contract rather than with other tests, and explains why test services show different values.
doi.org/10.1145/3517745.3561441 · Conference paper
Deng, X.; Hamilton, J.; Thorne, J.; Sivaraman, V. (2015): Measuring broadband performance using M-Lab: Why averages tell a poor tale. International Telecommunication Networks and Applications Conference (ITNAC 2015), pp. 24–29
Finding Data from almost 97,000 Australian households: even within the same household, performance varies considerably within a day and from day to day; averages hide this.
In the tool This provides the reason for the advice to measure several times and collect the values with timestamps, and explains why the official procedure requires 30 measurements on three days.
doi.org/10.1109/ATNAC.2015.7366784 · Conference paper
Høiland-Jørgensen, T.; Ahlgren, B.; Hurtig, P.; Brunstrom, A. (2016): Measuring Latency Variation in the Internet. Proceedings of ACM CoNEXT 2016, pp. 473–480
Finding Fluctuating round-trip times are common: in two thirds of the samples, latency varied by more than 100 ms. The main cause is queues that form when the line is full, over 200 ms at the median.
In the tool That is why the tool measures jitter as well as ping and does not rate telephony and video calls on bandwidth alone.
doi.org/10.1145/2999572.2999603 · Conference paper
Pei, C.; Zhao, Y.; Chen, G. et al. (2016): WiFi can be the weakest link of round trip network latency in the wild. IEEE INFOCOM 2016, pp. 1–9
Finding Two months of measurements at 47 access points: for more than half of the packets the Wi-Fi link alone cost over 20 ms, for one in ten over 100 ms, often more than half of the total round-trip time.
In the tool This supports the note that high ping and jitter often come from the Wi-Fi and that a measurement by cable settles the matter.
doi.org/10.1109/INFOCOM.2016.7524396 · Conference paper
MacMillan, K.; Mangla, T.; Saxon, J.; Feamster, N. (2021): Measuring the performance and network utilization of popular video conferencing applications. Proceedings of the ACM Internet Measurement Conference (IMC 2021), pp. 229–244
Finding Zoom, Google Meet and Teams used on average 0.8 to 1.9 Mbit/s on an uncongested line. After a brief drop in bandwidth, some took up to 50 seconds to recover, and in competition one application took more than three quarters of the line.
In the tool Explains why the tool uses the higher manufacturer recommendations and asks for a 50 per cent buffer: brief bottlenecks have a long after-effect in conferences.
doi.org/10.1145/3487552.3487842 · Conference paper
Chang, H.; Varvello, M.; Hao, F.; Mukherjee, S. (2021): Can you see me now? A measurement study of Zoom, Webex, and Meet. Proceedings of the ACM Internet Measurement Conference (IMC 2021), pp. 216–228
Finding Over 700 conference sessions: the delay depends on where the provider runs its servers, and the data rate changes with the number of participants and the state of the device.
In the tool This is the basis for the note that bandwidths for video calls are guide values and differ from case to case.
doi.org/10.1145/3487552.3487847 · Conference paper
Yang, X.; Lin, H.; Li, Z. et al. (2022): Mobile access bandwidth in practice: measurement, analysis, and implications. Proceedings of ACM SIGCOMM 2022, pp. 114–128
Finding Analysis of 3.54 million users of a commercial bandwidth test: a newly built test managed on average with one second instead of ten seconds of measuring time, at one fifteenth of the server costs.
In the tool Shows that a test does not have to run long to be useful. The tool measures about 8 seconds per direction and caps the amount of data so that it spares your data allowance.
doi.org/10.1145/3544216.3544237 · Conference paper
Standards, authorities and manufacturer data
- Internet speed: evidence procedure for fixed networks (Internetgeschwindigkeit: Nachweisverfahren Festnetz) · Federal Network Agency (Bundesnetzagentur)
- Right to be supplied with telecommunications services (Recht auf Versorgung mit Telekommunikationsdiensten) · Federal Network Agency (Bundesnetzagentur)
- § 57 Telecommunications Act (Telekommunikationsgesetz, TKG) · Federal Ministry of Justice (Bundesministerium der Justiz)
- Prepare your organization’s network for Microsoft Teams · Microsoft
- Monitor call and meeting quality in Microsoft Teams · Microsoft
- Zoom system requirements: Windows, macOS, Linux · Zoom
- Recommended internet speed (Empfohlene Internetgeschwindigkeit) · Netflix
- System requirements (Systemanforderungen) · YouTube
- How much data is transferred in internet calls? (Welche Datenmenge wird bei Internetgesprächen übertragen?) · AVM / FRITZ!
12 / MORE TOOLS
Still have a question?
- AT-W2 · Home & business
Windows 10 check
Does this PC still get security updates, and what needs doing by October 2027? - AT-W3 · Home & business
Wi-Fi planner
How many access points or mesh units does your home, office or warehouse need? - AT-W4 · Businesses
IT emergency check
How quickly could your business get back to work after an outage? - AT-W5 · Businesses
E-mail check
Are your e-mails landing in spam? Your domain’s SPF, DKIM and DMARC in plain language. - AT-W6 · Businesses
UPS calculator
Which UPS keeps your server, router and till running long enough during a power cut? - AT-W7 · Home
Digital legacy planning sheet
Which accounts, devices and contracts should your family know about?
Does your speed test show less than you pay for, or does the video call still stutter? We look at the connection, router and Wi-Fi, remotely or on site at your place, and tell you beforehand what it costs.
+49 3744 4399760[email protected]Mon–Fri 8:00–17:00 · Straße der Jugend 5, 08228 Rodewisch