Starlink (Generation 2) testing

For me, upload has been pretty consistent at 15Mbps+. It often peaks at 40Mbps or more briefly before settling down to ~15. Slowdowns during peak hours seem to mainly affect download speeds.

One good thing I've found is that if I'm downloading a torrent with hundreds or thousands of seeds, I will get speeds that are much faster than what I get during a speed test. Maybe all those simultaneous connections are brute-forcing more bandwidth? Not really sure. Speeds also increase quite a bit during off-peak hours, especially during the middle of the night. So if you were downloading a huge file or something, at least you can still get good speeds at those times.
Ok that is awesome! Really good news on the upload, though, I know that's pretty location dependent. Hopefully I end up with similar results.
I just spent some time building out a telegraf + grafana to measure out my speeds. Right now it uses speedtest CLI to run a test every 10 minutes (I have no idea if they'll count that frequency as abuse. Hope not). Right now seeing testing jump all over the place
1663921127568.png



The little dropdown near the upper left corner will let me view the same data for any multi-WAN setup I have. Once I have Starlink here and setup, I'll start pulling stats for it as well and compare directly.

I'll probably do "large downloads" over night, so, set that to download like, between midnight and 5am or something.
 
Alright, they expect the dish to ship out on the 3rd, and, no ETA yet on the ethernet adapter. Hopefully both show up at the same time. Already have pretty much everything plumbed for all the testing, just need the hardware.
 
Got the dish on Tuesday, and the ethernet adapter today. Just finished setting up a little test box with it. Basically I have a desktop that's whole life right now is just Starlink. It's plugged into the ethernet adapter directly, and logging speedtest runs every ~5 minutes or so.
The first run was atrocious. 13Mbps down, 4.0Mbps up. The app does say it'll take "Up to 6 more hours" for it to complete calibration. Also, in my current house, there isn't a great area from ground level that gets a full, clear shot to the sky. I'm not going to mount it to the roof, since, we're hoping to sell this place soon and I'm lazy. This should kind of be a worst case test. The new place we want to buy has wide open areas on the ground, and roof mounting would also be a thing I'd do if needed. Plus was less population density, so possibly less congestion.
I don't trust it where it's at (basically just sitting out in my driveway), and don't want someone walking off with it. I'll let it gather data for a couple of hours then see where it is at. Repeat the same thing a bit tomorrow morning and tomorrow night.
 
Little bit of data, so far.
"Calibration" took ~7 hours to complete. Performance did tick upward throughout, though, not substantially.
During times where the app says there's "congestion" (it's unclear what that exactly means), I get ~50Mbps down and 4-5Mbps up. When it doesn't say that, I get 85-100Mbps down and 7-12Mbps up. Those are averages for what that's worth. The peaks and valleys are pretty extraordinary - even on "non-congested" times, I get uploads as low as 3.27Mbps. Latency seems fairly consistent, between 55ms and 65ms.

Uptime is hanging out at about 98.7%. This is just a "simple" test where I ping two different endpoints every 5 seconds, log the latency, and calculate loss.

Overall, this data should still be taken with a grain of salt. The app says I have 3% obstruction in the sky, which doesn't sound like much, but can cause performance and availability issues. I'm not thrilled by the experience, was honestly hoping for a bit more performance out of it. It is serviceable and will be an alright bridge til I can get Comcast. Would not want or recommend for long term, unless you have no other option (or only slower alternatives). I did hold a couple zooms on it - one just a one-on-one, then a small group of 4, then tried a group of 10. The one on one was fine. The 4some showed video quality issues, and the group of 10 had long stutters and both audio and video issues.
 
During times where the app says there's "congestion" (it's unclear what that exactly means), I get ~50Mbps down and 4-5Mbps up. When it doesn't say that, I get 85-100Mbps down and 7-12Mbps up.

That is very interesting. I've had the service since last year and have never seen anything in the app about "congestion". Where in the app are you seeing that? Do you also see it when you visit http://192.168.100.1 (basically the web version of the app)? Although where I have mine, there are no obstructions at all, so maybe it's related to that.

Kind of interesting, last night at about 3am I was running some speed tests and was seeing some of the fastest speeds I've ever seen on Starlink. On one test I got over 400Mbps download and 37Mbps upload. Maybe that's a good sign in some way? Well at least overnight downloads will have access to some good bandwidth.
 
That is very interesting. I've had the service since last year and have never seen anything in the app about "congestion". Where in the app are you seeing that? Do you also see it when you visit http://192.168.100.1 (basically the web version of the app)? Although where I have mine, there are no obstructions at all, so maybe it's related to that.
In the app, if you tap the "Visibility" button, it takes you to a view of the sky with blue and red markers. I forget where but somewhere on that screen it had a %, but, now it doesn't. Maybe it was only showing that percent during calibration cos I can't find it now.
I'm certain the performance is also widly variable depending on the local node. If you're in an area with a lot of Starlink users, I'm sure it can get overloaded. In the app, it has a blue banner that pops up when it's "congested" - seems to frequently happen after 6pm on weeknights.
My peak download is only 259Mbps, peak upload is 19.3Mbps. Slowest download is 21.7Mbps, slowest upload 1.57Mbps.
1664659312400.png
 
Setup a 4 to 6 tunnel with Hurricane Electric.
I was very involved with the early Internet, etc. Went to the first InterOp conference, "Geeks By The Bay," knew Dave Crocker and Marshall Rose, once met Jon Postel, etc., etc. but that involvement pretty much ended by 1993. Amazing how much progress since then.

(y) Happy to learn about Hurricane Electric.
 
Well, there's now a 1TB cap on Starlink.
Data used between 11p and 7a won't count towards the cap, however (for now).
 
Well, there's now a 1TB cap on Starlink.
Data used between 11p and 7a won't count towards the cap, however (for now).

I don't know that "cap" is accurate.

Starlink said:
Under the Fair Use policy, all Residential customers will receive unlimited data, and will start each month with Priority Access, which means their data usage will be prioritized during times of network congestion.

Customers who exceed 1 TB of data use on a monthly basis (currently < 10% of users) will automatically be switched to Basic Access for the remainder of the billing cycle, which means their data usage will be deprioritized during times of network congestion, resulting in slower speeds.

Data used between 11pm - 7am will not count towards your Priority Access.

Starting today, you can now monitor your data usage on your account page. Read more in Starlink’s Fair Use policy and in the Terms of Service.

Thank you for being an early customer and for your continued support of Starlink!

Starlink Team

It's a disappointing development, no doubt, but compare that to the cap on XFinity (Comcast Residential) for example, where when you exceed the current 1.2TB cap, you are automatically billed $10 for every extra 50GB used (50GB is a pathetically tiny amount, and can add up very fast!). The XFinity cap is more along the lines of what a "cap" usually means among residential internet services. With Starlink, the only thing that happens when you go over your "cap" is that your data will be "deprioritized during times of network congestion". It's unknown what that means in practice. To exactly what extent will speeds be reduced? Will latency increase at that point, even during times of light usage? I will be testing this in the coming months...
 
I don't know that "cap" is accurate.



It's a disappointing development, no doubt, but compare that to the cap on XFinity (Comcast Residential) for example, where when you exceed the current 1.2TB cap, you are automatically billed $10 for every extra 50GB used (50GB is a pathetically tiny amount, and can add up very fast!). The XFinity cap is more along the lines of what a "cap" usually means among residential internet services. With Starlink, the only thing that happens when you go over your "cap" is that your data will be "deprioritized during times of network congestion". It's unknown what that means in practice. To exactly what extent will speeds be reduced? Will latency increase at that point, even during times of light usage? I will be testing this in the coming months...
Well, considering my speeds goes to ~7Mbps down and ~1.5Mbps up during current congestion, any further degradation of that will basically make it unusable for anything but simple web surfing.
 
Well, considering my speeds goes to ~7Mbps down and ~1.5Mbps up during current congestion, any further degradation of that will basically make it unusable for anything but simple web surfing.

My speeds have actually been improving a lot since late summer, and even back then I never saw significant slowdown on my upload; for me it was mainly download that was affected during peak hours, and I think the absolute worst I ever saw on the download was 15Mbps.

Anyone can theorize about what this means, but we won't know until we have some actual test results from people going over 1TB.

I just did a test here in Northern California, at 7pm and got 121Mbps download, 33Mbps upload. I think 7pm qualifies as "peak hours" and it actually gave me a warning about it during the speed test. That's not too bad IMO. I'm not sure why your speeds are so slow.
 
Last edited:
My speeds have actually been improving a lot since late summer, and even back then I never saw significant slowdown on my upload; for me it was mainly download that was affected during peak hours, and I think the absolute worst I ever saw on the download was 15Mbps.

Anyone can theorize about what this means, but we won't know until we have some actual test results from people going over 1TB.

I just did a test here in Northern California, at 7pm and got 121Mbps download, 33Mbps upload. I think 7pm qualifies as "peak hours" and it actually gave me a warning about it during the speed test. That's not too bad IMO. I'm not sure why your speeds are so slow.
It's not just mine, several other folks I know in the area that have it see similar results. Could just be our local cells are overloaded, but, there's no real way to determine where the issue is. Well, there are I guess, but, I don't feel like setting everything up to get that information, especially since it won't help anything :p
 
Sounds like typical ISP shenanigans going on. Offer unbelievable service and speeds, oversell, service and speeds go to crap.

I was really pulling for this to work. Guess only time will tell if things get better for Starlink customers or not.
 
Yeah, Starlink seems to be learning some bad habits from the big ISP's as it figures out how to become profitable.

They recently added a 1 TB data cap in the US and will start throttling users who exceed that cap.

I hope that they don't start adding bogus add-on fees like the "regulatory recovery fee" to their bills as well.
 
I was still hitting 175 down this morning. Despite all the potential shortcomings, Starlink is the best option for us by a very large margin.
 
My client's having issues with hers. I suspect it's the shitty router thats supplied. She's waiting on the ethernet adapter. Internet will just die out of no where even though the light on bottome is solid.
well hell mine does that every 4 minutes it's the sat signal with a mere 7% obstruction
 
Well, Starlink kicked the "data cap" back til April (at earliest). My overall speeds have stabilized lately, though it still does absolutely tank from ~6pm to ~10pm nightly (to like 10-15mbps download and 2mbps upload). Hopefully once we move that improves a bit.
 
Well, Starlink kicked the "data cap" back til April (at earliest). My overall speeds have stabilized lately, though it still does absolutely tank from ~6pm to ~10pm nightly (to like 10-15mbps download and 2mbps upload). Hopefully once we move that improves a bit.

Motherfucker... damn this is getting annoying. This is the 2nd time that they have pushed back the "data cap". Obviously, I'd rather that there was no data cap at all, but this is the 2nd time now where I've purposefully used more than 1TB (in a handful of days) to attempt to determine what "best effort" service (a useless term) actually means. We use considerably more than 1TB per month, so what the service behaves like past that point is going to basically determine if we are able to continue using Starlink or not going forward. Both times, I tested it, and the result was that the service seemed non-impacted, only to find out that they pushed the data cap "back"
rolleyes.gif


StarlinkFeb.jpg

Unfortunately we don't have a lot of time left to make a decision. This month (February), we have to decide whether to renew our cable contract (business w/ no caps) or not. But without knowing what we will end up with using Starlink, it's an impossible decision to make...

Overall, I'd say that Starlink is much worse than when we first got it. Speeds are way down, latency is up, reliability is down. This is especially apparent during the evening hours. When I first got it, it seemed like it would be a real alternative to traditional broadband, but now it's turning into something more like "better than Hughes or DSL" and not much more. I know that there is still hope for when Starlink starts launching the 2.0 satellites, but Internet is too important to bet on something that sucks now and will "hopefully" get better at some point in the future. They have clearly oversold the network at this point, and are obviously using some pretty advanced traffic analysis to slow-down certain types of traffic (torrents, etc).

I wish that there was some easier way to test out what "best effort" service actually means in practice.
 
i dont have any numbers to share, because i wasnt included in the install/testing like i was supposed to be, but we have now deployed starlink to our other two remote colony schools. the network guys seem to be really happy with them and theyre working better than any other rural options theyve tried. the nightly slowdowns wouldnt affect us though, as the schools are closed.
 
I am hearing the same thing from everyone locally that has had starlink since it became available, it was so good at first and now has became no better then xplorenet(hughes) and dsl.

10months ago there was only 15 users on this grid and now theres over 400
 
The good thing about the "data cap" is that it's still free for all from 11pm to 7am (reminds me of free after-hours minutes and texts on cell carriers in the early 00s haha). So if you can schedule your bulk workloads during that time, then you won't end up on the best effort tier.
Unfortunately where I'm about to move to, it's Starlink or ADSL for at least a year while providers expand into the area. I just need it to be good enough to work from home on, and since I'm no longer transporting containers or anything huge, just code, that should be fine. The county's contract with the local providers is supposed to be completed "late 2024", and I've confirmed my house is on the list to be connected. Not certain exactly what connectivity, but the county office said they're putting emphasis on fiber rollout, so, fingers crossed.
I can deal with the evening speed and availability drop... got plenty of things I do besides sit around on the internet, and I don't play competitive online games.
 
I am hearing the same thing from everyone locally that has had starlink since it became available, it was so good at first and now has became no better then xplorenet(hughes) and dsl.

During peak hours, it can get down to near DSL speeds, but latency is still good for gaming. Latency is still almost always <100ms, usually more like 40-60. It used to be more like 20-40ms. But that increase pales in comparison to the 800ms you get on Hughes, which is simply not good enough for online gaming. Hugues data caps are in the 15–100GB range, which again is nothing compared to the Starlink 1TB soft-cap.
 
Best thing about here is that it cant get any worse because the grid allotment is full, If the launch of the 2.0 sats improves speeds i will be taking over a buddys account as his road is getting a fiber expansion in the next 16 months and i will be building on my 50 acres of very rural property.
 
Best thing about here is that it cant get any worse because the grid allotment is full

Starlink has added a bunch of loopholes that still allow people to get service in areas that are "full". You can order mobile/RV service, which isn't tied to any specific address and thus can be used in areas that are already "full". Many people did this even though they fully intended to use it at their home. They also allow people to sign-up for "best effort" service even if their area is already "full". Regular users get "priority" over "best effort" users, but this kind of prioritization only goes so far, and in practice it still slows down the network for everyone to some extent.
 
Starlink has added a bunch of loopholes that still allow people to get service in areas that are "full". You can order mobile/RV service, which isn't tied to any specific address and thus can be used in areas that are already "full". Many people did this even though they fully intended to use it at their home. They also allow people to sign-up for "best effort" service even if their area is already "full". Regular users get "priority" over "best effort" users, but this kind of prioritization only goes so far, and in practice it still slows down the network for everyone to some extent.
They also seem to allow folks with existing service to (one-time) move into congested cells. I know of two cases where addresses were listed for service well into the future (2024), so folks bought in a region that allowed them instant access, then a couple of weeks later they "transfer" into the congested cell with no pushback or anything.
 
Motherfucker... damn this is getting annoying. This is the 2nd time that they have pushed back the "data cap". Obviously, I'd rather that there was no data cap at all, but this is the 2nd time now where I've purposefully used more than 1TB (in a handful of days) to attempt to determine what "best effort" service (a useless term) actually means. We use considerably more than 1TB per month, so what the service behaves like past that point is going to basically determine if we are able to continue using Starlink or not going forward. Both times, I tested it, and the result was that the service seemed non-impacted, only to find out that they pushed the data cap "back" View attachment 546585

View attachment 546584

Unfortunately we don't have a lot of time left to make a decision. This month (February), we have to decide whether to renew our cable contract (business w/ no caps) or not. But without knowing what we will end up with using Starlink, it's an impossible decision to make...

Overall, I'd say that Starlink is much worse than when we first got it. Speeds are way down, latency is up, reliability is down. This is especially apparent during the evening hours. When I first got it, it seemed like it would be a real alternative to traditional broadband, but now it's turning into something more like "better than Hughes or DSL" and not much more. I know that there is still hope for when Starlink starts launching the 2.0 satellites, but Internet is too important to bet on something that sucks now and will "hopefully" get better at some point in the future. They have clearly oversold the network at this point, and are obviously using some pretty advanced traffic analysis to slow-down certain types of traffic (torrents, etc).

I wish that there was some easier way to test out what "best effort" service actually means in practice.
I don't know if you want to consider running them both together for a while, but if you do, the tplink multi-wan router that's $60 should be able to handle both and allow you to see what really works well.
 
Starlink has added a bunch of loopholes that still allow people to get service in areas that are "full". You can order mobile/RV service, which isn't tied to any specific address and thus can be used in areas that are already "full". Many people did this even though they fully intended to use it at their home. They also allow people to sign-up for "best effort" service even if their area is already "full". Regular users get "priority" over "best effort" users, but this kind of prioritization only goes so far, and in practice it still slows down the network for everyone to some extent.
Wasnt sure how that worked, last summer i seen tons of dishes on RVs and a lot of offroad camping setups.

Is there any eta for their new Sat launches?
 
Wasnt sure how that worked, last summer i seen tons of dishes on RVs and a lot of offroad camping setups.

Is there any eta for their new Sat launches?
For gen2, all we have is "soon". There's been a few launches the past couple of months of gen1/1.5, those are still working their way up into correct orbit and should be joining the mesh "soon" as well (likely within a couple of weeks). It's kind of a race between sat deploys and how many customers they onboard to the service, but they're way over indexing on the onboarding part of that.
 
Wasnt sure how that worked, last summer i seen tons of dishes on RVs and a lot of offroad camping setups.

Is there any eta for their new Sat launches?

The holdup on Gen2 is because they are much larger and need to be launched via the new Starship rocket. The Starship rocket is being held up by regulatory hurdles, mainly because a lot of politicians decided that they don't like Elon Musk anymore
rolleyes.gif
 
The holdup on Gen2 is because they are much larger and need to be launched via the new Starship rocket. The Starship rocket is being held up by regulatory hurdles, mainly because a lot of politicians decided that they don't like Elon Musk anymore View attachment 546885
Nah. Since at least August of last year it's been self-induced hold ups. Previous to that there were valid concerns over it which have been rightfully addressed.

Edit: Not August, June of 2022. I got some dates mixed up.
 
Here's hoping they can launch a bunch quickly... what this manages to obfuscate a bit is while there's decent average upload, the mean isn't great, and there's lots of period of low (3-4Mbps) as well. And that outage was stupid long, with no explanation at all - the app kept saying everything was fine.

1677521713901.png
 
Yeah the slowdowns to sub-DSL speeds during peak hours seems to be getting even worse, it's easily noticeable even during casual web browsing at this point. So we really need these new Sats. I'm not sure if these little "mini v2" sats will have the same impact the big V2 sats were supposed to have, but every bit helps.

I will say, we just had a massive snow storm, which was worse than anything we've had in this part of Northern California in probably 30+ years. Usually we get maybe 1-2 days of snow per year where we get these tiny micro-flakes that are barely enough to coat the ground for about 30 minutes before melting. This time we got 14 inches... made the roads undrivable, knocked trees down, and of course knocked the power out. Being able to simply power the Starlink via generator was nice, and speeds were way up since I suppose many people in the area probably weren't using theirs. It was also the first time that I actually got to test out the snow-melt feature.
 
I will say, we just had a massive snow storm, which was worse than anything we've had in this part of Northern California in probably 30+ years. Usually we get maybe 1-2 days of snow per year where we get these tiny micro-flakes that are barely enough to coat the ground for about 30 minutes before melting. This time we got 14 inches... made the roads undrivable, knocked trees down, and of course knocked the power out. Being able to simply power the Starlink via generator was nice, and speeds were way up since I suppose many people in the area probably weren't using theirs. It was also the first time that I actually got to test out the snow-melt feature.
Thank you for the first hand-account on the snow there in CA. I heard about this, but though those were just mistakes in predictions as it sounded way off the charts. But it sounds like it was the real deal. :( Hopefully power doesn't stay off too long as that's not fun when it's cold outside.
 
Thank you for the first hand-account on the snow there in CA. I heard about this, but though those were just mistakes in predictions as it sounded way off the charts. But it sounds like it was the real deal. :( Hopefully power doesn't stay off too long as that's not fun when it's cold outside.
its real, seems to happen about every 40 years or so, last time was 80/81 iirc. almost like we are on a cycle or orbit or something...

It was also the first time that I actually got to test out the snow-melt feature.
how'd that work out? does it just warm the dish and the snow slides off?
 
its real, seems to happen about every 40 years or so, last time was 80/81 iirc. almost like we are on a cycle or orbit or something...


how'd that work out? does it just warm the dish and the snow slides off?
Chiming in from greater Seattle area. Yup, the dish gets warm and after just a little bit, the snow just ends up sliding off of it. Was neat to see. Didn't take too long for it to happen, either... maybe 15 minutes. It had about 4-5" of snow on it. It might not be as effective if the snow doesn't have anywhere to slide to and have to wait for it to all melt (and, then if there's liquid water, it could freeze again). But it seems to work good enough!
 
Chiming in from greater Seattle area. Yup, the dish gets warm and after just a little bit, the snow just ends up sliding off of it. Was neat to see. Didn't take too long for it to happen, either... maybe 15 minutes. It had about 4-5" of snow on it. It might not be as effective if the snow doesn't have anywhere to slide to and have to wait for it to all melt (and, then if there's liquid water, it could freeze again). But it seems to work good enough!
cool, thnx! i havent actually been able to see any of the ones we deployed as all the remote locations are way out in the boonies, in the opposite direction of where i work now.
 
Back
Top