Leaked SpaceX Starlink Speedtests Reveal Download Speeds of 11 to 60Mbps (arstechnica.com) 84
Some leaked speedtests from beta users of SpaceX's Starlink satellite-broadband service "aren't showing the gigabit speeds SpaceX teased," writes Ars Technica, "but it's early."
Long-time Slashdot reader AmiMoJo shared their report:
Beta users of SpaceX's Starlink satellite-broadband service are getting download speeds ranging from 11Mbps to 60Mbps, according to tests conducted using Ookla's speedtest.net tool. Speed tests showed upload speeds ranging from 5Mbps to 18Mbps. The same tests, conducted over the past two weeks, showed latencies or ping rates ranging from 31ms to 94ms. This isn't a comprehensive study of Starlink speeds and latency, so it's not clear whether this is what Internet users should expect once Starlink satellites are fully deployed and the service reaches commercial availability....
Links to 11 anonymized speed tests by Starlink users were posted by a Reddit user yesterday... A new Reddit post listing more speed tests shows some Starlink users getting even lower latency of 21ms and 20ms.
Beta testers must sign non-disclosure agreements, so these speed tests might be one of the only glimpses we get of real-world performance during the trials. SpaceX has told the Federal Communications Commission that Starlink would eventually hit gigabit speeds, saying in its 2016 application to the FCC that "once fully optimized through the Final Deployment, the system will be able to provide high bandwidth (up to 1Gbps per user), low latency broadband services for consumers and businesses in the US and globally." SpaceX has launched about 600 satellites so far and has FCC permission to launch nearly 12,000.
While 60Mbps isn't a gigabit, it's on par with some of the lower cable speed tiers and is much higher than speeds offered by many DSL services in the rural areas where SpaceX is likely to see plenty of interest.
Links to 11 anonymized speed tests by Starlink users were posted by a Reddit user yesterday... A new Reddit post listing more speed tests shows some Starlink users getting even lower latency of 21ms and 20ms.
Beta testers must sign non-disclosure agreements, so these speed tests might be one of the only glimpses we get of real-world performance during the trials. SpaceX has told the Federal Communications Commission that Starlink would eventually hit gigabit speeds, saying in its 2016 application to the FCC that "once fully optimized through the Final Deployment, the system will be able to provide high bandwidth (up to 1Gbps per user), low latency broadband services for consumers and businesses in the US and globally." SpaceX has launched about 600 satellites so far and has FCC permission to launch nearly 12,000.
While 60Mbps isn't a gigabit, it's on par with some of the lower cable speed tiers and is much higher than speeds offered by many DSL services in the rural areas where SpaceX is likely to see plenty of interest.
I WANT (Score:4, Informative)
11Mbps to 60Mbps? Give me that amount of terrible!
Re: (Score:3)
It's decent, and they're just getting started. I'm sure the sats are capable of faster to a single link, so I bet they're testing moderate saturation. They have a small constellation at this point, so even with relatively sparse beta customers they could be saturating individual sats, probably deliberately.
The latency is more interesting. Plenty of Slashdotters have insisted that latencies under 100ms would be physically impossible. Not so much. 50ms isn't great, though. Wonder if they'll improve that
Re: (Score:2)
Latency doesn't worry me so much, as long as it's not Hughes Satellite-level bad. Our corporate network has ping times over 100 ms to a lot of sites and that's certainly workable.
Re:I WANT (Score:4, Informative)
It's also worth mentioning that the initial test constellation is 550km and Ka/Ku band. However, by far, most Starlink satellites will be at around 340km (2/3rds the SNR), and Starlink will be expanding from Ka/Ku, to V, and even W band. That said, in adverse weather, only lower frequency bands will be available (W-band won't even go through thin clouds).
I'm not sure what other bandwidth limitations the test constellation may have. I'm pretty sure that at this point they don't have the optical links between satellites; it's only up and down to ground stations, and the number of ground stations is currently limited.
It's also a question as to where the user(s) in question were. In densely populated areas, you end up sharing the same patch sky (and thus bandwidth) with a large number of other users. In sparsely populated areas, you share it with few people, or even none. I also have no clue how big the beta is, to compare the number of users vs. the number of satellites which cross their locations, which is of course a critical factor.
Re: (Score:1)
Re: (Score:2)
It's a lot like mobile networks. Highly dependent on location, number of nearby users, available backhaul, signal quality and local interference etc.
Re: (Score:2)
I would think that in satellite links the bottleneck is the SNR in your downlink. If you can make the link between satellites using highly directional antennas the throughput between the satellites is going to be good. The satellites should know each other's positions, so using a sufficiently large phased array they should be able to point pencil beams at each other. I imagine that they probably cannot to muc beamforming in the downlink, and this will hurt the SNR and thus the throughput.
Re: (Score:2)
Re:I WANT (Score:4, Insightful)
11Mbps to 60Mbps? Give me that amount of terrible!
Especially in those rural areas where your choice is distant DSL that barely registers on the speed scale at all and flakey relayed wireless service. In my town there are movie star compounds that only have these choices.
Re: I WANT (Score:1)
Re: (Score:3)
You gotta crank up the radiation! (Score:3)
But testing is how things get better, this speedtest 'leak' is pointless.
Decent enough (Score:5, Insightful)
I'm looking forward to what they can do in the future.
Re:Decent enough (Score:5, Informative)
Yep, it beats the snot out of Exede on Viasat, which struggles to deliver 6Mbps even at non-peak hours.
Re: (Score:2)
It's what we expected, better than crappy rural broadband and what backwaters like the UK get in many places, but nothing like the hype from Musk.
It will raise the bar for crap internet but not complete with fibre.
Re:Decent enough (Score:4, Interesting)
I honestly will be happy with 10-60 Mbps if they can deliver on latency. I want interactivity, and I don't care much about HD video streaming. I'm happy with ye olde 480p for almost all content. But the latency on Exede is both massive, and variable.
Re: (Score:2)
Latency seems highly variable at the moment, between 20 and 100ms. Sub 30 is generally seen as reasonable for gaming, it's two frames at 60 FPS, but ideally you want sub 10ms.
Low latency is becoming mandatory for some games, e.g. Mario Maker 2. People with high latency tend to get on people's block lists. That's why this is a bit disappointing - the latency seen is just about workable now for some things, but every day it's getting less and less usable.
I think latency will probably vary with the position of
Re:Decent enough (Score:5, Interesting)
Keep in mind for rural areas, running that small fibre optic cable in a conduit along side a public road, is a small fraction of the cost of that public road. My GOD how did the country not go broke ten times over doing roads. Cost is a lie, every public road should have fibre optic running down one side, every single public road.
Re: (Score:3)
By ignoring maintenance and end-of-life reconstruction costs, and by depending on future growth to pay for yesterday's roads. It's truly a massive ponzi scheme. If the population ever stopped growing, our infrastructure would literally fall apart.
Re: (Score:2)
Re: (Score:2)
I suppose many politicians saw the value for a society when it comes to roads. Though there are exceptions, when I hear my countrymen from Romania for example tell me that various places only got asphalt roads in the last decade. I won
Re: (Score:2)
Another option is telegraph poles. In some countries they "unbundled" them so that other companies can get access to them, meaning fibre can be slung along them cheaply and accessed everywhere that has phone service.
Another option is just to require whoever owns the phone network to replace it with fibre and allow other companies to access it (for a fee). Or just nationalize it like other critical infrastructure such as roads.
That is an improvement (Score:1)
Even the low end of that is substantially faster than the fasted option I've been able to find for my mom, who lives in a rural area...
I'm really hoping I can get in on the beta before too long.
Exactly as expected? (Score:3, Interesting)
SpaceX has told the Federal Communications Commission that Starlink would eventually hit gigabit speeds, saying in its 2016 application to the FCC that "once fully optimized through the Final Deployment, the system will be able to provide high bandwidth (up to 1Gbps per user), low latency broadband services for consumers and businesses in the US and globally." SpaceX has launched about 600 satellites so far and has FCC permission to launch nearly 12,000.
600/12000 is 5%, and 5% of 1 Gbps is 50 Mbps.
If bandwidth scales near linearly with the number of satellites, links, bands and so forth (as big an assumption as that is) it may be exactly as expected given the partial deployment.
Re: (Score:3, Informative)
Re: (Score:1)
Re: (Score:2)
Perhaps. It depends upon the particularities of the way the network is constructed. I do not believe it is nearly so simple as you've assumed.
Not really, in satellite communications, transponder capacity is your limiting factor. You can have more transponders, but at some point you'll run out of frequencies. It's similar to a cell network, from that perspective.
You're also somewhat limited by the ground stations, and their connectivity. You can have 400G in transponders, but if the ground gateway only has 1G, you're toast.
Re: (Score:2)
600/12000 is 5%, and 5% of 1 Gbps is 50 Mbps.
So does this mean that a Starlink subscriber will be downloading from 20 satellites at once in order to get to 1Gbps? And none of these initial satellites are in a position to team up for a given test user?
Re: (Score:1)
I would assume that the web of satellites may transmit data in a peer-to-peer like arrangement. Packets are routed via internet backbones in a much more direct way, but once you get down to ISP infrastructure levels things look a lot more similar to a genuine web.
Don't ask me though, I only noticed the obvious similarity of magnitude in the ratios. I don't know much more than a layperson about such things.
Re: Exactly as expected? (Score:2)
As of today, no - itâ(TM)s just up to sat, down to ground station. However, they are planning on putting laser links on future versions of the sats to stream data in a web above us.
Re: (Score:2)
Re: (Score:3)
It's the old "up to" con. How about I post then "up to" a gigabuck for their service?
This could get interesting. In some parts of the world they can only advertise the average speed, not the theoretical maximum.
Re: (Score:3)
Re: (Score:3)
Re: (Score:2)
The overall network capacity but each satellite still has a limited amount of bandwidth and a limited number of users it can support. Maybe they can start doubling them up in some areas with heavy use but at best reaction times will be slow since they have to launch another satellite. A mobile network operator can just deploy some temporary masts.
It's comparable to Comcast's baseline (Score:2)
At the higher end of these results, anyway. I pretty consistently get 80 down, 6 up from Comcast.
Yeah, adding satellites will probably increase speed, but adding customers will likely counter that. So I guess it'll come down to price...
Re: (Score:3)
And service. I somehow doubt that Starlink's techsupport will ever rival Comcast for sheer awfulness.
Re:It's comparable to Comcast's baseline (Score:4, Insightful)
So just like other ISPs then (Score:3)
Re: (Score:3)
Yes, Starlink was clearly a waste of time. Let's destroy all the satellites. The US telcos can be relied upon to give cheap, fast service to rural areas /s
Re: (Score:2)
Re: (Score:2)
" and there's never a guaranteed minimum
Oh, there's a GUARANTEED MINIMUM all right -- 0 -- they just don't want to come out and say that.
Now if you get something less than that, you REALLY need to switch to a different carrier, even if it's a pigeon.
Re: (Score:2)
In some countries they are required to advertise the speed you are likely to get and have minimum speed guarantees. Rules generally only apply to fixed line operators though, for mobile service it's so dependent on location that they just let you cancel your contract in the first month or three if it doesn't work out.
When advertising the ISP is only allowed to show the average speed its users get and before you sign up it has to offer a personal speed estimate based on speeds in your area and if known the q
Re: (Score:2)
Putting lipstick on the pig (Score:2)
That may be a big improvement over the piss poor internet access that rural folks have had for decades the number of people who will be benefiting from this speed boost won't be very large. Are there really enough rural subscribers to pay for all the satellites, launches, and space junk mitigat
Re:Putting lipstick on the pig (Score:5, Interesting)
Are there really enough rural subscribers to pay for all the satellites, launches, and space junk mitigation?
Yes. Right now Starlink is authorized by the FCC to operate 5 million ground stations in the US. If you spitball the subscription cost of $80/month you get $400M/month gross revenue in the US alone. The current best guesswork is a launch with 60 satellites costs <$50 million ($250K/satellite + $30M launch cost.) There aren't any high costs associated with junk mitigation; the satellites disintegrate on reentry and due to the low orbit they inevitably reenter, even without guidance.
There are likely between 15 and 25 million rural households in the US alone that are potential subscribers. So yeah, Starlink will work, even without considering the rest of the planet.
Re: (Score:2)
Are there really enough rural subscribers to pay for all the satellites, launches, and space junk mitigation?
I am not a US citizen, but I was under the impression that 10s or 100s of millions of people in the US could accurately describe themselves as "rural subscribers"?
Re: (Score:2)
20% of 330 million is a decent working estimate.
There is some fuzziness to that. Where I live there is no cable TV service ( or over the air TV) but I do have fiber-optic to the house from the local power co-operative.
If you google Grant County WAshington ZIPP you can see what we have.
Re: (Score:1)
Don't forget all the non-rural customers who will switch to Starlink just because they hate whoever they are currently with.
latency is excellent, i think (Score:2)
i get 600 and 700ms round trip times, over xplornet satellite, pinging google.com
Re: (Score:2)
Re: (Score:2)
A small amount of jitter is usually OK, even if it's a large percentage of the RTT. A large amount will screw you, though.
Re: (Score:2)
however the satellite orbits are constant (enough) (Score:1)
the path length is not constant, however the latency delta is a function of the orbits of the satellites currently carrying the user signal, and should be easy to calculate.
i presume this would depend on starlink being smart enough make and report the calculations.
The problem is Latency (or RTT) (Score:3)
Or ping times for the unninitiated.
The way Ookola measures is to send a few pings and report the lowest one .
Normal methodology, including the FCC is to send a predetermined number of pings, at predetermined intervals and average the values, as long as those meet a certain criteria for std. deviation.
Some of those ping values are close to 100ms. If latency is avobe 100ms, starlink looses access to certain subsidies.
The speed of the electromagnetic wavefront from earth to the satellites and back will forever be the same, but as more and more people join, the delay introduced by the network equipment (specially routers) will grow.
Besides, many people forget, but there is a relationship between bandwidth and delay (and noise), so, even if you have gigabits of bandwidth, if the delay (and noise) are too much, you may not actually be able to use all your super-duper bandwidth in the first place. (have it on paper magazine. Link to buy PDF and support the IEEE: https://ieeexplore.ieee.org/do... [ieee.org] )
Only time will tell if starlink will be able to hit its goals of gigabit service with >100ms latency. I hope they do, for this would be a great solution for many countries, mine (Venezuela) included, but I think Musks optimism is misplaced this time (at least in the delay part).
Re: (Score:2)
It's a shame no-one on Reddit thought to just do a basic ping test, let alone use some more advanced tools, to see how consistent latency is.
How does it scale with more users? (Score:4, Insightful)
If a satellite handles 1000 users by splitting 11-60 Mbps among all of them, then not so great. Technically, Sprint's LTE towers are just as good as Verizon's. But Sprint resells service to so many MVNOs that their towers are heavily overloaded, and the speed each individual user sees sucks compared to Verizon's.
Re: (Score:2)
Raw drop in latency (Score:2)
Correct me if I'm wrong but the only advantage over people that already have fiber was in the latency drop, since the refractive index doesn't come into play for a vacuum. In which case it was only going to be at most a 33% drop in latency times anyway, but this LEO setup still requires base stations to connect with and translations to take place.
The reddit post with the 20ms round trip from seattle and users from LA, only 1500km away are getting 90ms, it seems incredibly disappointing for those of us wanti
Re: (Score:3)
Sub-100ms is extremely difficult. The speed of light itself means that the fastest circle of the world would be 133ms. So "the other side" of the world is easily 66.5ms away before you do ANYTHING else. That's literally the fastest you can traverse to the other side of the world, in the most optimal route (short of cutting through the Earth).
Add in routing, local connections, transmissions, conversions, different technlogies, etc.
You're not really going to see sub-100ms to everywhere on the globe in our
Re: (Score:3)
The statement that "latencies haven't really decreased much since the end of the modem era" is not entirely true for home users. The latency of broadband has also improved over the years. Newer generations of cable modems have less latency than older ones because of protocol improvements, which are a bigger factor than higher speed, and all of the fiber modems that have been deployed by FTTH systems like FIOS and Google Fiber have low latency.
The latency of cellular broadband has also decreased substantiall
Re: (Score:2)
Well most server are hosted in either the UK or the US, which are as the crow flies 15km away from aus, or being generous about 20km, half the rough circumference of the earth. I still get about 250ms to the US or 370ms average to the UK.
Cutting that time down to sub 100ms seemed possible to me at one point. These results say otherwise, at least for the time being.
and what are the FAP / CAPs like? (Score:2)
and what are the FAP / CAPs like?
I'd be interested to know (Score:2)
Re: (Score:2)
for now (Score:2)
speeds really don't really mean much right now. when they start getting subscribers, the speeds should go down because of heavier traffic.
Three points: (Score:2)
[1] These are 1st gen satellites and ground stations, with 1st gen software and firmware in BETA test (I'll come back to that in a sec...)
[2] This is with only a small fraction of the constellation in orbit.
[3] Musk is also working on his Starship rockets, which will be fully reusable and massive - able to lift HIS satellites in huge batches for essentially the cost of LOX and Methane. As a result, he'll be able to replace the current 1st gen satellites and add more at his leisure. If he wants to replace ce
Re: (Score:2)
Can't beat physics (Score:2)
There is always going to be a delay to shoot a signal hundred of miles up and then back down so that's something that unfortunately isn't going to go away. It's physics that dictate that a wired link will always work better than a wireless link. As long as StarLink can offer inexpensive and "reasonable" service to the rural community it will be welcomed.
Question for Musk fans (Score:1)
Better than what is available! (Score:2)
Latency should improve over time (Score:2)