Starlink is better than satellite and DSL but not as fast as fiber or cable | Hacker News Reader

Starlink is better than satellite and DSL but not as fast as fiber or cable

zdnet.com

46 pointsCrankyBear74 comments

They left out T-Mobile Home Internet from the comparison. It's a more than viable option if you're in the right location. You'll usually see faster speeds (about double) and lower latency (closer to standard broadband) than Starlink based on my analysis of a half-dozen recent reviews. It's also much cheaper, $0 up front and $60/mo vs Starlink's $499 up front and $99 a month. Starlink, of course, covers much more remote areas than T-Mobile Home Internet will.

Honestly I'm far more interested on Starlink's impact in the developing / unfree world than in access in rural areas in the US. Not that the latter isn't worth fixing, but it's a more gradual shift there rather than a paradigm shift.

What do you mean by "unfree world"? Is it America's pariahs or something?

I think anywhere with significant censorship, such as mainland china.

Anywhere without free internet access.

Flying some drones or planes every now and then will find most users of unapproved satellite terminals. Satellite internet won't save the world from censorship.

Why would China allow star link to operate in China without following Chinese laws? I don’t see Elon shipping clandestine transmitters to the people of Xinjiang, not if he wants any legitimate customers in China.

Really, star link is a business, they want paying customers-that means complying with local laws.

Besides, censorship is not a uniquely Chinese policy, America seems to be exporting more of it lately.

At least in MN I've consistently had friends and family in VERY rural areas have much faster internet options than I do in a metro suburb.

I suspect it's just much cheaper to run fiber in those areas where there's not many approaches or right of way issues to deal with

I concur that humanitarian aspect of Starlink is by far the most exciting, but fail to see how ability to have a stable fast speed connection anywhere is not a paradigm shift. Especially when you combine it with recent remote work transition.

Specifically if you wanted to start a company or even open an office either your business opportunities were limited to local area or you were locked into places with existing fiber connections. This means high paying work was also concentrated in places with existing fiber connections. I am hoping mega constellations will level economic landscape via increased geographic availability of high paying jobs.

You'd be surprised. I have relatives who live ten minutes from a McDonald's and still have no option for internet besides satellite with a bandwidth cap of a few gigabytes a month. They can't enjoy Netflix, YouTube, or Disney Plus. Their kids can't buy digital download video games. When Starlink drops their base station price a bit and I talk them into it it will be absolutely transformative for them.

I don't mean to suggest this won't affect them but... kids in rural Africa being able to connect to the internet for school / college / business is a lot more interesting to me than people in the rural US being able to consume a different form of television.

My nephew spent half of last year working from a take-home packet because he couldn't do Zoom class. It's not as different as you think.

In many more densly-settled rural areas like mine, outside Redmond WA, not even DSL is an option because Frontier refuses to add more DSL hookup hardware. T-mobile doesn't offer 5-g Internet in my area either. So I'm forced to use ordinary (Verizon) 4G hotspots for Internet.

It would need to get a lot cheaper. $500 setup and $100 per month is not even close to something someone can afford in this regions and those few that can already have internet access.

The actual dish costs like $2000 to $3000, and does not include the cost of rocket launches to get all of those satellites up there.

The $100 per month doesn't seem like it'd be sufficient to keep ~6000 satellites in orbit either (they're orbiting so low, that they'll need to constantly "boost" themselves to stay in their orbit ring).

So they're losing money per dish, and seemingly losing money per satellite launch, and I have my doubts that even 1-million customers at $100 / month or $1200 / year would be sufficient to keep their infrastructure costs (an ISP with only $1.2 Billion/year costs serving a million customers? Even without satellite launches, satellite development, that seems like a tall order. The ground-based infrastructure costs alone are already going to be costly... its not like bandwidth is free even if you're on the ground)

Presumably other satellites and space deliveries offset the cost per launch - and how many service years per satellite? How much are they paying their noc and customer service people, in addition to ground engineering and interconnects to ground stations? I'm guessing they've run the numbers and expect to make money.

Hugesnet has 3 Satellites not 30k like starlink has planned.

> I'm guessing they've run the numbers and expect to make money.

No one really does that in today's economy. Most businesses these days aren't about achieving profits, but are instead about first achieving dominance... and then hoping you achieve profits afterwards.

That's why we keep seeing companies like "Moviepass" pop up.

IIRC, the method of beamforming utilized by their satellites limits ground station density. This is not really an issue at the moment as they are not targeting dense areas.

Once they are 'out of beta', I expect either starlink themselves or a 3rd party to start offering 4g/5g service backed primarily by a ground uplink station. This would allow wireless sharing of one ground station in dense areas. I would imagine that there are many situations where being able to install a cell data station without the need for a physical internet connection would be cheap and effective.

I see speeds of up to 415Mbps down, 65Mbps up, and latency as low as ~16ms. I regularly see 200Mbps down, and I cap the upstream to 20Mbps to help with bufferbloat. Latency is usually around 20-40ms.

That's excellent! I have a friend in a very rural area served by only 5Mb/s download over CenturyLink DSL. This would be perfect for them.

I really hope Starlink gets big. I have no attachment to Musk or Starlink itself, but I hope to live more and more remotely as I get older and the prospect of having decent internet makes the dream feel more attainable (I want to be a hermit but still have reasonable amenities like access to information).

Musk Feb 2021 "Speed will double to ~300Mb/s & latency will drop to ~20ms later this year"

https://twitter.com/elonmusk/status/1363763858121256963?lang...

Again on Jun 2021 https://twitter.com/elonmusk/status/1272363466288820224?lang...

There are plenty of 300 Mbps+ speed tests posted on r/Starlink. But that's peak, vs the average that Speedtest is showing.

Starlink's bandwidth and latency are way past good enough at this point. The issue with Starlink currently is reliability. Many people experience dropouts and jitter that make it unsuitable for video conferencing or gaming. This is often caused by people being unable to meet the requirement that the dish have a continuous unobstructed view of most of the sky, unlike a traditional satellite dish that stays pointed at one tiny section of sky. This will improve as more satellites are launched.

There will be a market for large flag pole-like installations that are lightning resistant and are above the tree line.

Maybe I'm opening a can of worms here, but isn't Elon's Twitter an unreliable source of factual information when it comes to promises about the future capabilities of his companies?

Didn't the SEC have a chat with him about that?

The SEC only had a chat with him because it has to maintain the facade that it does something.

The mortgage crisis showed that the SEC is a toothless revolving door of rubberstampers. And still is.

Depends on the figure type. A specific spec is usually reliable, AI timelines are very unreliable. Product launch dates are inconsistent. Shipment volumes have always been met, but are almost always delayed.

> To no great surprise, Ookla found Starlink beats HughesNet and Viasat handly. The company found that "Starlink was the only satellite internet provider in the United States with fixed-broadband-like latency figures, and median download speeds fast enough to handle most of the needs of modern online life at 97.23 Megabits per second (Mbps) during Q2 2021. HughesNet was a distant second at 19.73 Mbps and Viasat third at 18.13 Mbps."

This is really deceptive. You can actually get really insane speeds on HughesNet/Viasat, the issue at hand is the service profiles available- Both HuguesNet/Viasat have 20Mbps down plans. If starlink goes full release at 100Mbps then sure? But that's highly unlikely unless they implement bandwidth caps or dont sell to the public (or limit subscribers to a tiny percentage?), as satellite bandwidth is extremely constrained (ESPECIALLY for leo constellations).

Where starlink is a clear undisputed winner is latency (not capacity). You'll actually be able to make video calls and play games on a starlink connection.

LEO constellations have way more aggregate bandwidth than GEO satellites

Aggregate yes, but not where you need it all the time! Customers are not evenly distributed across the earth, they are concentrated in regional areas on land. If you look at usable bandwidth (ie, satellites verhead at any point) LEOs drop off substantially.

Also weren't some ISPs caught fudging speedtest.net by specifically making the connection faster when doing a speed test? I remember Comcast doing first x MB bursts which would result in inaccurate fast results.

I can't believe I'm defending Comcast, but it does seem to be a good idea to serve requests with high initial bandwidth since most of them aren't sustained downloads.

That of course doesn't leave out that they'll detect being benchmarked like auto companies do with emissions testing and then switch to fake-out mode.

Viasat also heavily throttles video streams, so as not to allow a user to accidentally reach their monthly data cap in the first day or two.

I currently average 25-30 Mbps on speedtest.net and 700 kbps (not a typo) on fast.com. And still I reach the data cap in two weeks or less. Fortunately (?) it's a soft cap, not a hard one. All it means is no streaming anything in the evening till the month rolls over. (And, perversely, you don't want to do any large downloads until after hitting the cap!)

Hughes and viasat deserve their thrashing by starlink. They turned a technological triumph into rancid consumer exploitation, preying on ignorant satellite TV consumer markets. Source: I used to sell and troubleshoot that crap to people that had no business using anything more technical than a milk carton.

Thats the thing, it isn't going to BE a thrashing. Unless starlink has some way to create capacity from thin air, they'll end up with the same exact resource contention that ViaSat/HughesNet face now. Musk still has to obey the laws of physics- Starlink can (and will) provide a service with excellent usable latency. However they will either need to nearly double prices to maintain 100 MB/s and no bandwidth cap or dramatically reduce service (say, to 20MB/s a subscriber) with a cap at price levels comparable to ViaSat/HughesNet.

There's a concomitant number of ground stations augmenting the network. In my relatively small town starlink has 300gbps via fiber serving two base stations.

They're doing the infrastructure right with the aim of $100 gigabit internet service. Which Hughes and viasat could have pulled off if they'd been able to reach the same orbits, volume of satellites at the right price point, and been willing to invest in our take advantage of infrastructure on the ground.

The pay per byte model is fraudulent, whether your connection is satellite or coax or fiber. Once that is embraced, networks stagnate. Because established satellite isps are universally scumbags, I'm very glad starlink is breaking them. They'll have to compete with an unmetered service model, or die out.

It's interesting to me that they're using Ookla (speedtest.net) data for this analysis. For two reasons:

First, I know my old Comcast cable internet used to vary widely - at 7am, I could get 80 Mbps, but at 7pm, I was lucky to be able to download the whole speedtest.net page at all. I rarely would want to run a test when it was fast, but as people started to sign on and it degraded I think the natural impulse is to do a speed test.

Second, other providers like Google, Netflix, and ISPs offer integrated speed tests, so there's likely some selection bias for people who will still go to speedtest.net.

Note - I rejected my Comcast cable in favor of T-Mobile home internet, it's cheaper and more reliable. Which really says something about Comcast mismanagement of their resource, they've got a coaxial connection to my house but T-Mobile can do better wirelessly...

It sounds like you had a bad install. I had multiple techs come out back when I had Comcast with similar performance issues, and finally one of them took the time to diagnose and fix some wiring issue at the pole. All the flakiness disappeared immediately.

Consider this, you don't have to be a customer of Comcast, HughesNet, or AT&T anymore.

Yea but I’m not sure a Musk company is that much better. Tesla has lots of promises over the years about FSD combined with pricing swings.

I pay Comcast $100 for a Gigabit that’s reliable for voice calls. Looked into Starlink but couldn’t justify the download/upload speeds for the price or the loss of service.

Not sure why a lot of people say it’s for the country. I’m barely outside of a major metro area and was eligible for it a while back.

I literally hate Comcast.

You have no idea how bad it hurts me to use them. AT&T only offers crappy DSL at my place. They wired my neighborhood up for fiber and stopped 5 houses in each direction. They want thousands of dollars to complete the line to my house and the neighbors in between. My last house had AT&T fiber with symmetrical gigabit. Comcast is 1Gb up and 50Mb down, good luck ever seeing 50Mb and it used to be 25Mb max.

At my families vacation house Comcast is the only option. It was the same when I lived in apartments. They would make deals with the complex to be the sole provider.

Comcast is awful. We have comcast in the bay area, but we get constant outages throughout the day making the internet service extremely unreliable. AT&T DSL isn't a real contender to Comcast cable, so Comcast is the only realistic option.

If I could switch off Comcast (to another cable internet provider or fiber), I would do so in a heartbeat. Comcast has been a consistent, horrible monopoly.

Have there been total addressable market numbers published for starlink in US / Canada? Like lets say there are maybe 20% of people in US that starlink could address -- internet service typically services a household, so unclear what the math is there, but maybe 3 people per household, so total addressable market of say 350 M * 0.2 / 3 = 23 million households.. so a quick search on subscriber number for Hughesnet is 1M subs.. so maybe a target for SpaceX would be to get 10M subscribers? 10M of the access terminal things if they cost $500 each would be 5 billion just to pay for those.. $100 / mo for 10M people is 1B per month -- so seems like the target for subscribers world wide must be closer to 100M otherwise the economics don't seem too great...

Any current TAM would be leaving off the amount of people who live in cities and can now exit to live "off-grid" with tesla solar and starlink. I anticipate at least a few "we should live in a commune" groups making good on their longstanding threats.

My starlink connection is averaging about 0.08% packet loss to its terrestrial gateway (first hop of real world ARIN IP on the outside of the cgnat) over a period of 30 hours. I have a tree obstruction in 1/12th of the measurement system it uses.

The DOCSIS3 cable modem connection from the local last mile provider at the same location is doing much worse. Up to 1.5% over 3 hour periods and 1.2 to 3.5% over 30 hour periods.

This is as measured by a fairly rudimentary smokeping installation that's set to 60s intervals, a widely varied set of targets, and 20 'fping' per measurement.

"Starlink's median latency, 45 milliseconds (ms) is close to fixed broadband's 14 ms. " How is 3 times as much 'close'?

It's close compared to latencies over traditional sattelite internet where a ping of over 400ms is not uncommon. The article quotes a latency if over 600 or 700ms. At least 45 is in the same order of magnitude of common broadband.

45ms reminds me more of 3G than of any normal internet connection, but that's still much more useful than the latencies the sattelite competition offers.

We’re talking about humans here. A human might notice 45ms in a shooter, but not when just surfing the web. Compare that with 500-1000ms latency from older forms of satellite service, or 250ms from old dial–up service.

>A human might notice 45ms in a shooter, but not when just surfing the web.

You are forgetting round-trip request latency, handshakes, and dependent resources (don't know what to fetch until receiving the first thing, then the second level of things, etc., this can stack to multiple seconds).

High latency is very bad for browsing. The main things it is fine for are streaming audio/video (non-interactive), large downloads, and backing up/uploading files.

You’re not wrong, but not relevant either. If your RTT to the server is 15ms, then connecting to that server with TLS1.2 will take 4×15ms = 60ms. On Starlink that will currently take 4×45ms = 180ms. Yes, 180ms is three times 60ms, but to a human that is irrelevant. Humans can’t detect time intervals of less than 100ms, and can only barely detect intervals less than 200ms. In a direct comparison between the two, humans will rate the 60ms connection as instantaneous and the slower 180ms connection as nearly instantaneous. This makes them “close” as the article described, especially when compared to old–school satellite links where the RTT can be insane.

And the handshake time doesn’t compound; there is no need to open a new connection to the server for every resource needed by a webpage. The web browser just keeps reusing the same connection until it runs out of resources to fetch from the server.

> And the handshake time doesn’t compound; there is no need to open a new connection to the server for every resource needed by a webpage. The web browser just keeps reusing the same connection until it runs out of resources to fetch from the server.

Third party analytics and other resources often means it compounds even for handshaking, but dependent resources can compound too even on the same server even if they don't each need a handshake.

>By comparison, Viasat, 630 ms, and HughesNet, 724 ms, are almost unusable for these purposes.

They're saying it's close compared to starlink's competitors which aren't even the same order of magnitude. In context I think it's a fair assessment.

Yea a buddy tried out HughesNet and the latency was 1~1.5 seconds. Loading anything was death.

They are also thousands of miles further out. I believe hugesnet services almost the entire globe with just 3 satellites.

Who is getting 14ms on their fixed broadband. Last 3 places I can remember, I had cable and fiber to the home. My two cable connections would get 25-35ms at best. The fiber to the home wasn't much different but did go sub 20ms on occasion.

I am on Google Fiber and I get < 10 ms to anything served out of a CDN or to servers in my own state. If I ping google, reddit, amazon, etc they are all about 8 ms.

AT&T FTTH in Texas. Its 2-3ms from my home to Google's CDNs. 2-3ms to Cloudfront. 4ms to akamai. Its 3ms to my friend's house, 1ms to my work's office. Its 2ms to my colo in town.

40ms is me essentially connecting halfway across the country. A Digital Ocean droplet in SF or NY is ~40ms.

How are you measuring this? Because I just ran a ping from my colo server in a datacenter and am seeing 11-12 ms when I ping google CDN.

All of this was measured on my residential FTTH internet from a wired desktop just using the Windows ping utility.

11-12ms to a Google CDN from your colo? That's insane to me. Pinging from my colo to Google using the ping utility from iputils for 50 packets gives an average ping of 1.6ms. Akamai pings at 2.1ms.

Ok, so I ran a ping again from my data center server. And for CDN.google.com, I get average of less than 0.2ms ping. But, when pinging www.gstatic.com, that is when I get more like 7-10ms.

I'm not sure what the difference is between the two.

Just ran a speedtest on my phone; I get 19ms on a Farmer's Telecom fiber. 60-70ms on the work VPN.

The last time I had >20ms RTT "Fast Path ADSL" was still a thing.

Depends on where you live and where you measure to. Generally, it's fair to measure to the first IP hop, if you live far from any major internet exchange, that's going to add to your latency regardless.

I've seen fast-mode DSL around 7ms, interleaved DSL around 20 ms, cable usually 5-10 ms, but highly variable. ATT GPON was 5 ms which I thought was pretty high, although fiber should really be sub 1 ms.

I've seen LTE between 10 and 60 ms (or worse, when it's congested), older technologies are usually quite a bit more.

How are you measuring? Everyone is saying they get such low latency, I just ran a ping to 8.8.8.8 and get the following on my colo server in a Las Vegas data center:

PING 8.8.8.8 (8.8.8.8) 56(84) bytes of data. 64 bytes from 8.8.8.8: icmp_seq=1 ttl=115 time=12.7 ms 64 bytes from 8.8.8.8: icmp_seq=2 ttl=115 time=11.4 ms 64 bytes from 8.8.8.8: icmp_seq=3 ttl=115 time=10.6 ms 64 bytes from 8.8.8.8: icmp_seq=4 ttl=115 time=10.9 ms

I get about the same from my home cable internet connection.

Our rural fixed wireless has 5-10ms jitter, but 14ms (up to 25ish ms cause jitter) is easily observed when pinging CloudFlare via the KCIX exchange (where our ISP sends most of it's traffic).

Xfinity, Silicon Valley:

  --8<--

$ ping 8.8.8.8 PING 8.8.8.8 (8.8.8.8) 56(84) bytes of data. 64 bytes from 8.8.8.8: icmp_seq=1 ttl=117 time=14.7 ms 64 bytes from 8.8.8.8: icmp_seq=2 ttl=117 time=15.8 ms 64 bytes from 8.8.8.8: icmp_seq=3 ttl=117 time=14.8 ms ^C --- 8.8.8.8 ping statistics --- 3 packets transmitted, 3 received, 0% packet loss, time 2002ms rtt min/avg/max/mdev = 14.774/15.162/15.828/0.473 ms

-->8--

Download Speeds:

Alabama: 168 Mbps

Indiana: 68 Mbps

Canada: 87 Mbps

France: 140 Mbps

Germany: 108 Mbps

The wide range of service is due to the full constellation not being deployed yet? Once fully deployed, will service become more consistent?

I'd expect to continue to see regional variability based on density of subscribers, latitude, and quality of nearby ground stations.

Service should become faster in general though based on SpaceX statements, and based on the fact that they'll have a lot more bandwidth available with more satellites in the sky.

I'd imagine there will still be regional differences. Even with satellite-to-satellite links they're still going to want to route your data to the nearest ground station as fast as possible, so your performance is going to be related to how good your nearest ground station is, and how congested it is.

Great Title. Saved me a click. Thanks.

But seriously, if I was asked a multiple choice question of where Starlink would fall, this probably would have been my guess.

As others point out, it's really the cost/benefit trade-off that will matter.

I want to see a latency distribution, not summary stats.

Reply on news.ycombinator.com