31 thoughts on “5 Truths about Comcast’s super modem”

  1. I think your analysis nails it, Om. I have no idea if we’ll ever get FIOS in my nabe in San Francisco, but the second it arrives, I’m dumping our DSL. If the TV package Verizon offers is a better value than DirecTV, I’ll happily dump that too.

    I’ll never give a dime to Comcast but I’m sure there are tons of Comcast subscribers who will be thinking like me when they have a FIOS/TV alternative.

  2. Hey Om –

    Couple of comments to your post.

    First point/Downstream rates – 160 Mbps is the minimum specified (4 x 6 MHz channels). A shared gig is possible (if demand was there) now and Cisco is one of several CMTS vendors that could support this.

    Second point/Consumer speeds – Again, 4 channels is the minimum per bonding group. There are no technolgy hurdles to deploying more than one bonding group per node. If consumer demand supports this the technology supports this.

    Fourth point/Availability – modems will begin shipping in Q1 2008. In some cases, they may be pre-deployed (which means that the CMTS is scheduled for upgrade).


  3. Peter,

    I did mention the 1 gig shared upgrade path. The issue is when.

    on the more channels being available, it is going to be tough since they need all the bandwidth they can get for HD. I pointed that out in my previous post.

    The DBS guys are pressing their advantage. Echostar released their numbers earlier this week. that says it all.

  4. The intuition says that just comparing the access speeds between multiple access cable modem access and FiOS gives any indication on the user experience. Doesn’t it depend on the concentration placed by either system upstream? Isn’t it possible that even though Comcast puts less concentration upstream and FiOS has higher concentration so much so that the user experience is a flip of what you are suggesting? Given that Internet is a statistical multiplexing system, I would prefer to see a metric that reflects that statistical nature rather than a fixed deterministic value.

    Also I do not see the benefit of ever increasing access speed without a corresponding increase in the core. Currently users see the difference because the access speeds are low compared to the core speeds. But eventually I suspect that it will not matter.

  5. Hi –

    Your point on total capacity is the real issue. I think that D3 is only one of two necessary technologies needed to achieve “spectral happiness”. The other being SDV … and this is now a bit clouded thanks in large part to Rep. Anna Eshoo’s misguided concerns/comments at yesterdays hearing on the hill.

    By the way, I am a huge DISH fan …


  6. The reason for the pre-announcement might be Comcast’s recognition that it is going to end up in the bandwidth and not the content business. Hopeful sign is the quote from Roberts: “What consumers actually do with all this speed is up to the imagination of the entrepreneurs of tomorrow.”

    I agree with your skepticism on the actually obtainable speeds on the shared backbone.

    Posted more on all this http://blog.tomevslin.com/2007/05/wideband_via_ca.html

  7. I think the Omster has it on this one.

    I think ‘big cable’ has a big problem on their hands with FiOS on two fronts.

    1. the technology is more well thought out in my opinion.

    2. It already has SIGNIFICANT market ‘mind share penetration’ with the main stream user. Most casual users are aware of FiOS and think of it as the market leader in terms of performance bar none.

    I think cable figures they have to do something from a marketing perspective before its too late.

  8. I believe that the main audience here was Wall Street. Comcast needs to show that it can provide a service similar to Verizon’s without a major plant upgrade. Cable spectrum limitations are real, and much of the activity in The Cable Show was concentrated around BW saving technologies. Spectrum availability will be the main limitation to the speedy adoption of DOCSIS 3.0.

    In the Silicon Valley area, cable modem speeds are 8~12Mbps throughout the day. This comes on a single 38.8Mbps channel that is shared between ~750 homes (oversubscription statistics simply work). Scaling this oversubscription to 100Mbps requires ~50MHz of spectrum. However, since the BW heavy applications don’t yet exist, and not everyone will sign up to a more expensive service, I bet that it could be done with a 24MHz addition. But were can you find 24MHz in a crowded to the rim 750MHz plant?

    It is very likely that the first MSO to deploy this technology on a large scale will be COX, and sooner than you may think. Up until 2.5 years ago, Cox was constantly telling Wall Street how a 750MHz plant will suffice forever. Then Cox went private, and now it’s in the midst of a capital intensive 1GHz plant upgrade. 250MHz of new spectrum provide 1.6Gbps for every node, and a little of this can be allocated for DOCSIS 3.0. My guess is that the first to enjoy the new cable modem speeds will be the lucky residents in Cox’s flagship plant in San Diego. Alternatively, Cox can start with an area which overlaps Verizon’s FIOS deployment.

  9. Brian Roberts is an idiot. I remember being at a show 2 years ago and he said “I didn’t know what IP stood for so I called up Bill Gates to ask him.” Gates must’ve been chuckling after that call.

    Here’s the deal Brian: You can channel bond all you want but you only have 850 MHz to play with. You want to give people 150 Mb/s – well you better forget about delivering any linear channels or any VoD because your entire network will be devoted to broadband.

  10. The speeds are shared between multiple homes. However, that’s not a change from the current cable modem standard. The difference in DOCSIS 3.0 is that you’re sharing at least 4 channels, instead of a single channel in DOCSIS 2.0.

    I do think that cable operators will be able to implement DOCSIS 3.0 at the beginning of next year. That’s when the equipment is set to be certified.

    I’ve already heard of a Singapore cable company that’s implemented Pre-DOCSIS 3.0 (for 100 Mbps service). And I believe Pre-DOCSIS 3.0 is pretty much the same thing as DOCSIS 3.0 (I have a feeling the only difference between Pre-DOCSIS 3.0 and DOCSIS 3.0 is that the equipment just hasn’t been certified as DOCSIS 3.0 equipment).

    Now, whether any US cable company chooses to implement DOCSIS 3.0 at the beginning of next year, that’s another story.

    Actually, I don’t think they need to find 4 channels to do DOCSIS 3.0. They’re already using one channel to do DOCSIS 2.0, so to do DOCSIS 3.0, I think they need to find 3 channels.

    Typically, cable systems today have something like 120-130 channels available for use. Given all the crap that’s on TV these days, I bet I could easily find 3 channels that deserve getting the ax. That being said, I am not familiar with the agreements that cable companies make with content providers, so I’m not sure how easy it is to actually terminate a channel.

    And about Verizon. You have to remember that Verizon is only rolling out FIOS over their existing plant. Most of the US is served by AT&T, who is doing their U-Verse thing (which is a joke). And there’s also Qwest, who doesn’t seem to have a plan at this point.

    Given all that, I’m willing to bet that the only areas to get DOCSIS 3.0 in the near future, are the areas where FIOS is available.

    Also, you have to remember that HD is a problem for everybody. I’m not sure that Verizon, at least with its BPON architecture (which provides 622 Mbps per wavelength), is in any better of a position as far as HD goes. And, of course, AT&T and Qwest are basically nowhere on HD.

    But to answer your question on where cable will find the channels for DOCSIS 3.0, I have a feeling they’re going to try to implement SDV (switched digital video). I think Cablevision has already done some of this. Supposedly, you can get about a 40% reduction in bandwidth usage.

    FYI: other options for for getting the channels for DOCSIS 3.0 are: moving from MPEG-2 to MPEG-4, moving to an all digital television broadcast system, and upgrading the plant from 860 MHz amps to 1 Gig amps.

    And one last thing, its not like you can get 160 Mbps from anybody right now (I’m pretty sure Verizon doesn’t come close to this either). And why would you want to? Remember that even if you can receive 160 Mbps, unless the computer you are communicating with can transmit at 160 Mbps, you’re not doing to receive at 160 Mbps. So, basically, right now, I don’t think a 160 Mbps service is worth a whole lot for most people.

  11. Cablevision doesn’t seem to be scared. they already have a 30/5 package and the basic package is 15/2.

    Also they are testing out and deploying the narad tech to businesses wich provides 100/100 conenctions for each business. (cablevision is only doing 50/50 for the time being. )

    cable companies do have options.Just look at cablevision .They added Discovery hd and and national geographic hd and vs and golf hd in the future.

    I wish people would stop lumping all the cablecompanies together. they are not all the same.

  12. This reminds me of the megahertz wars when Intel and AMD would use clock speed to describe the capabilities of a CPU. It was a poor way to do that since processor speed had to do with so many factors determined by other things like cache sizes, applications etc and then real performance had to do with software, disks etc. Anyway, describing performance by using burst numbers is just a marketing tool. It is like saying that you can get from SF to San Jose in 40 minutes if no one is on the highway. . But on the other hands opening up more lanes is a good thing. I am actually interested in the minimum I can count on. In many cases, DSL actually is better at that I have written a bit more on this on my blog http://twothirdsdone.com/2007/05/11/congestion-ahead/.

  13. Good points majortom, but if you look at the territory overlaps – large swathes of Maryland, Philly-region and other states are where comcast and Verizon compete with each other.

    I did not lump all the cable companies together. by the way that is going to be an issue for all of them – verizon over laps with cox (virginia), time warner cable, and cablevision.

    cablevision is the only one with something on the market, so as you might have noticed that I have not mentioned them in this post, or others.

  14. Cable will be fine. The research labs will continue to finding new ways to add more bandwidth. The Next-Gen cable boxes will be out on the market around 2010, utilizing such technologies like MPEG-4, switched video and IPTV.

    I’ve written this many times before, real people don’t care how there service gets to them. Only geeks and techies care about the fiber thing. My service could come on a fishing line. If it works I’ll use it. Comcast showed that DOCSIS 3.0 is real and not an idea anymore. Verizon should be scared, not Comcast. For Verizon has to take video customer’s from Comcast. And with DOCSIS 3.0 cable companies will be able to complete.

    With future technolgy advances, the limit could be endless. But no one is ready to jump into the 100 Mpbs race yet. Even though Verizon can upgrade current customers right now, they won’t. The internet infrastructure can’t handle everyone at 100 Mbps symmetrically. Plus the price to move data off-network is costly. There will be caps imposed by every providers.

    Now that Comcast showed the future is very good for them, you all are “player hating” on Brian Roberts. Cable is the Old Dawg here, that wouldn’t go down without a good fight baby! Kudo’s to Brian Roberts on the successful Demo.

    Now he needs to just put his money where his mouth is and deploy it. I’m proud to be a Comcast worker and see a great future.


  15. I don’t give a flying whozis about what technical whichever does what. I’ve been online for 24 years, now. Whoever offers me faster gets the biz.

    Saying that – ain’t no one but ComCrap in my neck of the woods, anyway. DSL from Qwerst? Forget it.

    So, this weekend, I’ll be watching Fulham lurking just above relegation and there will be the hoardings from Pipex offering 8mb unlimited use for $25/month. That’s a 60% speed improvement over what ComCrap charges twice as much for!

    Screw ’em all!

  16. Brian Roberts is no dummy and there is a method to his madness. When push comes to shove he’d kick Schmidt’s arse from Mt View to Philly and back.

  17. Truedalife – you sound like a real genius.
    “Player hating”. Look, Einstein – cable is charging $45 a month for a service that costs them practically nothing to deliver. As soon as Verizon offers me a faster service for $35, I’m swtiching.

    I don’t care for any of these companies. Verizon is inept but G-PON is infinitely superior to HFC. Verizon will force Comcast and the rest to lower their prices which is good for consumers.

  18. And what will happen is the same thing that happens in every industry. You get low initial offers. Once the newcomer realizes they can make more $, they will raise rates to the same as before. Verizon has already upped FIOS/TV’s rates….

    We’ve seen it with cell phones, TV (ever wonder why DirecTV, Echostar, Comcast, and Verizon are about the same $), etc.

    Competition does not automatically guarantee low prices.

  19. Cable TV prices, satellite TV prices, and Verizon TV prices all reflect a basic cost they must pay for – that is programmming. For programming, the markup is typically around 100%. For broadband the markup is maybe 40 times that. There is very little cost to delivering high speed internet once the network is built.

    As far as cell phones go, I have had a cell phone for 18 years. Cell phone serice is certainly much cheaper than it was 18 years ago and cheaper than it was 5 years ago. Verizon charges me $79.99 for 1350 minutes – I couldn’t get anything close to that when I started with Verizon from AT&T in 2000. FiOS will lower cable’s prices on internet access. Trust me, that’s how competiton and free markets work.

  20. I don’t think Comcast (or any cable company) truly fears Verizon. The market overlap isn’t that great and FIOS is hideously expensive to deploy. The cable companies are the incumbents in the TV business… Verizon has to make a compelling case to the average consumer for why FIOS is important to them. The fact is, at the present time, they can’t make that case because 6mps, 10mps, or 50mps doesn’t make any difference to most people. Verizon can’t afford to get into a real price war… the costs of deployment are too high. They can’t exist on the concept of, “We lose money on every house we wire, but we’re making it up on volume.”

    The big hit for Verizon is in telephony. It’s a lot less expensive for cable to get into the telephone business than it is for the telcos to get into the TV business… and so, Comcast can get into into a price war over telephone service… and still earn a tidy profit.

    My view? Cable senses they’re at a peculiar moment in time. Verizon is mortgaging the company on FIOS — if cable can deny them a return on that investment, Verizon becomes a bit player in the residential market.

    I think Verizon’s prepared for that and almost expects that it will happen. I don’t think it’s a co-incidence that the major FIOS wiring is happening in areas that have a heavy business concentration. Ultimately, I think the cable companies will dominate the residential market, while Verizon concentrates on the commercial side. Not a bad idea for either… it’s what they’re both good at.

  21. The future is all about guaranteed bandwidth – cable’s peak bandwidth numbers are a joke. People are watching less and less TV and getting more and more content from the Internet. Just look at the recent headlines of how many viewers network TV is losing. CBS, NBC, and ABC used to be the TV world and now they are almost irrelevant.

    After CES this year, Gates declared we won’t recognize TV in 5 years when he saw Verizon introduce a 150 Mb/s connection. It will probably take a little more than 5 years but the future is about getting your content from the Internet and sending it to your TV via some new product from Apple or Microsoft.

    In Europe, many younger people have compelely dropped TV services the way people now have dropped landline phone service. They future is a cell for communication and the nternet for content.

    FiOS is simply better technology than the shared bandwidth of cable – even with DOCSIS 3.0 and channel bonding.

  22. “It is not going to be in couple of years. ”
    Not so, Japan’s biggest cable operator (and a few smaller ones in Japan and Korea) are already deploying a pre-Docsis 3.0 version of this:
    This is something you can get, now. Their solution is running in that same “super modem” shown by Comcast’s CEO.

  23. cablevision is the only one with something on the market,

    Well, Cox in the Fairfax area offers higher speeds than elsewhere. They do a 15Mbps down/2 Mbps up in FIOS areas.

    One of the amusing (but frustrating!) things about all of this was Verizon accurately responding by bragging that they could offer true 100Mbps symmetric right now if they wanted to, and that they are in tests right now, and anyway GPON will give them even more bandwidth to play with– and then trying to talk around why they’re not offering those sorts of speeds right now. “No truly compelling reason,” which is true insofar as that means that they don’t think that people would pay them more for the extra speed, so it wouldn’t increase profits. “We don’t need do until there’s real competition” would be another honest response, but for some reason they didn’t want to say that one.

  24. Om:

    Did Comcast tell you that their average node size is 400 homes? I am very close to them and I doubt seriously it is anywhere near 400 HPN. The present day Comcast is made up of many acquisitions, many of which were designed with 2000 and even 2500 homes pockets. They are trying to segment these nodes down to smaller sizes but I can assure you they are nowhere near 400 homes per node.

  25. Sorry for the long post but…

    Let’s face it, when it comes to network technology the average consumer is not very smart. I don’t mean this disparagingly, it’s just not a subject most consumers are intimately familiar with. So when a cable executive claims that they will someday enjoy 160Mbps broadband services, the average consumer probably believes that—someday—they’ll be able to download 160 million bits worth of stuff (music, videos, etc.) each second. While these prototypical consumers aren’t network-savvy, they do know when they’ve been taken. When they fail to achieve the aforementioned 160 megabits per second, they’ll know they’ve been suckered by a marketing pitch.

    The central problem is that broadband has always been described using a single number. Seven hundred kilobits. One point five megabits. One hundred and fifty megs. Consumers, naively, think this is the speed at which the broadband connection operates. They don’t understand that this is only a measure of peak downstream bit rate. In other words, if everyone in your neighborhood agrees to stay off the internet while you’re doing your work then, yes, these speeds might be achievable. But the moment other people begin using the internet you’ll begin to realize that your cable connection is shared. And your speed will drop significantly. Much like freeways, the sign may say that 65 miles per hour is the speed limit but try driving that fast during rush hour.

    There is an important quantitative measure of network performance about which consumers will become increasingly knowledgeable (and demanding): peak-to-average bandwidth ratio.

    Peak-to-average bandwidth ratios, basically a measure of network oversubscription, were often as high as 20-, 30-, or 40-to-1 back in the days when most internet traffic was HTTP and SMTP. This meant that service providers would sell 40 people a one megabit service despite the fact that those 40 people were sharing a single one megabit connection. That was ok because people mostly downloaded web pages and then spent a few minutes reading them. Or sent an email and didn’t know or care if it took one second or ten to reach its destination.

    However, with internet traffic increasingly comprised of streaming media, these oversubscription factors need to drop dramatically in order to preserve the illusion that the subscriber actually has that much bandwidth. Five- or ten-to-one might be more appropriate for today’s demanding internet user. In IPTV service delivery infrastructure the challenge is even more substantial. The peak-to-average bandwidth ratio for the portion of the access network allocated to IPTV must be 1-to-1. Yikes.

    It is here that telcos have a distinct advantage of the cable guys. This is because, while both access networks are oversubscribed, in the telco plant (whether it is fiber or copper) a network element (DSLAM, OLT, or MSAP) intelligently performs the oversubscription. Changing service quality for a single subscriber, with all due apologies to net neutrality socialists, is as simple as a software command. Changing service quality for all subscribers served by that network element involves an upgrade of a single network uplink. In the cable plant oversubscription is a function of the fact that the access network—analog or digital, DOCSIS 2.1 or 3.0—is a shared medium. Improving service quality for some or all users involves physical network reengineering. Much more complex, expensive, and time-consuming.

    It’s about time service providers truthfully described the product called broadband. If they did, the Brian Robertses of the world would have to admit that average downstream speeds of their 160 megabit product (assuming, say, a 500-home node with a 50 percent broadband take rate) are more like 640Kbps, a peak-to-average bandwidth ratio of 250. Double yikes.

  26. One thing Verizon can do with G-Pon is legitimately offer a guaranteed 50 Mb/s service and allow customers to subscribe to IPTV from third parties. Verizon may not sign that customer as a video sub but they will retain him as a voice sub and take away the broadband business from the competing MSO.

    Doing that would expose the cable quoted bandwidth numbers for what they are: bogus.
    For cable to offer a guaanteed 10 Mb/s service in a 400 home system, using 256 QAM, cable would require every bit of bandwidth on the plant leaving nothing for TV or VoD. There is no physical way that they can ever come close to 50 Mb/s guaranteed without a total plant rebuild.

  27. Om, as this thread demonstrates, you have lots of readers/commenters with advanced technical understanding of broadband infrastructure and how it affects competitive dynamics in the industry. Perhaps it would be useful to the ongoing discussion of this topic to provide some charts that definitively illustrate the bandwidth delivery capabilities and limitations of the primary architectures (750HFC 1000HFC SDV, switched fiber, PON fiber, DWDM/PON, etc). When one understands the relative architectural constraints and/or advantages of each player in each market, the competitive dynamics become much clearer. Admittedly, when viewed across the entire US map, this becomes quite complex; yet that complexity is the reality of the US broadband environment. And who do we have to thank for that – the FCC and the DOJ antitrust division. Anyway, this thread provided an excellent discussion of the subject. Thanks for kicking it off.

    Addendum – This Cisco overview of DWDM provides some insight into the functional differences among broadband architectures; it does not, however, include the critical economic data required to provide a basic starting point in understanding infrastructure competitive dynamics.

  28. The author Om Malik has used the term FUD incorrectly. The actions of Comcast here do not quite meet the requirements of ‘FUD’.

  29. One thing to keep in mind about the cable world of today. Most are HFC, Hybred Fiber Cable, meaning up to 100 miles of Fiber from the source and about 6000 ft of coax to the home. Fiber distribution with an average of 4-6 fibers to each fiber node in the streets.

    So lets assume of a moment that Telco Fiber to the home really started to take off. With the advent of IPTV in the next 5 years, and cable already is moving content between Master headends and HUB’s via IP instead of older proprietary solutions. The coming of D3.0 will provide a migratory path to making video distribution IP only, with SDV providing the bandwidth re-use near term opening up more bandwidth for more D3.0 channel. My point is going forward the distribution of IPVIDEO will be standard.

    Here is the kicker, remember the HFC, cable already has a fiber infrastructure to the neighborhood, all they have to do is replace the fiber node with a switch, and upgrade the ‘coax to fiber’ to the house. Not lay it in the ground from the central switch all the way to the house. At 10k per house, actual cost of the fiber to the house, not the 650 “hook up cost” telco’s tell wall street.

    So Cable and Telcos really both have the opportunity to provide fiber to the home, cable as an upgrade, telcos as a green field.

    IMHO Netrydr57

  30. SDV will help but not nearly as much as getting rid of analog. The problem for getting rid of analog for cable is the same as everything else – cap ex. All of a sudden, Comcast would be buying an additional tens of millions of STBs at $200-350 each and having to rent them cheap to not lose customers. Even with that cable still can’t provide anything close to the guaranteed bandwidths on broadband that Verizon’s G-PON can.

    Finally, $10K per home is wrong. According to you for Verizon to connect 20 million homes to FiOS will clost $200 billion. Uh no – Wall St isn’t that dumb.

    The way this battle will shape up is that cable will give away voice and Verizon will lose money on basic video as they each fight for triple play customers. Cablevision will be hurt the worst since they have 100% overlap with Verizon. And they know, it’s coming. So does Wall St.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.