news intel details next gen thunderbolt controllers

Intel details next-gen Thunderbolt controllers

Intel unveiled two new generations of Thunderbolt controllers at the National Association of Broadcasters show yesterday. Rumors of the Redwood Ridge and Falcon Ridge chips circulated this past summer, and it looks like they were correct. Redwood Ridge will be available in dual-port DSL4510 and single-port DSL4410 flavors, each with two channels per port. Like current implementations, the chips boast 10Gbps of bandwidth per channel. The interconnect’s DisplayPort support has been upgraded from version 1.1a to 1.2, and power management has purportedly been improved. 

Source: Intel

According to the presentation given at NAB (PDF), the new controllers also promise a reduction in "platform BOM cost and area." Motherboard makers have told us that existing Cactus Ridge chips cost $30-35 to implement, which is incredibly expensive compared to peripheral interfaces like USB 3.0. Thunderbolt cables are also pricey, although Intel says "lower-priced cables are available now." The cheapest one I see at Amazon is a 20" Apple unit for $29, with longer models running $39 and up. Cables used to cost about $50, so there’s admittedly been some improvement on that front.

Expect to see the Redwood Ridge chips in Haswell-based systems. According to Engadget, the new Thunderbolt tech will be built into some Haswell processors, as well. It’s unclear whether the controller will be incorporated into the die or will simply ride shotgun on the same package as the processor. Either way, I suspect the treatment will be reserved for highly integrated mobile parts rather than desktop chips. An ultra-thin convertible tablet with Thunderbolt would be pretty sweet.  

Source: Intel

Later this year, Intel expects to begin the initial production of another new Thunderbolt controller. Dubbed Falcon Ridge, this one promises to double the per-channel bandwidth to 20Gbps in each direction—enough to transfer and display 4K video simultaneously. It looks like there’s only one channel per port, so the aggregate bandwidth of the interface hasn’t changed. That’s likely why Falcon Ridge will be "fully backward compatible" with existing cables and connectors.

Intel demoed early Falcon Ridge silicon daisy chained to dual 2560×1440 displays and a pair of SSDs pushing 1200MB/s in IOMeter. That’s pretty impressive for a single cable, regardless of the cost.

Update — Intel has confirmed that Falcon Ridge combines the dual 10Gbps channels of current Thunderbolt implementations into a single, 20Gbps channel. The signal wiring for the cables and connectors is unchanged.

0 responses to “Intel details next-gen Thunderbolt controllers

  1. They are built for the home-server/ HTPC crowd. None of them *need* Thunderbolt or 10Gbps Ethernet or greater.

  2. RAID 0 is just stupid. That’s no excuse for it these days. A single SSD 2.5″ disk utterly destroys a sizeable RAID 0 of HDDs with none of the drawbacks expect capacity. If you need more bandwidth a PCIe SSD card outclasses a massive RAID 0 of HDDs and most SSD 2.5″ RAID 0 setups.

    RAID 1 is lazy-man’s back-up scheme. I/O performance has never been its strong-point. 😉

  3. Sorry, that situation is the definition of exotic/niche.

    You will be hard pressed to find something that has a genuine need for Thunderbolt and USB doesn’t cut it. Just give me an example that would be found commonplace.

    Intel wants to replace the USB dynasty on a faster standard and they have completed IP control over the standard (royalties). That’s why Intel has been so reluctant on releasing a native USB 3.0 controller on their platforms, When they finally do so, they don’t care if the controller has stupid issues (Haswell bug).

    Fiber cabling was deemed too fragile and expensive on the controller end. Fiber has enough trouble trying to overtake UTP in the Ethernet arena. I doubt Intel would have better luck.

  4. [quote<]I'm not sure you could get the Mini or MacBook on 10GbE.[/quote<] You can with a TB port ;D [quote<]That being the case, are you using an array of SSDs, or just a lot of nicer HDDs to get >300MB/s sustained transfers?[/quote<] Just a couple of WD blacks in RAID 0 smack against that. In my case I'm running 4 WD 2GB Blacks in Raid 0.

  5. RAW footage is murder, and trying to edit it across multiple machines has to be a pain, I feel for you!

    The 10GbE SAN idea makes a whole lot of business sense when a dozen or more people need fast access to the same repository, but yeah, that’s a huge investment for just one desk, and I’m not sure you could get the Mini or MacBook on 10GbE.

    That being the case, are you using an array of SSDs, or just a lot of nicer HDDs to get >300MB/s sustained transfers?

  6. [quote<]For storage, TB is still limited by the SATA/SAS interface, per drive- yeah, eSATA isn't as flexible, but if you need more than 300MB/s sustained writes out of a large volume then you're doing something pretty special. If you have an array that actually needs that much bandwidth, is it really going to be mobile? I mean, it's not like it can't be carried around, but is that really a solution?[/quote<] HD video editing on raw footage would quickly realize that even 300 MB/s quickly becomes a bottleneck for the simplest of things like scrubbing through the footage. With the TB array I am easily able to hook it up to my PC or the MBP or the MacMini and work on the footage one any of them. It is however not even just video editing that it also becomes a factor. Working with a complex dbase query also heavily depends on raw drive throughput. [quote<] Setting up a 10GbE network sounds like a better idea, either to a SAN, or to a workstation with an internal array.[/quote<] Setting up such a system far exceeds the cost of a TB setup. That may be a solution for the enterprise but not for the prosumer or home enthusiast. A single workstation array will work but you do lose the ability to use it on multiple machines ( something I do very often ).

  7. Man, you’re just asking to be beat up!

    Even though you’re correct, implying that RAID0/1 aren’t real ‘RAIDs’, many people would argue that both laptops and ITX systems are more than happy to support ‘RAIDs’. 🙂

  8. Let me backtrack a little- I agree that TB is perfect for video processing workflows, and I agree that it’s a much more capable solution than a number of other technologies available, such as USB, Firewire, Ethernet, and SATA, and that it has the potential to replace all of them- along with DisplayPort- at once.

    It also makes tremendous amounts of sense for Apple users/studios that rely primarily on iMacs which cannot be expanded internally.

    But it doesn’t make a lot of sense elsewhere, and that’s where I’m coming from.

    For storage, TB is still limited by the SATA/SAS interface, per drive- yeah, eSATA isn’t as flexible, but if you need more than 300MB/s sustained writes out of a large volume then you’re doing something pretty special. If you have an array that actually needs that much bandwidth, is it really going to be mobile? I mean, it’s not like it can’t be carried around, but is that really a solution? Setting up a 10GbE network sounds like a better idea, either to a SAN, or to a workstation with an internal array.

  9. I am talking about video capture devices such as a hauppauge colossus. Not video cards although that option still exists with TB. eSata is easily enough to saturate as well, especially as your drives are sharing one sata connection for all the bandwidth and are usually limited to Sata2 speeds. When you are using an eSata connection for multiple drives in an array you are splitting the bandwidth between the drives (this is how port multiplication works and is what external eSata enclosures use, two methods FIS or CMD).

    [url<][/url<] TB does not have this limitation as you can have a true pci-e sata controller which offers each drive it's own Sata link allowing for far greater bandwidth then an eSATA connection.

  10. [quote<]Thunderbolt *needs* a killer app[/quote<] eSATA/2560x1440 display/FW800/Gigabit Ethernet/USB/Thunderbolt drive - I'm driving all of that out of one port on my laptop. [quote<]in order for to have a decent chance at overtaking USB in the mainstream market. [/quote<] USB 3.0 and Thunderbolt are complimentary. This has nothing to do with one replacing the other. USB 3.0 can run over Thunderbolt for heavens sake. Thunderbolt is an amazing technology that has been hurt in a number of ways: 1) Lightpeak and the switch from optical to copper - Thunderbolt is viewed as something less because of that 2) Adopted by Apple first - kiss of death as far as the PC crowd is concerned 3) when one factors in the above points and then mixes in a high price you have condemned Thunderbolt to the realm of FireWire for many PC enthusiasts In hindsight I would have preferred that Intel had held out until they released Thunderbolt for their own products. A name change would have helped - something like PCIe over cable.

  11. Do you really want to move video cards around? I can understand the idea of having external GPUs, in some cases, but it’s still a stretch. If you need real mobile video power, is a desktop-replacement or ITX build too much?

    And is eSATA a no-go for external RAID? Is it not fast enough? I guess if you can saturate one SATA port, but you’d need what, six mechanical drives to do that in RAID5/6?

    My point is that Thunderbolt doesn’t add capability, it adds convenience, at a relatively extreme cost. Perfect for the Apple/media professionals crowd, but not so useful for consumers/enthusiasts.

  12. Bull, there are a ton of external RAID enclosures out there that are nowhere close to being built for the enterprise environments that you are talking about. (BTW, using a Gbit connection for a RAID setup for anything but redundancy is a waste as even a single 5400 rpm green drive is able to saturate the link).

  13. More like “I want a device that I can plug multiple systems into without ripping the thing apart every time. I want my system mounted to the back of my monitor but not lose out on expansion. I want to be use my video hardware on multiple systems. I want the full performance of a raid and not lose performance due to signal switching between drives…. etc.”

    While I realize that most people don’t ever even plug a expansion card into their full sized pc there are people that would like to utilize a device on multiple machines without giving up on performance, features or investing in duplicate hardware just so they have that capability on multiple machines.

  14. True, it even allows for solutions to questions that no one should be asking.

  15. Deanjo derping just to make a point:

    ‘I want a full-on PC with internal interface expansion stuff, but I want it to be in a bunch of little boxes that take up as much or more total volume as a normal PC. And I’ll gladly pay plenty for it!’

  16. I think we are in the process of the mainstream being redefined. It’s an open question as to whether the PC itself will continue to be “mainstream”.

    How’s this for an analogy — in the late 90s / early 2000s the PC was a healthy star. When netbooks came out, the star expanded into a red giant. Now the PC is contracting from red giant into a white dwarf. Perhaps Thunderbolt is white dwarf technology.

  17. Within certain niches that is.

    Outside of that. Thunderbolt simply costs too much and doesn’t yield any tangible benefits to warrant the costs.

    Thunderbolt *needs* a killer app in order for to have a decent chance at overtaking USB in the mainstream market. Unfortunately, there doesn’t seem to be anything in the foreseeable future.

  18. “RAIDs” aren’t being used in ITX form factors or laptops. They are often placed in dedicated NAS or SAN boxes that utilize Gigabit, 10Gbps and even 100Gps Ethernet (iSCSI).

  19. The biggest problem with Thunderbolt is that there’s no killer app for it in the mainstream market.

    Thunderbolt is inherently more expensive and complex then USB. (more tracings, dedicate controllers for cabling etc). Motherboard manufacturers operate on a tight bottom line. Thunderbolt costs too much to be used outside of high-end boards.

    The lack of killer app and cost makes Thunderbolt a redux of Firewire. A high-performance solution catered towards certain niches.

  20. I had zero issues with Thunderbolt under OS X and Windows 8.

    USB 3.0 is great up until a point. Beyond that you need Thunderbolt.

  21. This is just the first or second gen version… there must be a lot of flaws, which can only be identified with time.
    If I bring such a product to my home, most likely it will be outdated in a while, despite the high cost.

    My view is that, it takes 3-4 major revisions, to make a product successful.

    Well, honestly I want this tech, but I don’t need it yet

  22. The only thing ronch detailed was some drivel about motherboards, which I found quite irrelevant because nearly everyone who uses Thunderbolt devices are on Macs. Thunderbolt is really a Mac thing – I honestly wouldn’t even trust a PC Thunderbolt implementation. There are lots of Thunderbolt devices now and I am 100% sure they were tested quite well and are well supported on the Mac. I am 99% sure that very, very few of them bought crappy ASUS or GIGABYTE motherboards to test against. Why should they, to them the PC users who have Thunderbolt ports is the niche market.

  23. [quote<] thing is, nothing that you put outside of the enclosure actually needs it.[/quote<] Ever try to stick a PCI-e card in a Zotac/Mac-Mini/NUC for example? Ever try to get true raid speeds through USB? Not happening with even USB 3. Something as trivial as even checking a smart status or running of a low level diagnostic of a USB drive is often impeded by the USB controller.

  24. [quote=”TEAMSWITCHER”<]I do wish that Thunderbolt trickles down to the common man. It could revitalize the PC market with stunning new form factors that sacrifice nothing is connectivity and performance.[/quote<] They're already there, and ronch detailed the reasons that it doesn't matter above. Thunderbolt is just a standard for pushing PCIe outside of the enclosure; thing is, nothing that you put outside of the enclosure actually needs it. Combining cables is cool, but it doesn't enable more functionality, just slightly more convenience at a highly elevated platform cost. The only real advantage that it brings is, like you say, for special external high-bandwidth applications that made use of the bandwidth and latency advantage of Firewire.

  25. I think what they meant was: ” chained to dual 2560×1440 displays while a pair of SSDs pushed 1200MB/s in IOMeter.”

    Prior benchmarks have shown that TB 10Gb can do an effective rate of 1200MB/s, that’s after overhead.

  26. The new Thunderbolt spec was presented at the NAB Show. NAB stands for the National Association of Broadcasters. Thunderbolt is a very high-end spec for video professionals on the go, and I think they have far more interest in a $2000 MacBook Pro (2 Thunderbolt ports) than a $300 ASUS or Gigabyte motherboard. These guys already have plenty of equipment to lug around – they don’t want a PC box. At least I don’t think so….

    I do wish that Thunderbolt trickles down to the common man. It could revitalize the PC market with stunning new form factors that sacrifice nothing is connectivity and performance.

  27. IDF, Beijing, April 9, 2013: The US Government and Intel jointly today announced that they have trademarked the name “USAIn Bolt”. Anticipating a reaction from China’s state run China National Radio, Intel also announced that it was granting China a free license to the use of the trademark ‘CHAIn Bolt”. Not to be outdone, the DRAM industry, Intel and US automakers GM and Ford have jointly trademarked the term ‘DRAIn Bolt”.

  28. Not surprising that you won’t find them on intel boards. Their current line of boards have been out for a while (before the TB spec was really finalized) and intel is getting out of the motherboard business sadly.

    And btw, Asus offers a TB PCI-e card.

  29. at least USB 3 is practically out there on most new boards. i only found one or two motherboards with TB on them, and none were intel brand boards, lol. pci-e TB cards are nowhere to be found, so who really cares what intel upgrades with TB when you cant even buy it.

  30. Probably because of the losses on the board/connector/cable at 20Gb/s data rates. The repeater inside the cable is needed to handle the losses of the cable alone; if the board/connector losses are added in both ends, it might not be able to operate that fast.

  31. [quote<]And many of them are very finicky.[/quote<] Such as? My TB devices have given me zero issues.

  32. I think they might be referring to the “hooks” from some versions of the Thunderbolt add-in boards that will allow them to be powered down more completely than prior versions built for SB or IB.

  33. Don’t forget that of the few Thunderbolt devices out there, almost all of them are niche devices with huge price tags. And many of them are very finicky.

    At this point, I wish Intel would have just took their time to get the fiber version out and skipped all these copper versions since they aren’t making any impact anyway.

    For most purposes, USB3 is enough and it’s SO much cheaper and more universal.

  34. I must say, I love the private eye remark. It’s apropos AND funny.


  35. Is that still 20GB/s per channel and bi-directional? Or are the two 10GB/s channels now combined into one? I see the article says “double the per-channel bandwidth to 20Gbps in each direction”, but are there still two channels? Thunderbolt first gen had two 10GB/s upstreams and two 10GB/s downstreams (10d/10u and another 10d/10u) for example for an aggregate bandwidth of 40.

    EDIT: Found an answer, it is, indeed, just combining the channels


  36. It all boils down to price. Take USB 3.0, for example. It’s practically added at no cost to today’s $150 boards, so everybody is ok with it. I’m not sure about Thunderbolt, but if it means we can only get it with $300 boards plus not too many Thunderbolt-enabled devices, then at this point it’s just a niche piece of tech and largely irrelevant unless it trickles down to the mainstream. Of course a lot of new tech start out this way, but not all end up successful, that is, eventually being widely adopted.

  37. No thanks, my assistant will liaise with the CEO of Googley directly if they can be located somehow.

  38. Probably because it can be customized to the physical properties of the cable itself. Sort of like long/thin HDMI cables with Redmere.

  39. Sure you can find the correct address using Google too. Did you need some help with that?

  40. Intel Dept of Strategic Obfuscation Rule #Twoo:
    We don’t pick the names so people can understand them easily.

  41. Well it’s only a pair of SSDs. More than likely, they were both maxed out. Probably should have added a 3rd and 4th SSD.

  42. [quote<]Thunderbolt tech will be built into some Haswell processors, as well.[/quote<] What suffix will they use for that? "T" is already taken.

  43. Thanks, I’ll pass on that tip to my private eye. Should I pay you a finder’s fee?

  44. [quote<]possibly even hire a private eye to track them down[/quote<] Or learn how to use google. I didn't have much trouble finding my pci-e box, raid box or hd capture device.

  45. I think it’s a metaphor related to FDR… or perhaps the obscured back side of the coin represents 3 things Intel doesn’t want to show us: liberty, peace, and victory.

    Just my [s<]2[/s<] 10 cents 😛

  46. typo: [s<]10GBps[/s<] [b<]10Gbps[/b<]. Also, I find it unusual they used the number "1.2GB" of their IOMeter tests. According to the spec of TB v1.0, that number (10Gbps) should have already been possible to benchmark. I'd find it more impressive (and relevant) if they can do 2.4GBps while powering those two monitors. The bandwidth is also supposed to be bi-directional, so if we had a demo showcasing 20Gbps to and from at the same time, while powering a 4K monitor, that would actually have demoed the upgrade.

  47. Buy the cable and get a Free exclusive Intel RAID TB SSD enclosure fully populated with a 1TB array.

  48. Is the coin included in the Intel photo as a hint that you need to spend a lot of cash to buy TB peripherals/cables and possibly even hire a private eye to track them down?