How Intel Turned Thunderbolt From A Failure Into A Success

[Megalith]

Is Thunderbolt finally becoming a success? As many of this year’s top-tier laptops feature Thunderbolt 3 ports, that may not be such a crazy notion.

With Thunderbolt seemingly on the ropes, Intel had one last move—one that likely put the technology on a winning path at last. At last year’s Computex, to the surprise of many, the company announced a faster version of the spec called Thunderbolt 3, with speeds up to 40Gbps—and it could do it over the new USB Type-C connector, instead of the funky MDP cable. Intel essentially uses the same alternate mode that DisplayPort does to pass Thunderbolt signalling over PCIe. And by integrating a USB 3.1 10Gbps controller into the Thunderbolt 3 controller, it could fully support USB 3.1 too. What Thunderbolt 1 and 2 couldn’t do, Thunderbolt 3 has finally achieved in its vision of “one cable to rule them all.”

 

[Galvin]

Would be awesome if one day one port can do it all. No longer have to have different connectors.

 

[Skripka]

And how many devices out there have thunderbolt versus USB?

Calling that lack of penetration beyond motherboard interconnects a, "success" is really pushing it.

 

[funkydmunky]

Intel was forced to do this. Unless they wanted another RAMBUS' esq failure, they had to make concessions on their proprietary standard.

 

[rat]

Thunderbolt is still a failure considering what it had to do to change itself to become usable.

 

[pxc]

Intel was forced to do this. Unless they wanted another RAMBUS' esq failure, they had to make concessions on their proprietary standard.

Tech word salad of the day quota fulfilled.

Intel and Apple were pushing Thunderbolt through standardization even before the very first version was released, so I'm not sure why you call it "their proprietary standard". It's about as proprietary as Intel's other "proprietary" standards like USB, PCI and the ATX form factor.

 

[daglesj]

I wouldn't break out the champagne just yet. I'll wait for version 5 at least, when the cards cost $25 and the cables $2.

 

[Poseur]

I would not use success when describing Thunderbolt. I don't know anyone that has it or wants it. I do know people that would love >1Gb ethernet, but that's somehow more difficult to implement for the same money.

 

[SparkedFire]

I would not use success when describing Thunderbolt. I don't know anyone that has it or wants it. I do know people that would love >1Gb ethernet, but that's somehow more difficult to implement for the same money.

I wouldn't call Thunderbolt a success either. Seriously, when will 10Gb ethernet hit the consumer side?

 

[alxlwson]

I wouldn't call Thunderbolt a success either. Seriously, when will 10Gb ethernet hit the consumer side?

When there is a need for it. Netflix and pr0n don't even need that.
The driving factor will be digital media distribution, and even uncompressed 4K doesn't need that kind of throughput

 

[Krixon]

When there is a need for it. Netflix and pr0n don't even need that.
The driving factor will be digital media distribution, and even uncompressed 4K doesn't need that kind of throughput

The only use I can foresee coming up is streaming high-def interactive VR applications (much higher res than the Rift or Vive). Even then I'm not sure that someone won't come around with a better streaming compression protocol that would extend the lifespan of 1Gb ethernet.

 

[funkydmunky]

Tech word salad of the day quota fulfilled.

Intel and Apple were pushing Thunderbolt through standardization even before the very first version was released, so I'm not sure why you call it "their proprietary standard". It's about as proprietary as Intel's other "proprietary" standards like USB, PCI and the ATX form factor.

I was under the impression TB was not cost effective, like FW before it and was going to ultimately loose to USB. I assumed Intel was profiting. Or was it just the cable makers?

 

[nightanole]

I wouldn't call Thunderbolt a success either. Seriously, when will 10Gb ethernet hit the consumer side?

When you Ethernet port doesnt suck 25%+ of what your cpu idles at. People are not keen on loosing an hour of battery life just because a few people have applications that need above 100 meg a sec. Id love to have 10gbs network switch, but they are a wee bit expensive.

 

[sfsuphysics]

Well if the PR guy for Intel says it's a success who am I to argue!!

 

[jwcalla]

When there is a need for it. Netflix and pr0n don't even need that.
The driving factor will be digital media distribution, and even uncompressed 4K doesn't need that kind of throughput

If you have two computers networked and they both have SSDs, you're going to be bottlenecked by the 1 Gbps network when transferring files. Those of us who use these types of setups every day for work are dealing with this.

It would be nice if we could do ethernet over Thunderbolt some day. Otherwise I have no need for it right now.

 

[kandrey89]

I'll take a 4K 15.4" OLED Display in a 0.5" thick notebook with a lifespan of 2 days running at full load, 8TB NVMe SSD, and 8 USB-C (Thunderbolt3) ports (all of which support notebook charging at a rate of 0 to 90% in 2 hours).
Then I can at least use NVMe SSD to use up half the Thunderbolt3's bandwidth.

 

[nutzo]

I wouldn't call Thunderbolt a success either. Seriously, when will 10Gb ethernet hit the consumer side?

I'll be happy when the prices (switches and cards) come down enough that I can justify adding it to all my servers.
Been running Gb Ethernet for 9+ years, now using multiple ports on the servers, but I'd love to just plug in a single 10Gb connection.

 

[alxlwson]

If you have two computers networked and they both have SSDs, you're going to be bottlenecked by the 1 Gbps network when transferring files. Those of us who use these types of setups every day for work are dealing with this.

It would be nice if we could do ethernet over Thunderbolt some day. Otherwise I have no need for it right now.

I misread. I thought he said 10Gb internet, not ethernet.

 

[limitedaccess]

I wouldn't call Thunderbolt a success either. Seriously, when will 10Gb ethernet hit the consumer side?

Right now there might be a push for a 2.5Gb or 5Gb step up instead.

 

[Vyedmic]

It would be nice if we could do ethernet over Thunderbolt some day. Otherwise I have no need for it right now.

Something like this?

https://thunderbolttechnology.net/s...ing and Routing Instructional White Paper.pdf

Apple's stuff supports IP over Thunderbolt for a few years now, not sure about Win...

Or would you rather use this?

Products - SANLink2 Series | Thunderbolt 2 Adapter | PROMISE Technology

 

[mesyn191]

I wouldn't call Thunderbolt a success either. Seriously, when will 10Gb ethernet hit the consumer side?

There have been industry comments that 16nm processes from TSMC will allow for affordable 10Gbe. More recently the prices on it have seen some big drops. Netgear sells a $800-ish 10Gbe switch that is pretty decent. The NIC's can be found for a decent price too. Especially if you don't mind buying them used.

If you have two computers networked and they both have SSDs, you're going to be bottlenecked by the 1 Gbps network when transferring files.

Until 10Gbe gets cheap just use teaming. Teaming up 2-4Gb ports will get you pretty close to 2-4Gb transfers in real world scenarios and usually requires minimal software and hardware configuration to pull off. Intel based 2-4 port NIC's can be had for $50 used these days if your mobo doesn't have enough ports to do this.

 

[westrock2000]

Thunderbolt is pretty good stuff, the only issue I have with it is that generally only expensive gear implements it.

I would gladly purchase items using it if I could. I do have my monitor connected through Thunderbolt.

I think some of you are looking at purely from a speed perspective or something. Thunderbolt is only a transport, it is not a protocol. So you can run any protocol over it. Thus you can have a monitor attached, you can have a video card attached, and you have a hard drive attached. And all 3 would be able to use their native connection protocol. The computer will think that the video card is plugged into a PCI-E port and that the hard drive is plugged into a SATA port. USB 3 does have UASP protocol which replicates SATA through USB....but its an additional standard that had to be sorted out....it's not the same as allowing arbitrary protocols to be used over a standard medium.

To me this is a big deal becasue it allows you to have external gear, like hard drives, but still get the realiability of a native SATA for stuff like ZFS. ZFS does not like USB because USB is a little messy when it comes to data integrity. Like I said UASP does help fix that, but we had to wait for it come into existence first. With Thunderbolt, that is not an issue. Any new protocol is automatically supported.

 

[lazz]

My Gigabyte Z170X Gaming 7's USB 3.1 Gen2 type-C port got upgraded to a Thunderbolt3 port with a driver update a month after I bought it, since it already used the Intel USB3.1 chip. I don't yet have anything to run on it that could take advantage of the 40Gbps speeds, but the type-c port at least has been very useful already.

As the latest USB standard port is effectively merged with Thunderbolt, it becomes even more useful. Crazy data rates, DisplayPort 4k support, huge power delivery.. this is the first truly all-in-one connector.

 

[Quix]

Thunderbolt is eventually going to be included in Intel chipsets, when that happens everything will have it. By making it not a choice between USB-C and Thunderbolt 3, you can have both, they just made it easier for manufacturers to put it on everything.

 

[mesyn191]

The problem for Intel is that even if they put it in all of their chipsets for "free" there are still hardly any devices that will use it since it costs lots more than USB3.1. Remember fractions of pennies count here!

Maybe if they did a prolonged and concerted contra-revenue sales tactic and decided to lose money for 5yr+ straight on the chipsets that go into devices they'd have a shot but I don't see them doing anything like that.

 

[lazz]

The problem for Intel is that even if they put it in all of their chipsets for "free" there are still hardly any devices that will use it since it costs lots more than USB3.1. Remember fractions of pennies count here!

Maybe if they did a prolonged and concerted contra-revenue sales tactic and decided to lose money for 5yr+ straight on the chipsets that go into devices they'd have a shot but I don't see them doing anything like that.

It's kind of a chicken and egg scenario, for sure. But think of emergent industries like VR.. with a Thunderbolt 3 connection over a USB-C interface you could deliver video, power and data with one cable, and it could be long enough to suspend from a ceiling.

40Gbps is a lot. Do most people need it? Of course not. But most people don't spend $600 or more on GPUs either

 

[defaultluser]

Something like this?

https://thunderbolttechnology.net/sites/default/files/Thunderbolt™ Networking Bridging and Routing Instructional White Paper.pdf

Apple's stuff supports IP over Thunderbolt for a few years now, not sure about Win...

Or would you rather use this?

Products - SANLink2 Series | Thunderbolt 2 Adapter | PROMISE Technology

Yeah, Thunderbolt native networking sounds all well and good...until you try to buy switches. They still don't exist yet.

You can do point-to-point for cheap, but you can also do that with two 10GbT cards.

The other thing holding back Thunderbolt as a networking technology: the cables are seriously fucking expensive, and they're limited to 3 meters. That's an order of magnitude shorter of what you can get with Cat 6A, which is cheaper too.

Yeah, you can cheat and go optical, but that adds a hundred bucks for each converter. Might as well go Ethernet and call it a day

 

[Skripka]

Yeah, Thunderbolt native networking sounds all well and good...until you try to buy switches.

You can do point-to-point for cheap, but you can also do that with two 10GbT cards.

The other thing holding back Thunderbolt as a networking technology: the cables are seriously fucking expensive, and they're limited to 3 meters. That's a n order of magnitude shorter of what you can get with Cat 6A, which is cheaper too.

LOL. Reminds me of the good old days of HDMI up to v1.3 or so, where you'd get snowflakes on a digital interface over 3m because idiots decided error correction wasn't needed.

 

[defaultluser]

LOL. Reminds me of the good old days of HDMI up to v1.3 or so, where you'd get snowflakes on a digital interface over 3m because idiots decided error correction wasn't needed.

And yeah, you can use the new non-active 20Gbps Thunderbolt mode on USB type C for *way* cheaper cables, it's still limited to the exact same 3m length.

USB Cable Length Limitations And How To Break Them

In fact, the longest USB 3.1 cable you can buy at Monoprice is just 6 feet. So that spec limit on passive cables looks to be a little too high.

If you want to send data fast over long copper cables, you have to design the cable standard for that from the beginning. Not hack on the capability as a marketing check box.

And Thunderbolt is still desperate to find that "killer" marketing check box. In the new passive mode it's not vastly better than 3.1 at most use cases people need short fast external connections for. And in it's faster modes it will still be more expensive to implement than USB 3.1 for some time to come.

Intel didn't revitalize Thunderbolt, they just took it off life support. It's still just sitting there doing nothing for the vast majority of people.

 

[phillyboy]

^^ Yeah, you can get optical thunderbolt 2 cables from Corning that will let you go longer (I recently saw a 10m cable), but it goes back to the Thunderbolt 2 issue: Everything is expensive.

From what I have seen Thunderbolt 3 will help adoption in a few ways:

Combined with USB 3.1 on same chipset
Lower chipset licensing costs (I believe this is the case, but am fuzzy on it) so there is no longer a "$100 markup" on any Thunderbolt product
10/20 Gbps link speeds can be done over a *passive* Thunderbolt 3 cable - you only need an active cable for 40 Gbps

I think all three of those will help it each something more mainstream. However, it will most likely be relegated to eGPU support and the newer monitors coming out that support USB-C connectivity for a display / docking input. Other uses such as connecting to a raid box full of SSDs will be more of a niche thing.

 

[defaultluser]

^^ Yeah, you can get optical thunderbolt 2 cables from Corning that will let you go longer (I recently saw a 10m cable), but it goes back to the Thunderbolt 2 issue: Everything is expensive.

From what I have seen Thunderbolt 3 will help adoption in a few ways:

Combined with USB 3.1 on same chipset
Lower chipset licensing costs (I believe this is the case, but am fuzzy on it) so there is no longer a "$100 markup" on any Thunderbolt product
10/20 Gbps link speeds can be done over a *passive* Thunderbolt 3 cable - you only need an active cable for 40 Gbps

I think all three of those will help it each something more mainstream. However, it will most likely be relegated to eGPU support and the newer monitors coming out that support USB-C connectivity for a display / docking input. Other uses such as connecting to a raid box full of SSDs will be more of a niche thing.

But that's the thing: you don't have to have Thunderbolt support in a USB Type C connector. It's using the same alternate carry mode that DisplayPort uses.

And the Apple Thunderbolt Display "docking port" hackjob is a thing of the past as the new C cable can carry FULL BANDWIDTH DisplayPort and USB 2 at the same time (or the DP and USB3 lanes can split bandwidth if you need more accessory bandwidth). Why add Thunderbolt to that mess?

The aggregate data rate for USB 3.1and Thunderbolt over the same cheap passive cable is the same, so why do you think people will want the other?

I think USB 3.1 will get integrated on the Intel PCH much earlier than Thunderbolt, because it's much cheaper to validate 5Gbps per-lane than 10Gbps per-lane. And Thunderbolt will remain a high-end notebook expansion bus. Because the cost of an external PCIe box makes no sense on a desktop.

 

[phillyboy]

However, it will most likely be relegated to eGPU support and the newer monitors coming out that support USB-C connectivity for a display / docking input. Other uses such as connecting to a raid box full of SSDs will be more of a niche thing.

I think USB 3.1 will get integrated on the Intel PCH much earlier than Thunderbolt. And Thunderbolt will remain a high-end notebook expansion bus. Because the cost of an external PCIe box makes no sense on a desktop.

I think we are mostly on the same page, here, unless you need something to be argumentative about.

 

[Skripka]

^^ Yeah, you can get optical thunderbolt 2 cables from Corning that will let you go longer (I recently saw a 10m cable), but it goes back to the Thunderbolt 2 issue: Everything is expensive.

From what I have seen Thunderbolt 3 will help adoption in a few ways:

Combined with USB 3.1 on same chipset
Lower chipset licensing costs (I believe this is the case, but am fuzzy on it) so there is no longer a "$100 markup" on any Thunderbolt product
10/20 Gbps link speeds can be done over a *passive* Thunderbolt 3 cable - you only need an active cable for 40 Gbps

I think all three of those will help it each something more mainstream. However, it will most likely be relegated to eGPU support and the newer monitors coming out that support USB-C connectivity for a display / docking input. Other uses such as connecting to a raid box full of SSDs will be more of a niche thing.

Monitors are pretty much the only thing for such long cables.


Odds are if you're wanting to move large amounts of data between distant points you're either used to LANing it via Cat6 or wifi....or you use a USB dongle. The only circumstances where I like having 3m USB cables are for using a device while charging. Especially since most new construction mandates regular outlet spacing. Working in live productions, there's other scenarios-but all that professional A/V/lighting gear gear tends to be designed to talk over Cat6 cables if not wireless.