IDF 2014: Where is Thunderbolt Headed?
by Ganesh T S & Mahendra Lodha on September 14, 2014 7:45 AM EST- Posted in
- IDF 2014
- Intel
- Thunderbolt
Coverage of Thunderbolt devices has been expanded on our site over the last few months. At IDF, we took the opportunity to chat with Intel about where Thunderbolt was headed.
The current market perception is one of Thunderbolt being relevant only to Mac users. But, a look at the products that Intel showcases, indicate that there are plenty of PC components (motherboards as well as workstations) that come with the technology integrated. Thunderbolt users on Windows have traditionally found that the technology doesn't live up to its advertised potential. Common complaints include
- Purchased peripherals don't carry certification on Windows
- Hot plugging peripherals doesn't work reliably
- Performance in terms of both bandwidth and latency end up being better on Mac compared to Windows for the same workloads
These issues have turned out to be a vicious circle - Mac users end up getting targeted with more Thunderbolt peripherals (for example, storage manufacturers pre-format their devices in HFS+ format), and this, in turn, lowers the appeal of these devices to Windows users. Irrespective of one's personal preference, it is an undeniable fact that Windows still rules the desktop and notebook market by a big margin. Without extensive adoption on the PC side, there is no doubt that Thunderbolt would go the way of FireWire, a technology that slowly faded into oblivion because Apple was the only vendor who invested in promoting it.
Expanding Thunderbolt's Reach with PCIe Expansion Cards
In the process of migrating from the Z77 / Z87 to the Z97 chipset, Intel tried to drive up Thunderbolt adoption by allowing motherboard makers to provide support via an add-in card. This kept the price of the motherboard low by avoiding the cost of integrating and verifying the Thunderbolt ASIC. The board layout only had to support a Thunderbolt header. The routing of the Display Port was also done externally.
Only interested consumers needed to pay the premium for the interface. In their motherboard segmentation plans for Thunderbolt, Intel intended these Add-In Cards (AICs) only for the X79 and H87 / H97 / Q87 chipsets. Z87 and Z97 motherboards were supposed to have Thunderbolt silicon on board. Unfortunately for Intel, many motherboard makers (including Asus) decided that the add-on card would be the best way to go forward, and didn't release any Z97 boards with integrated Thunderbolt support. This meant that the DIY market, for the most part, completely ignored Thunderbolt.
Thunderbolt on PCs: A Crippled Experience
The reason for the far from optimal experience with Thunderbolt on PCs boils down to two different aspects, the hardware and the software. In terms of hardware, Intel has never allowed motherboard vendors to hang the Thunderbolt silicon / add-in card off the CPU's PCIe lanes. These have to hang off the platform controller hub (PCH). On the other hand, Apple was allowed to hook up the Thunderbolt silicon directly to the CPU. The reason behind this leads us to the software side of things.
Apple has full control over the operating system. Hanging Thunderbolt peripherals directly off the CPU's PCIe lanes requires extensive support from the operating system, particularly when it comes to hot plugging devices and/or waking up peripherals from sleep mode. Over the PCIe lanes off the PCH, Intel has more control via its chipset drivers. Ultimately, it looks like Microsoft dropped the ball and Intel decided to come up with a certification solution by only allowing Thunderbolt silicon to talk to the PCH for all PC boards.
While Microsoft continues to twiddle its thumbs, Intel has decided to come up with less restrictive hardware suggestions to bridge the Thunderbolt experience gap between Macs and PCs.
Thunderbolt's Future - X99 Brings Promise, Driver Features Add Utility
The X99 platform's Thunderbolt capabilities were barely touched upon in the initial Haswell-E reviews. Part of the problem was that none of the motherboards from major vendors had support with on-board silicon. However, it is actually a very important chipset launch for the Thunderbolt on PCs ecosystem because Intel has finally allowed the AICs to hang off the CPU's PCIe lanes. A restriction is that the AIC must have support for a switch to disable the sleep mode for all the Thunderbolt devices in the chain, though this could conceivably go away if Microsoft fixes Windows for this issue.
In terms of updates to Thunderbolt itself, we are yet to see a move to PCIe 3.0. However, the drivers have been updated to enable 'Thunderbolt Networking'. This involves linking multiple PCs / Macs with Thunderbolt cables. A 10 Gbps network is automatically created (in the form of a 'dummy' network adapter). Macs and PCs can talk with each other to share printers and folders. For small workgroups, this could be an effective way to achieve 10 Gbps networking without the costly and noisy switches.
In conclusion, Thunderbolt is making great progress and PC users can expect things to get better in the future. In addition to Apple, lots of other vendors are also throwing in support for Thunderbolt in their workstations and notebooks. It would have been great to have a new version of Thunderbolt with PCIe 3.0 along with the X99 launch. But, we already know it is not going to be the case till Skylake launches. That said, it will be a priority for Intel and Microsoft to get the performance and experience right with current silicon for now.
Facebook
Twitter
RSS
65 Comments
View All Comments
DanNeely - Monday, September 15, 2014 - link
Both do have an auxillary data sub-channel capable of carrying USB2. I don't think it ever got widespread adoption because it was an optional part of HDMI so OEMs couldn't count on it being there; I suspect the same, possibly combined with the monitor (GPU?) vendor needing to provide a driver, scuttled implementation on the DP side.USB3/3.1 would have the same problems; combined with being enough of a bandwidth pig that only the newest versions of the ports have the bandwidth needed to support it at mainstream resolutions. None of them could do it at their headline resolutions; and not offering it there would lead to widespread wailing and moaning from a chattering class much larger than the few of us who have 4/5k displays (@120hz for DP 1.3) that would be impacted.
repoman27 - Monday, September 15, 2014 - link
Intel slides regarding the Alpine Ridge Thunderbolt 3 controller indicate a new connector will be used, and a USB 3.0 signaling mode will be added to the mix along with DP 1.2, HDMI 2.0 and up to 100W power delivery. Looking over the USB 3.1, Power Deliver 2.0, and Type-C Cable and Connector specs, I get the feeling that the Type-C connector will also serve as the next Thunderbolt connector. One port to rule them all...Zertzable - Sunday, September 14, 2014 - link
Where is Thunderbolt headed? Nowhere. In a few years TB and FireWire will throw a party together.Morawka - Sunday, September 14, 2014 - link
its sad you guys didn't bring up the Price factor more in the article. When OEM's make anything thunderbolt related, they try and charge apple tax instead of a healthy 30% margin like everything else.Thunderbolt External SSD? HAH lets add $100 instead of $.89 cents, which is the cost difference from going to usb 3.0 to thunderbolt.
repoman27 - Monday, September 15, 2014 - link
The cost difference between implementing Thunderbolt and other consumer I/O interfaces is a *lot* more than 89¢. That being said, something clearly doesn't jibe with the retail pricing of Thunderbolt devices and the controller pricing Intel claims on ARK, and I'd love for someone to report on that in more depth.JDG1980 - Sunday, September 14, 2014 - link
I just don't see Thunderbolt having a future. External video cards have never been officially supported and Intel clearly doesn't want them to be supported. Video output can be handled with DisplayPort or mini-DP (1.2 can push 4K@60Hz), and everything else can be connected through USB. With USB 3.1 having reversible connectors, 10 Gbps throughput, and even charging capabilities, where's the niche for Thunderbolt?The fact that everything related to TB costs an arm and a leg certainly doesn't help, either.
repoman27 - Monday, September 15, 2014 - link
Thunderbolt solves a very particular problem for one of Intel's better customers. Since Apple decided to entirely stop making PC's with user accessible PCIe slots, they need Thunderbolt for the corner cases where USB alone cannot meet 100% of the expansion needs of their user base. Until Apple changes course or stops using Intel CPUs, Intel will continue to sell tens of millions of Thunderbolt controllers to them.Thunderbolt, like FireWire before it, is actually quite popular with the professionals who can actually leverage the additional bandwidth. For the time is money set, there's a value proposition to be had with Thunderbolt devices despite sticker prices that appear outrageous to the casual observer.
kyuu - Monday, September 15, 2014 - link
Yes, but being popular with a small population of "professionals" isn't good enough. Without broader acceptance, Thunderbolt is doomed to die just like Firewire before it as competing protocols catch up with its capabilities while boasting the very important advantages of cost and compatibility with what everyone else is using.Hell, USB3.x is already going to have a significant chunk of Thunderbolt's bandwidth. Combine that with the ability to supply higher wattage, and there's little reason for people to even consider the wildly expensive and poorly supported Thunderbolt standard. Outside of a small group of professionals, of course.
repoman27 - Monday, September 15, 2014 - link
Actually, it is. FireWire didn't just die because not enough people used it. It was used extensively over the past 15 years, became an essential technology for many market segments, and played a pivotal role in the advancement of certain product categories. In 2011-2012 it was largely displaced by both USB 3.0 and Thunderbolt. There will always be customers for technologies that offer more than the lowest common denominator, and they will generally provide way better margins for the OEMs that cater to them. They will also unceremoniously ditch old standards as soon as something comes along that helps them get their job done faster.The whole point of Thunderbolt is to offer the fastest form of serial I/O possible in a mass-market PC. Most people will never use it, but then again, the majority of consumers have also never bought a PCIe add-in card or ExpressCard either. Thunderbolt has been offering 5x the bandwidth of USB 3.0 and 2x that of USB 3.1 since 2011. It will double to 40 Gbit/s per port and inherit USB PD 2.0 style power delivery modes with Alpine Ridge which will accompany Skylake. USB 3.1, on the other hand, won't see Intel integration and thus ubiquity until Cannonlake arrives.
USB will keep getting faster, but it will always be the kind of fast that is also saddled with the requirement to remain cheap and ubiquitous. There will always be a way to design a faster interface if you don't care as much about price and compatibility, and there will always be people who are willing to pay for that kind of performance right now. You can't will away high-margin niche markets. It's not like NVIDIA and AMD are going to stop making $1000+ high-end GPUs or workstation cards just because most people don't buy them.
Spunjji - Wednesday, September 17, 2014 - link
"USB 3.1, on the other hand, won't see Intel integration and thus ubiquity until Cannonlake arrives."Call me curmudgeonly, but I wouldn't call Intel's continued conflict-of-interest based delays on implementing USB technologies an argument /for/ Thunderbolt so much as yet another one against it.
I broadly agree that this just isn't meant to be a mass-market technology though, so decrying it for not being one is not a great argument to make. I do wish they'd start catering to the external GPU market, though.