Why Thunderbolt Won't Come to the iPhone Anytime Soon
by Anand Lal Shimpi & Manveer Wasson on January 14, 2012 6:55 PM EST- Posted in
- Smartphones
- Apple
- Intel
- Thunderbolt
- Mobile
- iPhone
- SoCs
Before CES there was a bunch of silliness floating around about how Apple might bring Thunderbolt to the iPhone based on a creative interpretation of an Apple patent. Some folks even theorized that Thunderbolt would bring about faster sync. I meant to address the speculation before CES but I never got the chance to. Now that I'm done with most of what I needed to cover from the show, I can get to what I planned on earlier.
Lightpeak, the basis of Thunderbolt, was an Intel designed spec. When it came time to productize it, Apple and Intel collaborated to turn it into Thunderbolt. According to Intel, the bulk of their technology collaboration with Apple focused on the cable and connector design. The spec itself and the Thunderbolt trademark both remain property of Intel. Apple did some of the initial trademark filing for Thunderbolt, but eventually transfered ownership to Intel.
Intel is the sole owner of the Thunderbolt spec. Building Thunderbolt devices requires a license to use the spec but no royalties need to be paid to Intel. Intel is also the only supplier of Thunderbolt controllers. Without Intel's permission, no other company can make a Thunderbolt controller.
This last point is extremely important. The chances of Intel building a Thunderbolt controller for an ARM platform are very slim. Intel could eventually allow Apple and other companies to make their own Thunderbolt controllers, but that decision is Intel's alone to make.
Furthermore, Thunderbolt requires both DisplayPort and PCIe (x4) interfaces to work. To the best of my knowledge, no modern ARM based SoC (Apple's A5 included) features DP and sufficient PCIe lanes. Apple could admittedly add support in a future SoC, but again, Intel would have to build a controller designed for low power operation in a smartphone or tablet. The cost of the controller would also have to be reduced significantly. Current Thunderbolt controllers cost between $20 - $30, that's just as much if not more than Apple spends on building the A5. Thunderbolt controllers will eventually get down to the single digit dollar values, but not anytime soon.
Thunderbolt supports some extremely high speed signaling, which complicates board design, also contributing to higher costs. The cables are also fairly pricey, which Apple would presumably have to include in the box if TB were to truly supplant USB as the iDevice sync cable of choice. Finally, for today's usage models, Thunderbolt makes no sense to deploy on a smartphone or tablet. The NAND used in high-end mobile devices is 2-bit MLC in a single package of 1 - 8 die depending on storage capacity. The NAND is typically paired with a controller either on-package or integrated into the applications processor (this is how the iPhone works). That controller is good for performance in the tens of megabytes per second, not anywhere close to the 1GB/s that Thunderbolt can offer. Headroom is the enemy of good low power design. Implementing an interface that can transfer at 1GB/s when you're only going to use 5% of that is not good engineering.
Thunderbolt was built to further enable the notebook as a desktop usage model, and to enable notebooks to continue to shrink in size by allowing extra controllers and functionality to be delivered through external devices without compromising performance. Smartphones were never the target for Thunderbolt to begin with.
It's clear that WiFi will be used for both sync and driving external displays at some point in the not too distant future. With 802.11ac we'll likely be able to exceed the performance of NAND, even as ONFI 3.x based NAND arrives later this year and in 2013. Apple has already implemented AirPlay on the iPad and iPhone 4S, enabling some form of wireless display support. As WiFi Direct and WiFi Display become more prevalent however, this will likely be the path to connect smartphones/tablets to external displays.
All modern mobile SoCs include a hardware video encode block that is very low power. It's far more efficient to simply dump the frame buffer into a video, encode it in real-time to H.264 and transmit the low bandwidth signal over WiFi Direct to a display for decode/display. Using Thunderbolt in this case just wouldn't make sense. Mobile devices are all about portability and removing wires, not adding faster, more power hungry ones.
That's what I believe will happen. Apple may eventually move to USB 3.0, but sync and external displays will be done over wireless technologies. Thunderbolt remains a very high performance spec that we'll see limited to notebooks and desktops for the foreseeable future.
Facebook
Twitter
RSS
13 Comments
View All Comments
name99 - Sunday, January 15, 2012 - link
We don't need such a deep analysis.The claim that TB was coming to iPhone soon was always insane.
All we need is the obvious observation that the write speed of iPhone flash is currently 20MB/s --- ie easily served by USB2.
If we are lucky, this will be the year that iPad and iPhone get their flash upgraded, but if that upgrade is to 40MB/s, it's still basically USB2 speed. Meanwhile I imagine this year's Mac lineup will all ship with USB3, and in time iPhone/iPad will migrate to that.
Let's not forget, apart from anything else, that Apple is not in the business of selling iPads and iPhones that won't work with either older macs or the Windows installed base. and it's not like phones have so much space in them that "vanity" connectors, especially connections costing, what, an extra $25 or so, can be thrown in just to make some "political" point.
name99 - Sunday, January 15, 2012 - link
Having said the above, THIS:"With 802.11ac we'll likely be able to exceed the performance of NAND, even as ONFI 3.x based NAND arrives later this year and in 2013. Apple has already implemented AirPlay on the iPad and iPhone 4S, enabling some form of wireless display support. "
is ALSO a pretty dubious statement.
As far as I know, not a single phone vendor supports either MIMO or 40MHz bands for WiFi, meaning they are capped at 72Mbps bit encoding, and so about 5MB/s best case speeds (after MAC effects). To get to anything close to flash we'd need to first 40MHz support, then MIMO support --- and that's for a spec that's been out for what, six(?) years? I think talk of the glories of low-power phone-appropriate silicon for a spec that isn't even finalized yet are a little premature.
tipoo - Monday, January 16, 2012 - link
Lots of people (namely, Appleinsider readers and some other sites like that) seem to think that a thunderbolt connection would magically make the iProducts faster. The flash write speed limits them way before even the USB 2 bottleneck. And besides, Apple is moving to wire-free transfers.