A Thunderbolt chip inside Apple’s newest Mac Book Pro (image courtesy ifixit.com)
If Intel thought that launching Light Peak would help tamp down the nervousness over its new I/O technology, it certainly isn’t playing out that way.
Light Peak, now dubbed Thunderbolt, was never without controversy but now that it’s finally here, the critics aren’t ready to put away the slings. After its launch, the New York Times opined: Is Thunderbolt Really a Thunderbolt? and questioned its consumer value. Slate wondered if it was a worthless grasp at the past? and questioned why Intel should even pursue wired in an age of wireless. The Financial Times accused Intel and Apple of shunning USB 3.0 to the detriment of consumers while others called it Firewire 2.0 (an allusion to the failure of Firewire to win the standards war).
For those who don’t know, Thunderbolt is essentially one cable to rule them all. A slim, super fast cable and technology that can move 10Gb/s in data in both directions in each of its two channels. Intel is hesitant to oversell it as being capable of 40Gb/s, so 10Gb/s is the public figure but it’s clear higher speeds are possible. Thunderbolt will eventually scale to 100Gb/s if it makes it beyond the gestation phase. Originally planned for optical, Intel ditched it for pricing reasons but said optical cables will come later this year. Whatever the speed, it’s far faster than today’s USB 3.0 implementations and would let you transfer (if your read and write storage sub-systems were fast enough) a 1080p Blu-ray movie in less than a minute.
But if Thunderbolt is so fast, why all the player hating? Several factors play into it. Certainly, some of it comes from Intel’s King Kong-like stature in the industry which always engenders fear from the Fay Wray players in the industry. And despite Intel often pledging that it will support USB 3.0 in future chipsets, many still see Thunderbolt as a direct competitor to a common, low cost standard. The way Thunderbolt is being rolled out doesn’t help either.
When Intel created USB, SATA and many of the technologies it pioneered, the company created standards bodies, licensed its designs royalty free and built coalitions before going forward. There is no standards body for Thunderbolt and no third-party chip. Today, only Intel can make Thunderbolt controller chips and the details of the protocol are still guarded.
That‘s led AMD to croak that despite it being called a standard by some, Thunderbolt is essentially a proprietary technology. AMD has been quoted as forecasting failure for Thunderbolt, but an AMD official Maximum PC spoke to said the harshness in tone from previous stories was overstated. Instead, the official said, AMD is paying close attention to Thunderbolt and how well it does. However, the official pointed out, AMD believes consumers are well served today by USB 3.0 and SATA 6Gbp/s – both of which AMD are behind. And as a method of running external graphics, it’s a bit short in bandwidth as a single lane of the upcoming PCI-E 3.0 spec pretty much eats all of the bandwidth of gen 1 Thunderbolt. Translation: “Meh.”
Most of the acrimony comes from the supporters of the year-old USB 3.0 standard. Ever since Intel showed off Light Peak in 2009, USB 3.0 supporters have accused Intel of dragging butt on native chipset support so Thunderbolt could mature.
Intel, for its part, shrugs off the stones being thrown at it. In an interview with Maximum PC, the company again said that USB 3.0 will be supported in future chipsets and noted that some of the earliest USB devices from 15 years ago still work if you plug them in today.
“We recognize that you don’t throw that kind of eco system under a bus,” said Intel spokesman Dave Salvator. “We are committed to it.”
But, the Salvator said, the company looked forward in time three, five or ten years from now and saw that the amount of media people collect isn’t getting smaller while the resolutions are getting bigger. “We support USB today, we will support it going forward, but Thunderbolt is a way to take it to the next level.”
Why Apple? Apple was the test bed for Thunderbolt because the company was willing to jump in with both feet on it, he said. Despite PC vendors being absent from the Thunderbolt launch, they are interested, Salvator said. For what it’s worth, the major OEMs Maximum PC spoke with declined to disclose whether they will or will not support Thunderbolt. Support from HP, Dell and Acer may not matter though.
One industry analyst who forecast Apple adopting x86 years before it became a reality said the power of Intel plus Apple is like a freight train bearing down on your stalled Yugo.
“Like Microsoft, Intel is big enough to drive a standard alone, with Apple I don’t think they can be stopped,” said Rob Enderle of the Enderle Group. “Apple was the company that drove in a number of technologies including 3.5-inch disks and you could argue they were the big USB driver as well. That was when Apple was far weaker. As long as Apple continues to adopt this broadly it is likely a fait accompli.”
It’s clear there is support for Thunderbolt already. Storage vendors, such as Western Digital, Promise and La Cie are all aboard the T-bolt train as are many video-centric companies such as Avid, Blackmagic and Aja. Motherboard vendors we spoke with also said they will likely integrate Thunderbolt controllers in to professional-class motherboards and possibly enthusiast boards as a value add.
So why no standards body or licensing of the technology to third party vendors yet? Salvator said it’s so early that nothing has been decided yet.
One reason Intel may be taking unilateral action today is for speed. Shane Rau, an analyst with IDC, said by going it alone, Intel can avoid the design by committee roadbumps that can slow technology launches to a crawl. Rau believes that USB 3.0 is not a target, although he said Intel is only begrudgingly supporting it now.
Is Thunderbolt a replay of the DDR vs. Direct RDRAM battle? In that battle, Intel and Rambus’ superior, but revolutionary, RDRAM standard lost to the slower, conventional DDR SDRAM. Rau said no, a better analogy would be Blu-ray vs. Streaming. Blu-ray is the defecto standard today, but we all know that in the end, streaming will win. Rau said the cost issues Intel learned from the RDRAM battles aren’t lost on it nor are the lessons from trying to force revolutionary change on an industry living on razor thin margins.
And cost will certainly be a key issue if Thunderbolt is to eventually supplant USB, Firewire, eSATA and External PCI-E. Intel wouldn’t disclose the costs of the Thunderbolt controllers but it’s likely in excess of $10. For comparison, USB 3.0 host controllers vary from $5 to $2 with costs dropping to free once integrated into chipsets.
“Short to mid-term, USB is going to be here and there is enough opportunity for USB 3.0 to be worthwhile,” Rau said. “But there is not going to be a USB 4.0.”
Rau, however, said that applies to the transport layer, not necessarily the USB protocol. Within three years, 100 percent of new computers will have USB 3.0, he said, so USB will be here for a hell of a long time. What Rau doubts is that a new USB 4.0 transport layer will ever be launched. Instead, a USB 4.0 protocol would likely be built on top of the Thunderbolt infrastructure.
The USB-Implementers Forum, the standards group Intel formed to promote and police USB, may have another word on the matter.
The USB-IF was loathe to criticize Thunderbolt even if many see it as a competing technology but it had no problems highlighting how incredibly successful USB and low cost is today. The group’s president, Jeff Ravencraft, who also works for Intel, said roughly 3 billion USB devices are shipped every year. With the volume, component costs are dirt cheap, the connectors are backwards compatible and the connector was even just declared the standard power charger for all phones in Europe.
Ravencraft said SuperSpeed isn’t standing still either.
“The SuperSpeed USB protocol and host controller were designed to scale up to 25 Gb/s and beyond, so as requirements increase, SuperSpeed USB will be able to scale while getting complete reuse of the protocol and host controller, which means it will continue to be the low cost, high data rate solution now and in the future,” Ravencraft said.
In other words, bring it on.
|External PCI-E||Thunderbolt||USB 3.0||Firewire 800|
|Theoretical Bandwidth||4GB/s (x16 connection)||1.25GB/s (per channel. Two available)||600MB/s||100MB/s-400MB/s|
|Maximum Cable Length||Not specified||3 meters (optical versions will offer “tens of meters.”)||5 meters||100 meters|
|eSATA||USB 2.0||Firewire 400|
|Maximum Cable Length||Not specified||5 meters||4.5 meters|