three pages in this article, and still no real explanation of why TB exists. is there any evidence at all that Intel intends something more ambitious than yet another peripheral interface? I had hopes they were trying to do a sort of "IB without the pretenses". but the coverage so far makes it look like a "decommoditization" effort (against USB).
Intel's rationale is that the dual, bi-directional 10 Gbit/s links open up possibilities for faster synching, video transfer and Express outside the PC box--things it says are not possible with the single 5 Gbits/s max USB 3.0.
@ Warren: I think Thunderbolt does what Intel claims, but the other side suggests there are real costs and no killer apps for the 5-10G space it has exclusive claim to in PC I/O. @Iniewski: I will try to ping Intel to jump in here.
The ability to encapsulate PCI Express in a low latency way could very well make or break TB, particularly when they graduate to full Light Peak. USB can go optical too, so won't be bound to 5 Gbps, but will still be USB only. There is a big difference between a high speed i/o, like USB 3.0, and actually extending a system bus out of the box. Consider, for example, long range chip-to-chip interconnects. A cluster system would be much less like a bunch of computers networked together and much more like a single machine that just so happened to have its chips spread out amongst several different physical boxes.
@AlunWang: You are right, the specs need to be open if Intel wants others to join.
@rick.merritt: the 10Gig PHY/controllers can do the same today although their costs and power/thermal problems are still lingering. But volume adoption and better designs are forcing innovations there any way. Today you can AV-Bridge over RJ45. So there are cost effective existing solutions for faster synching and video transfer today!
Some advantages I see for TB is in displayport in lieu of USB. I don't buy the latency argument (where it matters, financial markets for round trip execution times, InfiniBand is firmly entrenched any way!); even Ethernet can reach microsecond latencies with offload engines.
Dr. MP Divakar
@docdivakar, interesting comments on latency...would you be interested in expanding them to a short book chapter? (say, compare TB, USB, InfiniBand etc)...if so pls let me know, email@example.com
A Book For All Reasons Bernard Cole1 Comment Robert Oshana's recent book "Software Engineering for Embedded Systems (Newnes/Elsevier)," written and edited with Mark Kraeling, is a 'book for all reasons.' At almost 1,200 pages, it ...