-
OT: TBolt eSATA adaptor is pretty darn slick at first glance.
-
[Alan Okey] “Just you wait and see, Apple is going to release a Super Duper Mac Mini Pro Workstation with infinite CPU and GPU power via Thunderbolt expansion. And everyone who buys one will get a free puppy.
/sarcasm*”
Sorry but it’s gotta come with a kitten. Puppies are so “Windows”
Cats & Macs.
And while I’m in a rhymin’ mood.
I want
Power not TowerTwo Proc Xeons and two 16xPCIe doesn’t need to be in a Tower form.
Thunderbolt is 4xPCIe made external and covers quite a bit of use scenarios and can be daisy chained for the most part.
Down the road when Thunderbolt goes completely optical some might argue about the two 16xPICe built in as well. -
I’ve been reading this thread with interest, but just read the subject line more carefully. DUDE! You did NOT say off-topic?!?! I’ve seen maybe 6 threads in the past year that I’d classify as off-topic, and this is definitely not one of them.
Thanks for the thought though. 🙂
Carry on,
TimTim Wilson
Associate Publisher, Editor-in-Chief
Creative COW Magazine
Twitter: timdoubleyou -
[Tim Wilson] “I’ve been reading this thread with interest, but just read the subject line more carefully. DUDE! You did NOT say off-topic?!?! I’ve seen maybe 6 threads in the past year that I’d classify as off-topic, and this is definitely not one of them.
Thanks for the thought though. 🙂
“Tim, just for future reference, is it the discussion of the Thunderbolt adapter or the simple culinary comments that keeps this thread topical for this forum?
-
[Chris Harlan] “I don’t believe thats the case. I could be wrong, but I believe the nature of Thunderbolt requires a single chain like SCSI.”
https://www.intel.com/content/dam/doc/technology-brief/thunderbolt-technology-brief.pdf
“A symmetric architecture that supports flexible topologies (star, tree, daisy chaining, etc.) and enables peer-to-peer communication (via software) between devices.”
Star/tree topologies should allow for hubs.
FireWire technically allows for hubs as well. A few were available, but they were never very popular.
—
Digital Workflow/Colorist, Nice Dissolve.You should follow me on Twitter here. Or read our blog.
-
[Chris Kenny] “[Chris Harlan] “I don’t believe thats the case. I could be wrong, but I believe the nature of Thunderbolt requires a single chain like SCSI.”
https://www.intel.com/content/dam/doc/technology-brief/thunderbolt-technolog...
“A symmetric architecture that supports flexible topologies (star, tree, daisy chaining, etc.) and enables peer-to-peer communication (via software) between devices.”
Star/tree topologies should allow for hubs.
FireWire technically allows for hubs as well. A few were available, but they were never very popular.
“Chris, I interpret this differently than you do. I think you are talking Apples and Oranges, here. But, as I say, I may be wrong. This is my understanding. I believe the “symmetric architecture” alluded to in this brochure exists on the Transport Layer (Apples) and allows “flexible topologies” at that layer, so that you can do things like build the LaCie and Belkin Hubs, and that you can run a variety of different protocols from them, and just generally have things sitting on the line, but that the Physical Layer (Oranges) requires a single lane circuit, which is why the daisy chain. This makes sense to me as these would be extensions of a PCIe lane, and is why I’m doubtful that you’ll ever see a hub, other than hubs for other protocols. Again, my understanding may be incomplete or wrong, but after cursory examination while deciding wether I was going to buy a new Macbook Pro, that’s how I came to understand it.
-
[Chris Harlan] “This makes sense to me as these would be extensions of a PCIe lane, and is why I’m doubtful that you’ll ever see a hub, other than hubs for other protocols.”
PCIe supports star topologies itself — that’s how external PCIe expander boxes like the Magma work. And in any event, Thunderbolt isn’t quite native PCIe — PCIe is just one protocol that can be multiplexed over it. Even if PCIe didn’t support star topologies, Thunderbolt controllers could disguise what was going on from PCIe devices.
—
Digital Workflow/Colorist, Nice Dissolve.You should follow me on Twitter here. Or read our blog.
-
[Chris Kenny] “[Chris Harlan] “This makes sense to me as these would be extensions of a PCIe lane, and is why I’m doubtful that you’ll ever see a hub, other than hubs for other protocols.”
PCIe supports star topologies itself — that’s how external PCIe expander boxes like the Magma work. And in any event, Thunderbolt isn’t quite native PCIe — PCIe is just one protocol that can be multiplexed over it. Even if PCIe didn’t support star topologies, Thunderbolt controllers could disguise what was going on from PCIe devices.
“Again, I think you’re mistaken and are mixing up abstraction layers. I’d like to be wrong. I’d like a hub. The one thing I know you are mistaken about in the above paragraph is “PCIe is just one protocol that can be multiplexed over it.” PCIe and DisplayPort are the ONLY two bus level protocols muxed by Thunderbolt at that abstraction layer. Other protocols run on top of them. In my searches, I’ve found many references to hubs or potential hubs for upper level protocols, but NONE for anything at that level. ALL references at that level are to daisy-chaining. Check out the following Intel brief. Notice the focus on the daisy chain as the mode of connection.
https://www.intel.com/content/dam/doc/technology-brief/thunderbolt-technology-brief.pdf
“With Thunderbolt products, performance, simplicity and flexibility all come together. Users can add high-performance features to their PC over a cable, daisy chaining one after another, up to a total of 7 devices, 1 or 2 of which can be high-resolution DisplayPort v1.1a displays (depending on the controller configuration in the host PC). Because Thunderbolt technology delivers two full-bandwidth channels, the user can realize high bandwidth on not only the first device attached, but on downstream devices as well.
Users can always connect to their other non-Thunderbolt products at the end of a daisy chain by using Thunderbolt technology adapters (e.g., to connect to native PCI Express devices like eSata, Firewire). These adapters can be easily built using a Thunderbolt controller with off-the-shelf PCI Express-to-“other technology” controllers.” -
[Chris Harlan] “The one thing I know you are mistaken about in the above paragraph is “PCIe is just one protocol that can be multiplexed over it.” PCIe and DisplayPort are the ONLY two bus level protocols muxed by Thunderbolt at that abstraction layer.”
Um, yes, I know. How does this make me “mistaken”? PCIe is only one of the protocols that can be multiplexed over Thunderbolt. There’s one other protocol that has that status now, and hypothetically there could be more in the future (though PCIe is pretty general purpose, and multiplexing more protocols would require more complex cross-connects in the host system, so it might not be worth the trouble).
In any event, PCIe’s ability to handle star/tree topologies wouldn’t matter even if it were the only protocol that could be multiplexed over Thunderbolt — the abstraction provided by Thunderbolt would still allow for topologies PCIe didn’t support. (And, again, PCIe actually does support star topologies, so even if Thunderbolt literally were just external PCIe hubs would be possible.)
[Chris Harlan] ” I’ve found many references to hubs or potential hubs for upper level protocols, but NONE for anything at that level.”
Because you’re ignoring a clear reference saying Thunderbolt supports tree/star topologies. Again, look at the PDF I linked to. This is clearly labeled as a transport layer feature. It’s true that nobody has announced a product yet, but hub products weren’t very common for FireWire either — most people’s needs are met by daisy chaining (with technologies that support it), so hubs tend to be rare.
[Chris Harlan] “Notice the focus on the daisy chain as the mode of connection.”
Doesn’t mean very much. FireWire absolutely supports hubs (there were a few on the market), but they were rarely used, and a brief overview of how the technology worked might well ignore the possibility.
—
Digital Workflow/Colorist, Nice Dissolve.You should follow me on Twitter here. Or read our blog.
-
[Chris Kenny] “Because you’re ignoring a clear reference saying Thunderbolt supports tree/star topologies. “
No. I’m not. I’m saying that you are quite possibly misunderstanding that reference. Let me as you this, Chris–since you are so obstinate about such things–after a full year of availability, when a hub would be the easiest, most convenient solution for a majority of Thunderbolt users (instead of a daisy chain with all of its organizational limitations), why doesn’t one exist? Why hasn’t one been announced? It were possible, or at least easily achievable, why isn’t anyone marketing what would be THE most obvious peripheral?
-
[Chris Harlan] “No. I’m not. I’m saying that you are quite possibly misunderstanding that reference.”
But you can’t seem to explain how. What exactly is your interpretation of Intel’s statement that Thunderbolt’s transport layer supports tree/star topologies? The transport layer lives below the layer of the multiplexed PCIe Thunderbolt supports, which puts it two levels down from the kind of multi-interface ‘hubs’ you’re claiming are all Thunderbolt can support.
[Chris Harlan] “Let me as you this, Chris–since you are so obstinate about such things–after a full year of availability, when a hub would be the easiest, most convenient solution for a majority of Thunderbolt users (instead of a daisy chain with all of its organizational limitations), why doesn’t one exist? Why hasn’t one been announced? It were possible, or at least easily achievable, why isn’t anyone marketing what would be THE most obvious peripheral?”
I’ve already addressed this. Hubs are possible on FireWire, but very rarely used. Daisy chaining is sufficient for most users’ needs. Given Thunderbolt’s similar use cases, why would you expect a type of peripheral that has been so unpopular in the FireWire world to be a top priority for vendors in the Thunderbolt world?
It’s also possible (probable, now that I think about it) that while the Thunderbolt protocol supports hubs, a hub isn’t something that can be constructed from the Thunderbolt controllers Intel currently has on the market. By way of rough analogy, the hardware needed to build an ethernet switch is different from the hardware needed to build an ethernet NIC. So hubs may have to wait for new Intel hardware, or for Intel to start licensing third-parties to build Thunderbolt ICs.
—
Digital Workflow/Colorist, Nice Dissolve.You should follow me on Twitter here. Or read our blog.
Reply to this Discussion! Login or Sign Up