Thunderbolt Still Has Problems For Linux

[AnWe]

I seem to have fumbled there, it was meant to read:
Given that USB 3 has about half the bandwitdh of a TB 1 channel, and more overhead as well, how could TB loose?

Not to argue, you are more than free to your opinion, but personally I'm not too worried about DMA attacks, unless the thunderbolt equipment I connected also has network capabilities.
Otherwise someone would need physical access to pull off an attack, in which case I'd be screwed anyway.
I've already got firewire on most of my systems (for DAW use), so Thunderbolt wouldn't make much difference to me in that respect.
(I'd worry more about getting struck by lightning indoors than an intruder with a specialty device trying to access my systems without me noticing )

As for use cases, bandwidth (and perhaps latency in really extreme cases) is the only real limit.
This is extreme (and a bit nonsenical) but you can't do this with USB, no matter what speed you crank it up to.
Thunderbolt is a low-level attachment (almost) directly to the processor, USB is not.

Say for instance you had a specialty PCIe card that you want to use with a laptop. Piece of cake with Thunderbolt. No can do with USB.
A more common use case could be an external monitor for your laptop, that includes lots of PCIe peripherals.
With thunderbolt the performance would be close or on par with a desktop with that hardware built in.
With USB the peripherals would have to be USB devices (limits the choice), and you'd almost certainly have inferior performance, both bandwith and latency-wise.
Going further you might be able to connect your (future) smartphone to a thunderbolt device, and have access to any type of peripheral you could connect to a desktop.

This said, I don't even own a Thunderbolt capable system yet (last time I built one, there were no z87 mATX TB capable motherboards, and I *really* looked for one).
I'm just excited about a technology that hopefully opens new doors to how a system can be built.
I just felt it was sad that GKH, who really should know better, slags it on unfounded grounds (although he has a point on the availability issues).

(And no I'm not an Intel or Apple fanboi if anyone is thinking that.
I couldn't care less what brand my computers components are as long as they do their job well.
As long as I can put Debian on it of course )

/Andreas

[tga.d]

Given that USB 3 has about half the bandwitdh of a TB 1 channel, and less overhead as well, how could TB loose?

As you yourself even said, that comparison isn't valid, and if you're measuring in terms of pure bandwidth, USB 3.1 and Thunderbolt 1 are the same (10 GBps, though Gen 2 doubles this, at least theoretically). As I mentioned in an earlier post, given the limited hardware support of Thunderbolt and it's susceptibility to DMA attacks, I am far more inclined to use USB 3.0. Maybe there are some usecases I'm missing, but I for one am incredibly glad that I only have to deal with Thunderbolt on my desktop and not my laptop.

[AnWe]

Sad

This is truly sad...
So sad that I felt inclined to register here and degrade myself to thread necromancy.

I thougt more of GKH than this.
The fact that the USB 3 vs Thunderbolt review he linked used different SSD's for the comparison makes it severly flawed and completely irrelevant.
The TB drive used a lowly Micron SSD and the USB 3 drive used a Samsung 840 Pro. Essentially they were comapring SSDs, not interfaces.
Try comparing a Quad SSD enclosure via Thunderbolt and USB 3 and I'm confident the picture changes entirely.

However, that is completetly beside the point.
1. If you buy/use Apple hardware and expect it not to be quirky using a third party OS, you have misunderstood the core philosophy of Apple.
They don't care about anyone using anything other than their own products. Specs and standards are only useful to Apple as long as they suit Apples needs.
As long as it works inside their walled garden, they couldn't care less.

Someone will probably (as in this case) come along and fix the issues, but you can never count on that.
Buy Apple hardware at your own peril if you intend to run Linux.
And don't confuse the specific implementation of interfaces on said hardware with the general implementation of those interfaces on any hardware.
GKH being the linux driver guy of course already knows this.
It seems he assumes only Apple will ever release hardware with Thunderbolt ports, which is a bit far fetched.

2. Comparing USB to Firewire or Thunderbolt is a bit ignorant. All three are designed for and are capable of very different things.
Firewire can do stuff that neither USB or Thunderbolt can. ("host"-less bus)
Thunderbolt can do stuff that neither USB or Firewire can.

GKH is well aware of this as he states "Thunderbolt is PCI Express hotplug".
Given that (and the implications of that statement), I don't understand how he fails to spot the obvious flaws in the USB/TB comparison he linked to.
Given that USB 3 has about half the bandwitdh of a TB 1 channel, and less overhead as well, how could TB loose?

I agree that Thunderbolt 1.0 seems to be comercially dead in the (windows/linux) PC market right now.
Try to find a Z87 mATX board with TB for example. There is none.
Even the motherboard manufacturers seem to have missed one of the major points of Thunderbolt.
Why are they otherwise only implenting the port on their top of the range motherboards, that already have all the bells & whistles?
(not to mention more PCI Express slots than you could shake a stick at)
It would be truly useful on laptops, mITX and embedded systems, where there is severly limited room for traditional expansion.

I'm hoping most manufacturers just have been waiting for TB 2 and more innovative peripherals, but I'm not willing to bet the farm on it.
As seen with Firewire, these things are never based on technical superiority alone.
Licensing politics also seem a bit restrictive, but hopefully that is not a bad thing in the long run.
Implementation costs are of course important, and then there's the chicken and egg problem with useful peripherals.

TB 1 is probably dead in the water outside the Apple market, yes.
TB 2 looks more promising with flexible bandwith allocation fixing one of the flaws of TB 1.
("Thunderbolt ready" looks like a band-aid interrim solution. Hopefully the cost of integrating TB directly on mortherboards will come down and this type of hack won't be necessary any more.)

To sum up, I think GKH is wrong.
Granted though, TB is a niche technology for now, but with some very interesting potential use cases.
For example, I'd like to see some embedded systems with Thunderbolt ports (maybe there are some already?).
If they get Thunderbolt 3 just right (enough bandwith for high end external multi-GPU boxes for example), I think Intel is onto a winner.

/Andreas

[uid313]

There is very little need to go above 4GB of RAM at this point unless you like to open tons of big applications all at once which is not something 99% of people do. The (relatively)small amount of storage is because you're supposed to store everything in Google Drive. The display is what makes it so expensive. That's just what happens if you want a good display at the moment. I guess now you could say Google has the same "tax"(high-end hardware) that Apple does.

Developers are advanced users and often do things like virtualization.
Example Windows developers who run Linux in a virtual machine.
Or Linux users that run Windows or another Linux guest.

For virtualization, you preferably do want more than 4 GB RAM.

Or if you're a developer doing Java or .NET then that can take quite much RAM.

[Mafjrk]

Thunderbolt was actually created by Intel. Apple is just the only one using it.

[curaga]

According wikipedia it only allows 3 meter long cables.

Comparable range to IRDA then

[TheBlackCat]

You can use it as a low-latency point-to-point network if you like. Better than 1Gbs ethernet at latency-bound loads despite the nominal 800Mbps speed.

According wikipedia it only allows 3 meter long cables.

[caryhartline]

Thunderbolt is dead. It won't see any adoption.
It is not royalty-free, so it cost much money.
Also it is Intel-owned with no third-party manufacturers, so you have to buy the circuits from Intel, no competition.

Also, USB 3 which is royalty-free is "good enough".
Lets just wait for USB 4.

Also Chromebook Pixel sucks, it costs very much, its expensive and only has 4 GB RAM and 32 or 64 GB SSD. So it is only good as a web browser.

There is very little need to go above 4GB of RAM at this point unless you like to open tons of big applications all at once which is not something 99% of people do. The (relatively)small amount of storage is because you're supposed to store everything in Google Drive. The display is what makes it so expensive. That's just what happens if you want a good display at the moment. I guess now you could say Google has the same "tax"(high-end hardware) that Apple does.

[curaga]

I just got an Intel motherboard with a Thunderbolt port. The fact it has one was mostly seen as a negative for me, since it potentially opens up DMA attacks and I don't really have any use for it. As of now, I don't even have a way of testing if it works, and have no idea what to do with it. I've heard of people gluing theirs shut, but I figure there has to be something it's useful for, right? Other than a display port of course, though even that is useless to me right now.

You can use it as a low-latency point-to-point network if you like. Better than 1Gbs ethernet at latency-bound loads despite the nominal 800Mbps speed.

[erendorn]

New? ...firewire dates back to the very late 80's or early 90's. Obviously, thunderbolt is somewhat new 2010 or 2011..? i can't remember off-hand the exact years...

He was comparing what hindered the firewire adoption in the 90s to what is hindering thunderbolt adoption now, not saying that firewire is new today.