Mixing TB 2 and TB 1

[Cubemmal]

What happens in these scenarios, searching hasn't found me an answer


nMP->TB2 device->TB1 device

A) Does the TB2 device operate at TB 2 speeds (20 Gbps), and TB 1 at 10 speeds when it gets to the second link (10 Gbps)?

B) Or does the whole chain default to 1? With the prior scenario, that means that the TB1 device (say, a dock with video) gets no mini-dp video signal?


nMP->TB1 device->TB2 device

C) I'm guessing that the whole chain will default to TB 1.

 

[Bones13]

Assumptions only for me.

Any chain of TB devices will be limited by the weakest link.

With 3 TB channels, separating the chains by TB type would make sense.

 

[Cubemmal]

Assumptions only for me.

Any chain of TB devices will be limited by the weakest link.

Depends on how the bus is designed. Its possible that the master queries every device, and if it sees a single TB 1 then it negotiates a TB 1 link with the next in the daisy chain, which propagates down the line.

However that might be a stupid design, as what if you simply have a leaf node (an old Thunderbolt 1 hard drive say), should that then bring down the whole chain? No that makes no sense as it's not using the video signal anyhow. So maybe if you only have a device that listens to the video signal, but is the master sure of that anyhow? Do all the devices report exactly what capabilities they have and use?

Only negotiating with the immediate peer would be the easiest and make the most sense IMO.

 

[Killerbob]

Your assumptions A and C are correct. The bus will be TB2 until it reaches a TB1 device...

 

[Cubemmal]

Your assumptions A and C are correct. The bus will be TB2 until it reaches a TB1 device...

Nice, that's what I hoped to hear. It maintains TB 2 until it hits a TB 1 link. That means that a downstream TB 1 link will NOT get a video signal. Do you have a reference?

 

[xraydoc]

Nice, that's what I hoped to hear. It maintains TB 2 until it hits a TB 1 link. That means that a downstream TB 1 link will NOT get a video signal. Do you have a reference?

I'm not sure that's true (about the video signal). I thought the DisplayPort video signal is carried separately from the PCIe data stream, thus as long as a TB device had a proper outbound port, the video signal will continue to go down the line until it hits a display.

I could be wrong, but I've seen Apple's TB display at the ends of multi-device TB chains.

 

[Cubemmal]

I'm not sure that's true (about the video signal). I thought the DisplayPort video signal is carried separately from the PCIe data stream, thus as long as a TB device had a proper outbound port, the video signal will continue to go down the line until it hits a display.

I could be wrong, but I've seen Apple's TB display at the ends of multi-device TB chains.

TB 1 has four channels, two which are dedicated to video (DisplayPort standard) and the other two as PCIe data for 10 Gbps (actually I think it's 10 G transfers per second technically), both bidirectional (I believe, the DP signal probably isn't in reality).

TB 2 takes the two DP video channels and uses them as additional PCIe data channels to the existing two channels for data, so now its 20 Gbps. It's all data now with no legacy mini-DP video signal. However the Display Port channels are also upgraded to 1.2 which supports 4k screens.

 

[theSeb]

I'm not sure that's true (about the video signal). I thought the DisplayPort video signal is carried separately from the PCIe data stream, thus as long as a TB device had a proper outbound port, the video signal will continue to go down the line until it hits a display.

I could be wrong, but I've seen Apple's TB display at the ends of multi-device TB chains.

Apple's TB device cannot be used as the display port output device. But it won't break the chain.

So

Computer --> Thunderbolt display --> Mini Display Port display (will not work)
but

Computer --> Thunderbolt display --> some other two port device like a hard drive enclosure --> Mini Display port (will work)

 

[Cubemmal]

Apple's TB device cannot be used as the display port output device. But it won't break the chain.

So

Computer --> Thunderbolt display --> Mini Display Port display (will not work)
but

Computer --> Thunderbolt display --> some other two port device like a hard drive enclosure --> Mini Display port (will work)

That's a particular Apple wrinkle - true. Something specific to them, I don't know why they have it that way.

 

[theSeb]

That's a particular Apple wrinkle - true. Something specific to them, I don't know why they have it that way.

Indeed. Also computer --> thunderbolt display --> thunderbolt display (will work)

 

[borad]

It maintains TB 2 until it hits a TB 1 link. That means that a downstream TB 1 link will NOT get a video signal.

You don’t even need TB2 to have a parallel video signal.

I’m currently using Mac Mini --> drive box --> DisplayPort monitor.

 

[Cubemmal]

You don’t even need TB2 to have a parallel video signal.

I’m currently using Mac Mini --> drive box --> DisplayPort monitor.

Obviously, with TB2 you don't even HAVE the video. That's what I was saying, the video channels are taken over for data.

 

[deconstruct60]

TB 1 has four channels, two which are dedicated to video (DisplayPort standard)

Technically not quite right. It is Thunderbolt encoded DisplayPort data. This data is segregated from the Thunderbolt encoded PCI-e data but it is Thunderbolt data going down all four channels.

and the other two as PCIe data for 10 Gbps (actually I think it's 10 G transfers per second technically), both bidirectional (I believe, the DP signal probably isn't in reality).

DP is bi-directional.
" ... A bi-directional, half-duplex auxiliary channel carries device management and device control data for the Main Link, ... "
http://en.wikipedia.org/wiki/Display_Port

The amount of data traffic is skewed heavily to one side but it is still bi-directional.

TB 2 takes the two DP video channels and uses them as additional PCIe data channels to the existing two channels for data, so now its 20 Gbps.

Absolutely not. All it does is not segrated the Thunderbolt data. There is no PCI-e or DP data traveling down Thunderbolt networks. It is Thunderbolt data. It is de-coded at the endpoints back into those other formats but it isn't in the native formats.

What TB2 does is not segregate the data. That has upsides and downsides. If there is "too much " PCI-e or DP data that can 'eat into' the available bandwidth for the other. The TB v1 set up made sure that each of the two flows at least got some minimal level of traffic and that surges in data traffic of the other type didn't impact latency and/or throughput on the other.

TB v2 has to throw more hardware at being able to better juggle and route that two, somewhat competing, flavors of TB data traveling across the TB network. Probably has a bigger transistor budget so not as much of a problem.

TB v2 also has to deai with DisplayPort 1.2 data encoding/decoding. TB v1 cannot. So getting a DP v1.2 signal out to some TB v1 device isn't going to happen.

It's all data now with no legacy mini-DP video signal. However the Display Port channels are also upgraded to 1.2 which supports 4k screens.

DP v1.1 supports 4K screens. Not at 60Hz but it does do 4K screens.

----------

That's a particular Apple wrinkle - true. Something specific to them, I don't know why they have it that way.

It isn't really specific to Apple. It is specific to TB devices that are monitors. Since , Apple is the only vendor of a display docking station it impacts them, but it isn't unique to Apple.

The TB controllers up until now have one TB->DP decode mechanism.

[ Only one DP "out" port. (which needs to be feed back if in by-pass mode.
http://www.tomshardware.com/reviews/thunderbolt-performance-z77a-gd80,3205-4.html ]

If the TB display docking station is using that to pump single to the internal LCD panel there isn't another TB->DP decoder to pump data to the TB socket being utilized in legacy by-pass to DP mode ("backwards compatible
" mode). If the TB device isn't using the decoder (doesn't have a internal panel that needs signal) then the the signal can be sent to the other TB port in legacy by-pass to DP mode.

----------

Indeed. Also computer --> thunderbolt display --> thunderbolt display (will work)

Because what is going out of the first TB display docking station is a TB signal. Not a DP signal. All two port TB devices can pass TB signals out of the "other" TB socket. Would be pretty hard to daisy chain them if they couldn't.

----------

With 3 TB channels, separating the chains by TB type would make sense.

In the context of the Mac Pro 2013, it is more accurate to say there are 3 TB controllers. If segregate TB v1 devices to one controller none of this v1 before v2 or vice versa stuff matters much. [ it is a bigger deal for 2013 MBP models with TB v2 with only one controller. ]

 

[Cubemmal]

Thanks for the informative reply, all good points

TB v2 also has to deai with DisplayPort 1.2 data encoding/decoding. TB v1 cannot. So getting a DP v1.2 signal out to some TB v1 device isn't going to happen.

True, another reason why a leaf TB v1 from a TB v2 won't get a video signal

Absolutely not. All it does is not segrated the Thunderbolt data. There is no PCI-e or DP data traveling down Thunderbolt networks. It is Thunderbolt data. It is de-coded at the endpoints back into those other formats but it isn't in the native formats.

Correct, however what I meant is that there are two channels, and from what I understand you only get the 20 Gbps if there is no video on the link, so both channels are dedicated to data.

In the context of the Mac Pro 2013, it is more accurate to say there are 3 TB controllers. If segregate TB v1 devices to one controller none of this v1 before v2 or vice versa stuff matters much. [ it is a bigger deal for 2013 MBP models with TB v2 with only one controller. ]

The reason I ask is because I'm looking at getting one of those $300+ Corning TB cables, so would like to use only one obviously.

 

[deconstruct60]

True, another reason why a leaf TB v1 from a TB v2 won't get a video signal

There is a "handshake" for DisplayPort connections. It isn't another reason. If the leaf TB devices says it is DPv1.1 it isn't going to be a huge problem for the Mac Pro to send it DPv1.1 signals. Mixing DPv1.2 and v1.1 traffic down one single daisy chain might be an issue, but there are 6 starting points for a Mac Pro. Pragmatically, it isn't a big problem to pass a DPv1.1 (or two) down a single chain with TB v2.

Little reason why

MP2013 -> TBv2DriveEnclosure -> TB-display -> TB-display

would not work.

Correct, however what I meant is that there are two channels, and from what I understand you only get the 20 Gbps if there is no video on the link, so both channels are dedicated to data.

To any one specific host don't get 20 Gbps. The Thunderbolt data transverses at 20 Gbps. That isn't necessarily what gets converted back into the native formats. There is bandwidth on TB for inter-PCIe card/controller data transport. It isn't all necessarily all going to the host.

All the channels are dedictated to data. It is all Thunderbolt data. You distinction of flavors of data don't particularly change the fact they this is all data. Meeting isochronous/latency/bandwidth constraints is really the principle difference between TB v1 and TB v2. The total aggregate data bandwidth is exactly the same.

The reason I ask is because I'm looking at getting one of those $300+ Corning TB cables, so would like to use only one obviously.

It would be odd to put the bulk of the TB connections on the other side of a single "long distance" optical connection. A TB display docking station (with keyboard , mouse/track pad , etc ) and perhaps another TB v1 device on the other end but most stuff that can fully leverage the cable.

If there is some TB v2 device (or 4K monitor) than can saturate the cable then really need to examine if need to additionally daisy chain further what is pragmatically an exhausted cable.