MP 7,1 Mac Pro Thunderbolt 3 ports

[bob_stan]

Just want someone with more knowledge than I to confirm there is no major down side to chaining thunderbolt 3 devices. Planning to daisychain two thunderbolt 3 disk arrays and one Thunderbolt 2 array as last item in the chain on one port and 1 4K monitor and 1 1440x2560 monitor on the other.

 

[casperes1996]

It depends is the real answer here. Think of the Thunderbolt connection on your Mac as a motorway. It has a certain number of lanes. You can keep putting cars on the lanes with no issue but if you keep adding traffic eventually you'll get traffic jams where one car/device is hindering the speed of another.
A good example is that eGPUs are usually not recommended to be chained at all because they can use all the bandwidth themselves.
Displays can use a different set of pins on the connector meaning they may not impact anything at all.

Also the devices won't block each other if they aren't actively running. That is aside from slight extra latency from having to move through an extra Thunderbolt controller to get to the device.

In short, depending on the speed of the drives and how much you tax them the difference may be imperceptible but there is definitely a difference. Thunderbolt 3 is fast enough though that it's most likely going to be imperceptible unless we're talking very high end SSD arrays

 

[OkiRun]

It depends is the real answer here. Think of the Thunderbolt connection on your Mac as a motorway. It has a certain number of lanes. You can keep putting cars on the lanes with no issue but if you keep adding traffic eventually you'll get traffic jams where one car/device is hindering the speed of another.
A good example is that eGPUs are usually not recommended to be chained at all because they can use all the bandwidth themselves.
Displays can use a different set of pins on the connector meaning they may not impact anything at all.

Also the devices won't block each other if they aren't actively running. That is aside from slight extra latency from having to move through an extra Thunderbolt controller to get to the device.

In short, depending on the speed of the drives and how much you tax them the difference may be imperceptible but there is definitely a difference. Thunderbolt 3 is fast enough though that it's most likely going to be imperceptible unless we're talking very high end SSD arrays

I sent a private message...

 

[deconstruct60]

Just want someone with more knowledge than I to confirm there is no major down side to chaining thunderbolt 3 devices. Planning to daisychain two thunderbolt 3 disk arrays and one Thunderbolt 2 array as last item in the chain on one port and 1 4K monitor and 1 1440x2560 monitor on the other.

Depends upon the throughput speeds to/from the drive you are looking to get. If these are relatively low speed HDDs in modest numbers then there can be not much of a difference. If relatively high stripe width and in relatively high numbers ( > 10 ) then probably not the best path. Especially on a new Mac Pro where there are at least two pairs by default.

Pragmatically Thunderbolt ports are provisioned in pairs ( off of a Thunderbolt controller. Apple tends to essentially refer to that as a "Thunderbolt bus". ). If you hook the same set of drive arrays to the same TB "bus" then there really is not difference between

TB-bus-1 ---> TBv3_drive1 -> displayPort -> 2k_monitor
|----------> TBv3_drive2 -> TB_4K -> TBv3_to_v2_adapter -> TBv2_drive1

and what you proposed above. You aren't "speeding up" the TB drives by segregating them away from the the video much if they are just 2-4 HDD arrays.

If those were TBv3 SSD drive arrays which each are generating around x2 PCI-e v3 worth of concurrent same direction bandwidth demands then not just daisy chaining is ill advised, but putting them on the same TB "bus" would be too. ( putting on the adjacent port controlled by the same controller being feed through the same host PCI-e bus is a gating factor also).

One issue is that monitors are basically unidirectional but drives can be both direction. So if TBv3_drive2 is primarily being used for reading static source material (and TBv2-drive1 is something like TimeMachine backup ) then there isn't much impact of putting the outbound 4K data stream there ( TimeMachine is relatively slow and it is to a v2 drive anyway).
[automerge]1573053082[/automerge]

...
Displays can use a different set of pins on the connector meaning they may not impact anything at all.
....

That is not how Thunderbolt v3 works. Both display and PCI-e data streams are being sent over the same set of wires. The data is encoded and multiplexed. It is just bandwidth demands by the collection of the devices that primarily matters.

4K (and up) tend to be unidirectional hogs on outbound (from system) traffic. But inbound they are not a problem at all. Thunderbolt is also bidirectional. The bidirectional is on separate pins/wires.

 

[casperes1996]

That is not how Thunderbolt v3 works. Both display and PCI-e data streams are being sent over the same set of wires. The data is encoded and multiplexed. It is just bandwidth demands by the collection of the devices that primarily matters.

Are you sue? I'll gladly stand corrected, but at least with TB2, which of course used the Mini Display port connector, the TB bandwidth was separate from the video stream.
I figured TB3, using a display-alt mode, would similarly not interfere with the data stream.

 

[bnumerick]

Just want someone with more knowledge than I to confirm there is no major down side to chaining thunderbolt 3 devices. Planning to daisychain two thunderbolt 3 disk arrays and one Thunderbolt 2 array as last item in the chain on one port and 1 4K monitor and 1 1440x2560 monitor on the other.

Just an FYI I have 3 TB2 arrays and I had a lot of issues with them plugged into the TB3 port with Apple's adapter with my iMac Pro and 2018 Mac Mini. They kept causing the systems to lockup or reset. Took quite some time to figure out that was what was causing the issues. I ended up just hooking them up to my 2011 iMac and using them for time machine backups.

 

[deconstruct60]

Are you sue? I'll gladly stand corrected, but at least with TB2, which of course used the Mini Display port connector, the TB bandwidth was separate from the video stream.

Not really true for TBv2 either. If go back to the original TB v1 controllers Intel had cut some corners on those implemetations and segregated encoded video onto two channels and PCI-e data onto the other two channels. That was more a restriction on the silicon implementation, not the basic Thunderbolt design/protocol, in implementing Quality of Service and aggregation support. Intel removed that with the next generation.

Thunderbolt 2 removed that.

"...Each channel however is fully independent. Although PCIe and DisplayPort are muxed from the cable perspective, you can only send one or the other over each channel. That limits max performance for a single storage device to 10Gbps (minus overhead), and it similarly limits the max display bandwidth to 10Gbps as well. ...
.....
....Thunderbolt 2 provides that solution. By combining the channels together, Thunderbolt 2 enables two 20Gbps bi-direction channels instead of two sets of 10Gbps channels. There's no overall increase in bandwidth, but the solution is now more capable.

Intel's Thunderbolt 2: Everything You Need to Know

www.anandtech.com

I figured TB3, using a display-alt mode, would similarly not interfere with the data stream.

If in DisplayPort alt mode then the socket is not in TB mode ( TB is another alternative that doesn't go lower than all four channels. It doesn't "share" primarily because the whole protocol is build around shared/mux the channels in the first place. ).

TBv2 and TBv1 likewise had a DP compatiblity mode where you could plug in a DP cable and get just DP throughput. But that isn't really shared concurrently with TB. TB bandwidth and data traffic is basically turned on on that port in that mode.

Type-C has some split grouping channel sets up but those are dead end the same reasons they were dead in back in TBv1. It is not a very efficient use of the bandwidth available when doing multiple high data rate types of steams.