Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Mago

macrumors 68030
Original poster
Aug 16, 2011
2,789
912
Beyond the Thunderdome
According iFixit's teardown, the mp 7,1 instead Intel Titan Ridge tb3 headers it implements 2x Diodes Incorporated PI3DBS16 PCIe Thunderbolt 3 signal mux ( functional equivalent to Intel Titan Ridge).



Digging whiting it's specifications, it's notorious this part is full pcie 4 compatible, which makes more sense seing it inside an Ryzen motherboard.

It could means either Intel titan ridge too expensive, Apple depart Intel as supplier, AMD Zen 2 on the way.

This Chip could implement full tb3 with just 2x pcie4 lines, supports full dp.1.4 (6k/8k) etc.

What you think?
 

tsialex

Contributor
Jun 13, 2016
13,460
13,609
According iFixit's teardown, the mp 7,1 instead Intel Titan Ridge tb3 headers it implements 2x Diodes Incorporated PI3DBS16 PCIe Thunderbolt 3 signal mux ( functional equivalent to Intel Titan Ridge).



Digging whiting it's specifications, it's notorious this part is full pcie 4 compatible, which makes more sense seing it inside an Ryzen motherboard.

It could means either Intel titan ridge too expensive, Apple depart Intel as supplier, AMD Zen 2 on the way.

This Chip could implement full tb3 with just 2x pcie4 lines, supports full dp.1.4 (6k/8k) etc.

What you think?
I think you reading to much on this. Since MP7,1 slots are behind a PLX switch, Apple had to use another solution for TB3 muxing.
 

T'hain Esh Kelch

macrumors 603
Aug 5, 2001
6,485
7,461
Denmark
Interesting!

It may also pave the way for Thunderbolt to go with ARM architecture, which I find more likely than AMD, unfortunately.
 

tsialex

Contributor
Jun 13, 2016
13,460
13,609
This has nothing to do with AMD/ARM/etc, it's just that Apple needed a more robust solution since MP7,1 have so much slots/lanes to be muxed.

That's also why Apple used four PI3PCIE3242 PCIe exchange switches.

Btw, MP6,1 uses a similar, but simpler, approach.
 

tsialex

Contributor
Jun 13, 2016
13,460
13,609
Titan Ridge tb3 header can be behind a pcie switch, as does ASRock with some of it's x570 motherboards (Ryzen 3000s pcie4).
You did not understand my point.

MP7,1 has 8 PCIe slots, Apple have to mux/route TB3/video/audio/USB signals to any of the slots. Titan Ridge can't do that at all since it's just a PCIe device in the chain, that's why Apple used the 6 muxes and the 4 PCIe exchanges switches.
 
Last edited:

Mago

macrumors 68030
Original poster
Aug 16, 2011
2,789
912
Beyond the Thunderdome
Interesting!

It may also pave the way for Thunderbolt to go with ARM architecture, which I find more likely than AMD, unfortunately.
As long I know there is no pcie4 support yet in ARM, considering only AMD Zen2 and IBM power 8 are the only CPU architecture already out and supporting pcie4, I consider safe to bet is an AMD Zen2 based the next Mac platform, it comes sense as AMD also license Zen IP and this IP is also compatible with GloGO 7nn process where Apple owns a number of waffers in excess to keep it CPU supply chain, there's evidence in macOS apple is testing AMD APUs which implies Zen 2 CPUs, add this there is no leaks about Mali, sgi, or Apple GPU support in macOS suggesting Apple is testing a desktop or laptop full ARM, it makes sense for apple to move to AMD CPUs now, as Ryzen Threadripper are now the new Halo CPUs, also Zen CPUs have not suffer the security issues whi9 affected Intel CPU (with performance penalization), AMD Zen2 is not just faster, cheaper, attractive, safer, available, it also maybe customized for Apple by Apple as apple and manufactured by Apple did with ARM, and do not need an new version of macOS as current macOS run on amd64 instructions (AMD licensed it's 64 bit x86 implementation to Intel three decades ago), so no nightmares debugging new kernels, drivers neither having to port applications to an new CPU.

To me not just make sense, it's also a safe bet to believe there is soon a Mac loaded with AMD Zen2 based CPU.

My personal dream Mac Pro-lite would be an trashcan reinstall loaded with 8-16 core zen2 CPU each having a single Navi GPU (5500xt or 5700xt) on board and supporting 256gb ECC RAM it could be as cheap as an iMac 21 and even more powerful than most imMP configurations. Of course it won't happen.

BTW I'll be happy too having an iMac Pro with Threadripper 3 in 24-32 cores with single Vega II GPU or single WX5700 (most likely to happen).
[automerge]1576681827[/automerge]
You did not understand my point.

MP7,1 has 8 PCIe slots, Apple have to mux/route TB3/video/audio/USB signals to any of the slots. Titan Ridge can't do that at all since it's just a PCIe device in the chain, that's why Apple used the 6 muxes and the 4 PCIe exchanges switches.
There is no difference among titan ridge tb3 header and PI3DBS16, both by definition are muxers and muxers are just an logic pcie switch peripheral both have it's own pcie address interrupt etc like another pcie peripheral, muxers only special is they mixes different kind of digital signal data buses among the target interface (as pcie ,USB, display port, audio, DC power etc).

Take a minute on Google at least educating yourself about.
 
Last edited:

tsialex

Contributor
Jun 13, 2016
13,460
13,609
As long I know there is no pcie4 support yet in ARM, considering only AMD Zen2 and IBM power 8 are the only CPU architecture already out and supporting pcie4, I consider sage to bet is an AMD Zen2 based the next Mac platform, it comes sense as AMD also license Zen IP and this IP is also compatible with GloGO 7nn process where Apple owns a number of waffers in excess to keep it CPU supply chain, there's evidence in macOS apple is testing AMD APUs which implies Zen 2 CPUs, add this there is no leaks about Mali, sgi, or Apple GPU support in macOS suggesting Apple is testing a desktop or laptop full ARM, it makes sense for apple to move to AMD CPUs now, as Ryzen Threadripper are now the new Halo CPUs, also Zen CPUs have not suffer the security issues whi9 affected Intel CPU (with performance penalization), AMD Zen2 is not just faster, cheaper, attractive, safer, available, it also maybe customized for Apple by Apple as apple and manufactured by Apple did with ARM, and do not need an new version of macOS as current macOS run on amd64 instructions (AMD licensed it's 64 bit x86 implementation to Intel three decades ago), so no nightmares debugging new kernels, drivers neither having to port applications to an new CPU.

To mi not just make sense, it's also a sage bet to believe there is soon a Mac loaded with AMD Zen2 based CPU.

My personal dream Mac Pro-lite would be an trashcan reinstall loaded with 8-16 core zen2 CPU each having a single Navi GPU (5500xt or 5700xt) on board and supporting 256gb ECC RAM it could be as cheap as an iMac 21 and even more powerful than most imMP configurations. Of course it won't happen.

BTW I'll be happy too having an iMac Pro with Threadripper 3 in 24-32 cores with single Vega II GPU or single WX5700 (most likely to happen).
The only thing that this show is that you really would like Apple to move to AMD CPUs, not that Apple will do it.

Apple having to use a more robust solution for TB3 signal routing/muxing proves nothing AMD or ARM related, just that Intel can't offer a ready made solution for an Apple problem.

Apple has been using PCIe switches and muxes with Mac Pro for years, the only news is the MP7,1 complexity.
 
Last edited:

tsialex

Contributor
Jun 13, 2016
13,460
13,609
There is no difference among titan ridge tb3 header and PI3DBS16, both by definition are muxers and muxers are just an logic pcie switch peripheral both have it's own pcie address interrupt etc like another pcie peripheral, muxers only special is they mixes different kind of digital signal data buses among the target interface (as pcie ,USB, display port, audio, DC power etc).

Take a minute on Google at least educating yourself about.

Apple designed route/muxing in a way that any MPX slot can route the signals, Apple MPX GPUs have fully functional TB3 buses and output to the Apple I/O card/slot-8 too. Please show me how a Titan Ridge controller can route/mux signals for the Mac Pro configuration?
 
Last edited:
  • Like
Reactions: thisisnotmyname

Mago

macrumors 68030
Original poster
Aug 16, 2011
2,789
912
Beyond the Thunderdome
Apple designed route/muxing in a way that any MPX slot can route the signals, Apple MPX GPUs have fully functional TB3 buses and output to the Apple I/O card/slot-8 too. Please show me how a Titan Ridge controller can route/mux signals for the Mac Pro configuration?

There is nothing special in mpx module, it has an extension to connect Display port 1.4 signals from the GPU back to the motherboard where the muxer (aka tb3 header) muxes with pcie signals etc and route back to the mpx module the actual tb3 signals to it's physical usb-c/tb3 ports, almost exactly as on the trashcan but having the TB ports in the same module instead an I/o module.

Apple has been using PCIe switches and muxes with Mac Pro for years, the only news is the MP7,1 complexity.
No complexity, actually it's an bold simple clean sustainable Orthodox design. If you see complexity on it, simple you're not an hw designer, read at iFixit's comments at least.
 
Last edited by a moderator:

tsialex

Contributor
Jun 13, 2016
13,460
13,609
@tsialex do you think the base i/o 7.1 card could be tested without risk on our 5.1 since it looks like the thunderbolt controler is on the card?
I don't think so, Apple warned that the I/O Card can only be used with slot-8. For now, seems that the I/O Card is not a standard PCIe x4 card at all (see the photo), seems to have a lot more proprietary pins. Probably needs several signals like USB 2.0 from the chipset (for USB audio?) and others from the muxes on the logic board, maybe even UARTs too.

We will know more about the I/O Card when the MP7,1 Apple Technician Guide, or the block diagram, leaks.

o61yEMRCMOMmqjHX.huge
 

edgerider

macrumors 6502
Apr 30, 2018
281
149
Definitely won't fit a MP5,1, it's a proprietary slot, see the photo:

View attachment 883568
smart mofos!????
it looks like even the base model with 8 core and rx580 will outperform a maxed out twin x5690 almost in almost evry domain, i’ve put all my **** for sale, and see how it goes.
i’m definitely going for the rack version if i have to buy one.
the added benefit for me is that I no longer need all my pcie extension chassis.
 

bsbeamer

macrumors 601
Sep 19, 2012
4,313
2,713
Based on the released Apple support documents, it APPEARS that updated I/O cards may be available in the future, or at least that is the initial plan.

The question in another thread is about Afterburner "fitting" in an MP5,1. Even if it fits within dimension, do not believe it will work or be recognized.
 

edgerider

macrumors 6502
Apr 30, 2018
281
149
@bsbeamer , yeah and i think we are beating a dead horse now...
I now have 5 maxed out 5.1 that i am going to sale for 1000€ and shelf the extra 3/4000 to buy a rack version
 

atonaldenim

macrumors regular
Jun 11, 2018
239
316
I don't think so, Apple warned that the I/O Card can only be used with slot-8. For now, seems that the I/O Card is not a standard PCIe x4 card at all (see the photo), seems to have a lot more proprietary pins. Probably needs several signals like USB 2.0 from the chipset (for USB audio?) and others from the muxes on the logic board, maybe even UARTs too.

We will know more about the I/O Card when the MP7,1 Apple Technician Guide, or the block diagram, leaks.

o61yEMRCMOMmqjHX.huge
Damn, I was really wishing for a 7,1 I/O card in my 5,1!
 

Petri Krohn

macrumors regular
Feb 15, 2019
114
124
Helsinki, Finland
The Diodes Incorporated chips have nothing to do with sourcing a Thunderbolt interface. They are analog devises that switch differential pairs between inputs and outputs. They do not care if the signals are PCIe, DisplayPort or Thunderbolt. The PI3DBS16-12 switches PCIe lanes while the PI3PCIE3242A switches DisplayPort lanes.

The basic Mac Pro has a total of four Thunderbolt 3 ports. Two of them may originate from the Apple interface card. Each Thunderbolt port needs a DisplayPort video signal to be passed through.

The Radeon Pro 580X provides four DisplayPort outputs that are directed to the extended connector in the MPX slot. Each slot has four DP inputs. The user can place the 580X is either one of the slots, so a switch is needed.

The Radeon Vega II modules have only two DP outputs. If two Radeon Vega IIs are in use, then the four internal Thunderbolt ports must take their DP video feed from two ports from each of the slots. Here again a switch is needed.

Each PI3PCIE3242A has two high-speed differential inputs and two outputs. Depending on the select pin, the signals either go straight through, input A to output A and input B to output B. With the select signals in the other position input A crosses to output B and vice versa.
 

deconstruct60

macrumors G5
Mar 10, 2009
12,493
4,053
According iFixit's teardown, the mp 7,1 instead Intel Titan Ridge tb3 headers it implements 2x Diodes Incorporated PI3DBS16 PCIe Thunderbolt 3 signal mux ( functional equivalent to Intel Titan Ridge).
....

IFixit's teardown of the Mac Pro was rather weak. It showed nothing about the Thunderbolt ports at all. Nothing. They did not do a chip analysis of the I/O card. They didn't even show the I/O card embedded in the top of the case barely at all ( left alone a chip analysis). They should the ribbon cable connector on the main board to the 'top' I/O board but not the board itself.

As outlined above those Muxes are about moving DisplayPort from the MPX cards to the rear and top Thunderbolt controllers. There are not replacements for the Thunderbolt controllers at all.

Being able to put dual DisplayPort streams from optional two different MPX connectors to the top port means Apple needs a couple of switches there. Same with being able to put one DP stream to the back and one to the top but also pushing those two only to the back means another couple of DP switches.

MacPro_DP_Switching.jpg
MacPro_DP_Switching.jpg
 
  • Like
Reactions: thisisnotmyname

Mago

macrumors 68030
Original poster
Aug 16, 2011
2,789
912
Beyond the Thunderdome
Disagree with both about Diodes Inc. Muxers, by definition Titan Ridge headers are muxers, it has nothing to do with port distribution, please take your time reading the specifications from the link above.
 

AidenShaw

macrumors P6
Feb 8, 2003
18,667
4,677
The Peninsula
IFixit's teardown of the Mac Pro was rather weak. It showed nothing about the Thunderbolt ports at all. Nothing. They did not do a chip analysis of the I/O card. They didn't even show the I/O card embedded in the top of the case barely at all ( left alone a chip analysis). They should the ribbon cable connector on the main board to the 'top' I/O board but not the board itself.

As outlined above those Muxes are about moving DisplayPort from the MPX cards to the rear and top Thunderbolt controllers. There are not replacements for the Thunderbolt controllers at all.
In other words, the MP7,1 could have been the Mac Pro 2016 - except for the brain-dead decision to require DisplayPort tunneled over T-Bolt. (The key word is *require*.)
 
  • Like
Reactions: ssgbryan

deconstruct60

macrumors G5
Mar 10, 2009
12,493
4,053
In other words, the MP7,1 could have been the Mac Pro 2016 - except for the brain-dead decision to require DisplayPort tunneled over T-Bolt. (The key word is *require*.)

Those are your made up words; not mine. Nor inferred from Apple's design constraints.

DisplayPort 2.0 is going to moving to Thunderbolt tech. It is part of the standard.


"... The increased bandwidth is the result of VESA leveraging Thunderbolt 3's physical layer. ..."


"..Speaking of the link itself, the move to Thunderbolt 3 technology also means that DisplayPort inherits Thunderbolt 3’s signal encoding scheme. ..."

If Intel (and Apple ) had not been working with VESA/DisplayPort folks all along there probably wouldn't have been this merging. Excluding DP years back would have done Thunderbolt no good long term. Hardly brain dead. Just not myopic and backward rearview mirror looking.
 
  • Like
Reactions: thisisnotmyname
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.