MP 1,1-5,1 Mac OS 11.3 has broken support for older Mac Pros

[Dayo]

Has this been tested on the 5,1 for ≥11.3?

Came up yesterday but not sure if it has been actually tested in a full way:

Mac OS 11.3 has broken support for older Mac Pros

Any hypothesis on why it would seem to be working for everything else apart from cMP 4,1 and 5,1? Notice the working reports are with the OCLP which may have involved some level of Rigid Patching. Anyone confirm with OCLP on a 5,1?

forums.macrumors.com

Mac OS 11.3 has broken support for older Mac Pros

I believe it has options for System Volume Patching of some sort. Might be worth actually running it fully to be 100% sure this does not have an impact. You will need to take a back up first.

forums.macrumors.com

 

[startergo]

Has this been tested on the 5,1 for ≥11.3?

Just tried it. On the first attempt:

On the second attempt I booted to 11.4 beta 1 for the first time:

Hardware Overview:

  Model Name:    Mac Pro
  Model Identifier:    MacPro7,1
  Processor Name:    6-Core Intel Xeon
  Processor Speed:    3.46 GHz
  Number of Processors:    2
  Total Number of Cores:    12
  L2 Cache (per Core):    256 KB
  L3 Cache (per Processor):    12 MB
  Hyper-Threading Technology:    Enabled
  Memory:    128 GB
  System Firmware Version:    9999.999.999.999.999
  SMC Version (system):    9.9999
 
   System Version:    macOS 11.4 (20F5046g)
  Kernel Version:    Darwin 20.5.0
  Boot Volume:    BigSur
  Boot Mode:    Normal

 

[cdf]

On the second attempt I booted to 11.4 beta 1 for the first time:

Interesting. Perhaps something gets carried over from previous boot attempts. This seems inline with my earlier observations:

• The same installation always boots when moved to a supported Mac
• Interestingly, the first boot after moving the installation back is successful

 

[startergo]

Don't get your hopes high. I cannot repeat the success yet.

 

[eVasilis]

Just tried it. On the first attempt:
View attachment 1762916
On the second attempt I booted to 11.4 beta 1 for the first time:

Hardware Overview:

  Model Name:    Mac Pro
  Model Identifier:    MacPro7,1
  Processor Name:    6-Core Intel Xeon
  Processor Speed:    3.46 GHz
  Number of Processors:    2
  Total Number of Cores:    12
  L2 Cache (per Core):    256 KB
  L3 Cache (per Processor):    12 MB
  Hyper-Threading Technology:    Enabled
  Memory:    128 GB
  System Firmware Version:    9999.999.999.999.999
  SMC Version (system):    9.9999

   System Version:    macOS 11.4 (20F5046g)
  Kernel Version:    Darwin 20.5.0
  Boot Volume:    BigSur
  Boot Mode:    Normal

How about consecutive boots?

 

[startergo]

How about consecutive boots?

Not very good. I was using for the first 2 tests NvmExpressDxe.efi.
The tests without it result in an unsuccessful boot. With it I have partial success.

 

[JohnD]

How can this be tested on Apple supported hardware? I can only see a MP7,1 booting BS beta from an external PCIE card with non-apple NVMe drives for testing. Otherwise Apple will not look at a report from an obsolete hardware I guess?

Funny thing is I reported this bug back when 11.3b3 was released. Can't hurt to report it. I said while I'm on an unsupported machine, I'm afraid this will affect the 7,1 as well. Probably won't, but wanted it to get some attention.

 

[cdf]

Can't hurt to report it.

Not sure about that. If the issue only affects unsupported machines, then it might be in Apple's interest to preserve it...

 

[startergo]

Yes we need a report from a supported machine with non-Apple NVMe.

 

[tsialex]

I don't doubt that this NVMe chase here is just a very stupid red-herring. The same problem also happens when you use a PCIe AHCI blade, even Apple made ones.

This is PCIe related, maybe something related to initialisation of the IDT PLX switch. Remember that MP3,1s don't have one…

 

[cdf]

I don't doubt that this NVMe chase here is just a very stupid red-herring. The same problem also happens when you use a PCIe AHCI blade, even Apple made ones.

This is PCIe related, maybe something related to initialisation of the IDT switch. Remember that MP3,1s don't have one…

I concur. 11.4 is unbootable in my case even with no PCIe cards (except for GPU and wireless).

 

[tsialex]

I concur. 11.4 is unbootable in my case even with no PCIe cards (except for GPU and wireless).

I think I know a way to physically disable the IDT PLX switch shutting down the power input to it, I'll take a look at it Sunday.

If without the switch logically present it magically starts to work, then will now for sure.

Btw, lot's of the very high end PC boards also have a PCIe switch, anyone following hackintoshes have read about any problems with 11.3 or 11.4?

 

[Dayo]

I don't doubt that this NVMe chase here is just a very stupid red-herring ... maybe something related to initialisation of the IDT switch. Remember that MP3,1s don't have one…

I don't think there has been any conclusion that the issue is related to NVMe or anything else so it is a bit extreme to use "stupid" (might be a language thing so let's ignore that).

Anyway, the point is to try to find out what the issue is and in the absence of leads, the only option is to identify potential items and eliminate them one by one. First question I asked you yesterday was what is the difference between 3,1 and 5,1 PCIe slot and now that you have brought this IDT switch thing up, is there any suggestion on this angle?

What does IDT stand for btw?

EDIT: Good idea to see what's happening with Hacks.

I concur. 11.4 is unbootable in my case even with no PCIe cards (except for GPU and wireless).

Presumably it boots with just the GPU ... I.E., it is the wireless that's triggering issues?
Reason for query to cross reference to this:

OpenCore on the Mac Pro

Every 11.3 beta version has upgraded and booted fine on my HDD in an HDD sled bay. Every 11.3 beta version has NOT upgrade OR booted fine from my SATA SSA on a PCIe card. It takes several hard power cycles to get the install to take on a PCIe card, and sometimes it boots fine. I'm running 11.4b1...

forums.macrumors.com

Says only has issue when booting from PCIe Slot. Yours presumably doesn't work from SATA when some items are connected to the slots even when not booting from these.

 

[cdf]

I think I know a way to physically disable the IDT switch shutting down the power input to it, I'll take a look at it Sunday.

Nice.

Btw, lot's of the very high end PC boards also have a PCIe switch, anyone following hackintoshes have read about any problems with 11.3 or 11.4?

The only issue I've come across concerns flashed Titan Ridge cards:

Gigabyte Z490 Vision D (Thunderbolt 3) + i5-10400 + AMD RX 580

OpenCore itself needs to detect the type of CPU that is present in the system. This allows OpenCore to apply conditional treatments to different CPU types. Alas, this does not mean that OpenCore gives macOS native support for those CPUs. Only Apple can do that! ok thnx

www.tonymacx86.com

Presumably it boots with just the GPU ... I.E., it is the wireless that's triggering issues?
Reason for query to cross reference to this:

OpenCore on the Mac Pro

Every 11.3 beta version has upgraded and booted fine on my HDD in an HDD sled bay. Every 11.3 beta version has NOT upgrade OR booted fine from my SATA SSA on a PCIe card. It takes several hard power cycles to get the install to take on a PCIe card, and sometimes it boots fine. I'm running 11.4b1...

forums.macrumors.com

Says only has issue when booting from PCIe Slot. Yours presumably doesn't work from SATA when some items are connected to the slots even when not booting from these.

Probably worth looking into. Perhaps @JohnD could let us know if his Mini PCie slot is populated.

 

[tsialex]

I don't think there has been any conclusion that the issue is related to NVMe or anything else so it is a bit extreme to use "stupid" (might be a language thing so let's let that pass).

Anyway, the point is to try to find out what the issue is and in the absence of leads, the only option is to identify potential items and eliminate them one by one. First question I asked you yesterday was what is the difference between 3,1 and 5,1 PCIe slot and now that you have brought this IDT switch thing up, is there any suggestion on this angle?

What does IDT stand for btw?

I've explained to you that Mac Pros with PCIe SATA or PCIe AHCI drives (even Apple made ones) can't also boot 11.3, the only reliable reports of early-2009 to mid-2012 booting successfully 11.3 are from people that is using SATA drives connected to the southbrige SATA ports. Since no one using non NVMe PCIe drives can boot 11.3 reliably, so this eliminate NVMe as the real motive of the crashes.

Sorry, IDT is the manufacturer of the clock generator, I'm always confounding PLX with IDT, it's a PLX switch. I'll correct my previous posts.

All Mac Pros released after the early-2008 model have PCIe switches to overcome the necessity of more PCIe lanes, even the trashcan. Apple used different switches with different upstream/downstream connections over the years:

• early-2009 to mid2012 have a PCIe 2.0 switch from PLX (PLX Technology, Inc. PEX 8608) to share the northbrige 4 remaining PCIe lanes between slots 3 and 4.
• late-2013 have a PCIe 2.0 switch from PLX (PLX Technology, Inc. PEX 8723) to share 8 lanes between all the Thunderbolt controllers and the USB 3.0 ports.
• 2019 Mac Pro has a PCIe 3.0 from PLX, now renamed Broadcom, PEX 8798 to share the 16 lanes upstream to 64-lanes (Slot-2, 4, 5, 6, 7, with pool-A and slot-8+Thunderbolt controllers with pool-B) downstream.

 

[Tachanka]

I still can't get 11.4 to boot on SATA. And this is after I've removed my NVMe Drive as well as my USB C card. The Radeon VII being the only card left in.

 

[Tachanka]

It installed 11.4 curiously just can't boot it. Keep getting the Prohibitory Symbol.

 

[Tachanka]

Why is it necessary to run Big Sur on an older Mac Pro? So you can use a Navi GPU? Mojave supports Vega 20 so you should be able to run anything up to a Radeon Pro VII. That should be more than enough juice to handle most anything. If you need more then you need to get a 7,1 Mac Pro. If you need Big Sur for certain apps, but don't need 7,1 level hardware then a 6,1 Mac Pro would be a good choice. I love mine.

No one here is blaming Apple for not supporting an unsupported Mac. The reason why so many people still use these old machines is because Apple has not made an expandable modular Mac tower with PCIe under $6k since the discontinuation of these Mac Pros. Believe me we would buy newer Macs if we could.



“We are bringing PCI Expansion back to the Mac in a big way” - John Ternus of Apple when introducing the Mac Pro 7,1 at WWDC 2019

But only if you have $6000 lol



The Apple towers of yore were nowhere near as expensive in terms of base price.

 

[Tachanka]

No one here is blaming Apple for not supporting an unsupported Mac. The reason why so many people still use these old machines is because Apple has not made an expandable modular Mac tower with PCIe under $6k since the discontinuation of these Mac Pros. Believe me we would buy newer Macs if we could.



“We are bringing PCI Expansion back to the Mac in a big way” - John Ternus of Apple when introducing the Mac Pro 7,1 at WWDC 2019

But only if you have $6000 lol



The Apple towers of yore were nowhere near as expensive in terms of base price.

Also why the hell is wanting a Mac with PCIe for under $6k so controversial?

 

[Stex]

Also why the hell is wanting a Mac with PCIe for under $6k so controversial?

@Tachanka and @MisterAndrew:

this is a friendly note from a newbie: the issues you raise are interesting, valid, and worth debating. but please do that in a separate thread.

like other commenters mentioned before me, let's keep this thread focused and on topic, please.

Many of us are following this thread closely and are very grateful to the OP and the experts who are generously and proactively trying to identify the problem.

 

[JohnD]

All Mac Pros released after the early-2008 model have PCIe switches to overcome the necessity of more PCIe lanes, even the trashcan. Apple used different switches with different upstream/downstream connections over the years:

• early-2009 to mid2012 have a PCIe 2.0 switch from PLX (PLX Technology, Inc. PEX 8608) to share the northbrige 4 remaining PCIe lanes between slots 3 and 4.
• late-2013 have a PCIe 2.0 switch from PLX (PLX Technology, Inc. PEX 8723) to share 8 lanes between all the Thunderbolt controllers and the USB 3.0 ports.
• 2019 Mac Pro has a PCIe 3.0 from PLX, now renamed Broadcom, PEX 8798 to share the 16 lanes upstream to 64-lanes (Slot-2, 4, 5, 6, 7, with pool-A and slot-8+Thunderbolt controllers with pool-B) downstream.

I assume you're referring to the optical ports as being "directly attached to the SouthBridge", and then is the PLX switch the "SATA/SAS Mux" chip? That chip is responsible for the warm boot issue where drives don't show up, unless cold booted. Same drives always show up when connected to the optical ports.

From what I've understood, the muxer chip is for the Apple RAID card, which allows for SAS drives to connect.

 

[tsialex]

I assume you're referring to the optical ports as being "directly attached to the SouthBridge", and then is the PLX switch the "SATA/SAS Mux" chip? That chip is responsible for the warm boot issue where drives don't show up, unless cold booted. Same drives always show up when connected to the optical ports.

From what I've understood, the muxer chip is for the Apple RAID card, which allows for SAS drives to connect.

View attachment 1762961

Nope. You looked at the wrong part of the diagram. The SATA SAXSAS Mux, just alternate the connection of the drive bays SATA/SAS connector between the southbridge SATA and the SAS from the Apple RAID card installed on PCIe slot-4 and don't do any dynamic switching, it's one or the other.

 

[JohnD]

Perhaps @JohnD could let us know if his Mini PCie slot is populated.

Here's my 11.4b1 report so far. I've been running 11.4b1 the past couple days, nonstop, testing whatever I can. I'm using a spare 2010 Mac Pro with dual X5677's, 48GB RAM, MVC-flashed GTX780, original WiFi/BT, and an LSI eSAS HBA in PCIe slot 3 that's not connected to anything, and drivers don't load in Big Sur anyway.

I have a SATA SSD in sled bay 1, with Martin Lo's OpenCore 0.6.8 package, and a Mojave install. I have a SATA HDD in sled bay 4, which was a CCC clone of my primary machine's 11.2.3 system, upgraded with every 11.3 beta, and is now running 11.4b1. I have a SATA SSD on a PCIe card in PCIe slot 4, which is now a clean install of 11.2.3, upgraded to 11.4b1. Finally, I have 2x 3TB HDD's in a RAID-1 mirror for all testing data, shared between OS's.

I booted 11.4b1 and ran about 14 hours of file copies to/from my FW800 RAID. I've not had any kernel panics or unexpected behavior once booted. I previously copied several TB's of data between my PCIe SATA SSD and the RAID in one of the later 11.3 beta's, also without issue.

Now for the boot reports... I rebooted 29 times from my SATA HDD in sled bay 4 - 26 successful, 3 prohibited symbols. I rebooted 23 times from the SATA SSD on PCIe card - 13 successful, 10 stalls/hangs that led to prohibited symbols.

 

[startergo]

Nope. You looked at the wrong part of the diagram. The SATA SAX Mux, just alternate the connection between SATA and SAS connections and don't do any dynamic switching.

But then anything installed in PCIE 2 should not be affected by the switch correct?

 

[tsialex]

But then anything installed in PCIE 2 should not be affected by the switch correct?

This is not so simple.

First, anything connected to the slots 3 and 4 are downstream to the PCIe switch. Only the slot 1 and 2 and the devices connected to the PCIe lines of the southbridge, like the FW and mini-PCIe, are not directly connected to the same lines of the PCIe switch.

Second, even if you have nothing connected to slots 3 and 4, you still have the switch logically enabled, that's why I'll look at disabling it interrupting the power feed.

2019 Mac Pro had serious problems with a second switch downstream of the first one, it would not be to crazy to think that Apple did something to solve this problem that caused our current one.