Updating a Mac (Pro)s CPU Microcode

[highvoltage12v]

interesting results

an iMac10,1 is very much a machine I wish to add to my collection and play with (especially as I have a large number of LGA775 CPUs and LGA771 CPUs id like to test in one)

did you try the Core 2 Quad before modding the BootROM? its worth checking what it does before you start fiddling with things so you have a baseline so to speak

its interesting to see that it does not POST, from what little information there is on the matter. and from some own personal poking around at an ioreg kindly provided to me by @AphoticD the iMac10,1 should at least post with a C2Q however only 2 cores will be detected/usable (because of the APIC tables in BootROM)

Yes I tried it stock and then again by adding 10677 and finally by updating the stock 1067a microcode to 2015 1067a. I saw someone on ifixit having a q6600 in their 10,1 but they only had 2 cores working. I doubt the microcode for kentsfield will fix that.

 

[LightBulbFun]

Yes I tried it stock and then again by adding 10677 and finally by updating the stock 1067a microcode to 2015 1067a. I saw someone on ifixit having a q6600 in their 10,1 but they only had 2 cores working. I doubt the microcode for kentsfield will fix that.

looking online the 8200S is a R0 stepping CPU so CPUID 1067A so thats the microcode it would potentially use

BTW did you check if the iMac still booted with its stock CPU on the updated Microcode BootROM?

it would be very interesting to try something like a X5270, its the Xeon LGA771 equivalent to the unreleased Core 2 Duo E8700, still Dual Core but 3.5Ghz

(for this CPU you probably will have to add the right microcode for it as it uses a different platform ID)

(I have a C0 10676 QX9770 here it would be fun to try that in an iMac10,1 )

 

[TheSynthax]

Not Mac Pro related, but I'm considering a similar project with an iMac 2011 and a 3770K. Would you happen to know if that CPU is physically supported or not? The socket and chipset should support it, though no one has had success in booting with one, so I'm assuming it's due to lack of microcode. Also on a separate topic, would you have any idea about an EFI mod or a vBIOS mod to enable boot screens?

 

[LightBulbFun]

while the hardware in theory (Z68 chipset) is compatible with an ivy bridge CPU

theres more to it then just that and microcode, it seems the Firmware has to support the CPU in general first

its worth a shot if you have the means to recover from a hardware brick, but I wouldn't bank on just adding microcodes to get ivy bridge working in a 2011 iMac

 

[highvoltage12v]

Not Mac Pro related, but I'm considering a similar project with an iMac 2011 and a 3770K. Would you happen to know if that CPU is physically supported or not? The socket and chipset should support it, though no one has had success in booting with one, so I'm assuming it's due to lack of microcode. Also on a separate topic, would you have any idea about an EFI mod or a vBIOS mod to enable boot screens?

The way I have gotten around the EFI boot screen issue was taking an EFI Flashed 780TI putting it in an akito node and i have EFI boot screen over Thunderbolt 1.0. Since my iMac also has a 780m in it, the frame buffer loads for both cards and I can use them simultaneously. Also discussed on the iMac GPU upgrade thread with @Nick [D]vB Even if changing Microcodes allowed the iMac to boot, Intel's ME will be broken and the iMac may not load the OS fully or be slow/laggy if it does. This is all just assumed information and has not been tested.

Link to my discovery of EFI Boot screens over thunderbolt. I use this setup daily and have since water-cooled the 780ti.
https://forums.macrumors.com/threads/2011-imac-graphics-card-upgrade.1596614/page-56#post-26905493

 

[dlxmax]

Today I tried to make a Core 2 Quad Q8200S (65w-tdp) work in a 2009 iMac 21.5"...Any other ideas before I give up on this?

I really want to see you succeed here, partially because I've got the same machine in my faculty office, and it's been feeling sluggish even with an SSD and 8gb just running productivity apps. I've been contemplating putting linux or Chromium on it...

Anyhow, have you tested that new cpu in another system to verify that it works?
That's all I can think of for the moment...

 

[dlxmax]

Further thoughts for highvoltage12v...

We should assume there's no code in the ACPI for quad core processors and it'll have to be added.

DFI put out a motherboard with the same chipset, and it supported quad core. This bios can potentially donate the ACPI code you need: https://drivers.softpedia.com/get/BIOS/Dfi/DFI-LANPARTY-JR-GF9400-T2RS-Bios-N7EDC01.shtml

Your iMac bios can possibly be manipulated with standard bios tools. It's then just a matter of determining which module needs to be copied over.

 

[highvoltage12v]

I really want to see you succeed here, partially because I've got the same machine in my faculty office, and it's been feeling sluggish even with an SSD and 8gb just running productivity apps. I've been contemplating putting linux or Chromium on it...

Anyhow, have you tested that new cpu in another system to verify that it works?
That's all I can think of for the moment...

My Q6600 came from my old p5n-d, the q8300 came from the p5k-d and the q8200s works fine. The 2009 iMac did a security update which installed firmware update installing 132.0.0.0.0 which undid all the changes i did including adding NVMe.dxe to the ROM, and the one Microcode that was added. I'm burnt out on this topic, I ended up installing an e8600 and putting the iMac back together.

 

[dlxmax]

The 2009 iMac did a security update...which undid all the changes i did...

That's dispiriting, though support and updates for High Sierra are liable to cease around September next year, when this iMac series should be transitioned to a different operating system.

Could you post your firmware dump? I'd like to play around with it.

 

[handheldgames]

Hmmm. Based on the forum rules we all love to follow... This thread seems to have drifted off topic from the mac pro. Perhaps it belongs over in the imac forum or at tonymacx86.

 

[LightBulbFun]

Hmmm. Based on the forum rules we all love to follow... This thread seems to have drifted off topic from the mac pro. Perhaps it belongs over in the imac forum or at tonymacx86.

well I started this thread updating the microcode on a MacBook2,1 soo do you want me to move this entire thread over to the MacBook forum? LOL

 

[highvoltage12v]

That's dispiriting, though support and updates for High Sierra are liable to cease around September next year, when this iMac series should be transitioned to a different operating system.

Could you post your firmware dump? I'd like to play around with it.

Yes, tomorrow I'll upload the latest dump (I actually had to use El Capitan because with the latest HS it would freeze dumping) I just need to edit the serial number out of my ROM before posting. That bios link you posted ended up going to a 403 Forbidden location unfortunately. Also I think we have 2 more years (a little less now) of security updates. It's 2 years after the 1st mainstream year.

 

[tsialex]

Yes, tomorrow I'll upload the latest dump (I actually had to use El Capitan because with the latest HS it would freeze dumping) I just need to edit the serial number out of my ROM before posting. That bios link you posted ended up going to a 403 Forbidden location unfortunately. Also I think we have 2 more years (a little less now) of security updates. It's 2 years after the 1st mainstream year.

You should edit:

• 3rd stream of NVRAM - SSN
• 3rd stream of NVRAM - SON
• LBSN_BD sector - LBSN
• LBSN_BD sector - BD

 

[highvoltage12v]

You should edit:

• 3rd stream of NVRAM - SSN
• 3rd stream of NVRAM - SON
• LBSN_BD sector - LBSN
• LBSN_BD sector - BD

Thanks for letting me know! Should I send it to you before making it public? I know you're busy.

 

[tsialex]

Thanks for letting me know! Should I send it to you before making it public? I know you're busy.

To test this type of thing, a dummy NVRAM volume with correct checksums and LBSN_BD sector should be created.

If you just edit, all checksums will be wrong.

 

[Petri Krohn]

Microcode may not be all that is needed to enable X5400/E5400 series Xeons to run on Mac Pro 1,1. The system may also need the Intel® Firmware Support Package to do memory and processor initialization. These may have to be extracted from the Mac Pro 3,1 firmware. (Would these tools work?)

In my mind we have to options to succeed in running Sierra, High Sierra or even Mojave on an MacPro1,1/2,1 (besides any needed hardware modifications):

1. Supporting XEON 54xx: Back porting the MacPro 3,1 EFI to the MacPro1,1/2,1. It will overcome the 32Bit limitations...
2. Staying with XEON 53xx/51xx: Finalizing pike-R-alpha´s boot loader and create a kext to emulate the SSE4.x instruction set using this library provided by Intel for developers...

There is a third option:

• Porting Coreboot to Mac Pro 1,1.

This would enable using any processor supported by compatible with the socket and running any bootloader or firmware as the payload of Coreboot. This should be possible as:

1. We have the tools to flash any firmware to Mac Pro 1,1 - including firmware that will brick it.
2. Libreboot has been ported to Macbook 1,1/2,1 and iMac 5,2. Coreboot should be easier to get working than Libreboot, as it uses the proprietary closed source Intel FSF blobs.

All that is needed to run 64 bit macOS is for the firmware to post and load the master boot record from a MBR drive. This method has actually been used to run 64 El Capitan on a Mac Pro 1,1, but most people seem to prefer the boot64.efi -> boot.efi method. The bootloader on the MBR drive (Chameleon, Clover, rEFInd) would implement the EFI firmware and emulate the Mac Pro 1,1 hardware.

If this works, then it would be best to move as much of the code as possible to the flash ROM chip on the Mac Pro 1,1 mainboard. For development and testing the preferred hardware solution might be a CompactFlash memory card in the second optical drive bay connected to the IDE cable with a CF to IDE adapter. (I have a whole box full of cheap adapters and cards I used to build m0n0wall routers some 14 years ago.)

 

[razorfsh]

Just something to be aware of.......
I took my mac pro 5.1 boot rom to 140....... but i found that after re-installing the 2019 security update for sierra..
it was back down to 138xxxx

 

[tsialex]

Just something to be aware of.......
I took my mac pro 5.1 boot rom to 140....... but i found that after re-installing the 2019 security update for sierra..
it was back down to 138xxxx

Impossible, Sierra don’t even have a Mac Pro BootROM and Apple efi2flasher don’t flash a previous version.

If your BootROM was downgraded, it was something else.

 

[crjackson2134]

Just something to be aware of.......
I took my mac pro 5.1 boot rom to 140....... but i found that after re-installing the 2019 security update for sierra..
it was back down to 138xxxx

Updating the bootrom on a cMP requires specific user intention and physical interaction on your part. You can't change the bootrom without 1st physically putting your cMP into programming mode. Also, as @tsialex stated, the Apple efi2flasher can't automatically decrease your version.

You may have thought that's what happened, but you are mistaken.

Perhaps you never achieved firmware version 140.0.0.0.0 to begin with, but were under the impression that you had, and when you checked more thoroughly you found that 138.0.0.0.0 was installed instead.

I suspect you attempted a firmware update to 140.0.0.0.0 and it appeared to be successful but actually failed, and left you with the impression you were on 140.0.0.0.0. Now that you have actually checked the version, you have discovered you are on 138.0.0.0.0, and are incorrectly assuming that the security update regressed your firmware version (which is just not possible).

 

[Petri Krohn]

Let's get back to the topic of updating Mac Pro firmware to add support to newer processors, specifically enabling use of 5400 series Harpertown Xeons on Mac Pro 1,1 and 2,1.

• Porting Coreboot to Mac Pro 1,1.

Super Micro made many dual Xeon motherboards that are very similar to the Mac Pro 1,1 logic board. Models X7DAE, X7DAL-E, X7DA8, and X7DGU share the Intel 5000X north bridge with Mac Pro 1,1. I have not checked if the other components on the board are the same, but both Apple and SuperMicro seem to follow Intel's reference designs. All the boards support Harpertown Xeons.

Would it be possible to flash a Supermicro or Intel BIOS on the Mac Pro logic board. (Or more importantly, would the board post with the Supermicro firmware?)

I checked if Coreboot had been ported to any of these Supermicro motherboards. There is one at least partialy supported board, the Supermicro X7DB8. It has the Intel 5000P north bridge which lacks the snoop filter of the 5000X. Support for Supermicro X7DB8 was dropped in June 2018 with the release of Coreboot 4.8. The latest version could however function as the basis for porting Coreboot to Mac Pro 1,1,.

 

[Petri Krohn]

I downloaded the Clovertown and Harpertown datasheets and compared them looking for hardware differences. I noticed that the Harpertown Xeons 5400s have TESTHI12 signal on pin AE3 while on the Clovertown Xeon 5300s the pin is marked as RESERVED. I went to Google searching for "TESTHI12 Harpertown". The only two results were the Harpertown datasheet and page 5 of this discussion.

I have spend some time reading through the data sheets of the XEON 5300 and XEON 5400.
There is a difference in the use of TESTHI* PINs, which may create an issue as unused (reserved) pins are supposed to be left open while the TESTHI*-Pins are supposed to be terminated by dedicated resistors.

                    XEON 5300                                     XEON 5400
Signal Name  Pin #   Type                      Signal Name  Pin #   Type
RESERVED     AE3                               TESTHI12     AE3     Power/Other     Input
TESTHI00     F26     Power/Other     Input     RESERVED     F26    
TESTHI01     W3      Power/Other     Input     RESERVED     W3    
TESTHI02     F25     Power/Other     Input     RESERVED     F25    
TESTHI03     G25     Power/Other     Input     RESERVED     G25    
TESTHI04     G27     Power/Other     Input     RESERVED     G27    
TESTHI05     G26     Power/Other     Input     RESERVED     G26    
TESTHI06     G24     Power/Other     Input     RESERVED     G24    
TESTHI07     F24     Power/Other     Input     RESERVED     F24

The hardware modifications needed to fit a Harpertown Xeon on a Clovertown motherboard are of about the same complexity as the BSEL mods done ten years ago to overclock LGA775 Core 2 processors. Or the Chinese address pin swapping trick that allows LGA771 Xeons to run on almost any LGA775 motherboard. One would isolate with tape of maybe nail polish the pins / lands used for test signals TESTHI00 to TESTHI09. One could use and adapter similar to the LGA771 to LGA775 adapter to swap pins AE3 and W3. The reserved pins would be left open and TESTHI12 would be pulled up to Vtt with a 51 Ω resistor.

I do not think any of this is actually necessary and I doubt it would boot a Harpertown on s Mac Pro 1,1. It is worth a test though.

All Intel and Supermicro 5000X motherboards support Harpertown processors without hardware modifications, at least with an updated bios. Pin-swapped 5400 Xeons seem to work on just about any LGA775 motherboard. (I have a Xeon X5460 running on a Gigabyte G41 motherboard.)

P.S. - Page 34 of the technical product specification of the Intel® Workstation Board S5000XVN lists the addresses used by the SMBUS. It is highly likely that Apple used the same addresses. (This is useful for porting Coreboot to Mac Pro 1,1.)

 

[Petri Krohn]

Update:

...Intel 5000X motherboards support Harpertown processors without hardware modifications...

Actually they do not. Intel made some minor modification to their motherboards. The revised motherboards are identified by the letter "R" at the end of their product code. The same BIOS is used in both versions. Based on the previous discussion it seems that the hardware modification consisted of terminating pin AE3 to Vtt with a 50 ohm resistor.

Would a 5400 series Harpertown Xeon work without TESTHI12 on land AE3 being terminated to Vtt? There is a simple test to find out. Take a system with a working 5400 Xeon. Isolate land AE3. See if the processor still boots. (I have a Xeon X5460 running on an LGA775 motherboard, but I cannot do the test now, as the system is still in use.)

The mod:

If AE3 is the reason for Harpertown Xeons not working on Mac Pro 1,1, then there are three ways of modifying the hardware to make it work.

1. Modify the motherboard. Most LGA775/LGA771 motherboards have the pins exposed on the underside of the board by vias. I looked at photos of the Mac Pro 1,1 logic board and most pins are exposed. I did not find a closeup, so I am not sure if pin AE3 is included. People doing BSEL modifications on Thinkpads have soldered wires to similar vias.
2. Create an adapter similar to the "No Need Adapter" used for seating LGA771 Xeons on LGA775 motherboards. The adapter would connect land AE3 of the processor to pin W3 of the socket. On Clowertown Xeons land W3 is used for TESTHI1, but on Harpertowns it should be left unconnected.
3. Modify the processor by connecting land AE3 to land W3. The connection can be routed over unneeded lands AE2, AE1 and Y1, Y2, Y3. The lands are either Vss = ground, "reserved", or Test Clock. (I do not think the processor Test Bus is needed for operation.) Three layers of paint would be painted over the unneeded lands. 1) isolation, 2) conducting paint, 3) isolation.

The bad news: The above modifications may not solve the issue. I checked all the datasheets for LGA775 processors I could find on Intel's site plus some that were impossible to find. This included processors from Pentium 4s to 45 nm quad core Yorkfields. All of the processors shared the same pinout as 5300 series Xeons, with AE3 "reserved" and F[24] to G[27] used for TESTHI test signals. Xeon 5400s work fine on most LGA775 motherboards with the use of the Chinese address pin swapping adapter. I do not think these LGA775 motherboards have terminated pin AE3. More research is needed...

 

[swimllamaswim]

Update:


Actually they do not. Intel made some minor modification to their motherboards. The revised motherboards are identified by the letter "R" at the end of their product code. The same BIOS is used in both versions. Based on the previous discussion it seems that the hardware modification consisted of terminating pin AE3 to Vtt with a 50 ohm resistor.

Would a 5400 series Harpertown Xeon work without TESTHI12 on land AE3 being terminated to Vtt? There is a simple test to find out. Take a system with a working 5400 Xeon. Isolate land AE3. See if the processor still boots. (I have a Xeon X5460 running on an LGA775 motherboard, but I cannot do the test now, as the system is still in use.)

The mod:

If AE3 is the reason for Harpertown Xeons not working on Mac Pro 1,1, then there are three ways of modifying the hardware to make it work.

1. Modify the motherboard. Most LGA775/LGA771 motherboards have the pins exposed on the underside of the board by vias. I looked at photos of the Mac Pro 1,1 logic board and most pins are exposed. I did not find a closeup, so I am not sure if pin AE3 is included. People doing BSEL modifications on Thinkpads have soldered wires to similar vias.
2. Create an adapter similar to the "No Need Adapter" used for seating LGA771 Xeons on LGA775 motherboards. The adapter would connect land AE3 of the processor to pin W3 of the socket. On Clowertown Xeons land W3 is used for TESTHI1, but on Harpertowns it should be left unconnected.
3. Modify the processor by connecting land AE3 to land W3. The connection can be routed over unneeded lands AE2, AE1 and Y1, Y2, Y3. The lands are either Vss = ground, "reserved", or Test Clock. (I do not think the processor Test Bus is needed for operation.) Three layers of paint would be painted over the unneeded lands. 1) isolation, 2) conducting paint, 3) isolation.

The bad news: The above modifications may not solve the issue. I checked all the datasheets for LGA775 processors I could find on Intel's site plus some that were impossible to find. This included processors from Pentium 4s to 45 nm quad core Yorkfields. All of the processors shared the same pinout as 5300 series Xeons, with AE3 "reserved" and F[24] to G[27] used for TESTHI test signals. Xeon 5400s work fine on most LGA775 motherboards with the use of the Chinese address pin swapping adapter. I do not think these LGA775 motherboards have terminated pin AE3. More research is needed...

Just picked up a 1,1, waiting on a couple RAM daughterboards for it, but I have a set of E5420s in a Supermicro X7DCU thats not being used. Pretty sure I have like 3 775 socket systems put away as well. Going to attempt to isolate AE3 on the supermicro board, and I'll report back soon if it worked. If it does, I'll try it again on my 1,1 once my ram boards get here.

 

[swimllamaswim]

Just picked up a 1,1, waiting on a couple RAM daughterboards for it, but I have a set of E5420s in a Supermicro X7DCU thats not being used. Pretty sure I have like 3 775 socket systems put away as well. Going to attempt to isolate AE3 on the supermicro board, and I'll report back soon if it worked. If it does, I'll try it again on my 1,1 once my ram boards get here.

Cool update, the CPUs are booting on my Supermicro board. Rather than trying to isolate AE3 on the board itself, I just covered W3 with conformal coating, ran a wire from AE3 over where W3 goes, and covered AE3 with conformal coating. So far, everything seems to work fine, it detects all of the RAM, both CPUs, correct clocks, everything. I'm getting an Ubuntu image ready to boot it up and run some tests, but so far promising.

Can't test it on my Mac Pro yet though, my RAM cards came in, and I realized I don't have any FB-DIMMs. I could have sworn I did, but oh well. Gotta wait on those now.

 

[Yojx21]

Cool update, the CPUs are booting on my Supermicro board. Rather than trying to isolate AE3 on the board itself, I just covered W3 with conformal coating, ran a wire from AE3 over where W3 goes, and covered AE3 with conformal coating. So far, everything seems to work fine, it detects all of the RAM, both CPUs, correct clocks, everything. I'm getting an Ubuntu image ready to boot it up and run some tests, but so far promising.

Can't test it on my Mac Pro yet though, my RAM cards came in, and I realized I don't have any FB-DIMMs. I could have sworn I did, but oh well. Gotta wait on those now.

Very interesting !! Tell us the result when you get your Ram !