MP All Models Mac Pro CPU Compatibility List

[TomMuc]

MP21.007F.B08 firmware microcodes:

View attachment 947725

Is it correct to assume that last 6FB entry isn't there in the B06 firmware?

BTW, do you usually run these machines with SIP enabled or disabled? I only discovered it was enabled in El Capitan when ROMTool told me to disable it. I had mistakenly assumed it was disabled by the patcher.

how do you guys use those tools to get all the info out of the rom? i saw where to download and already dumped some of our roms. but how and what tools to use and what infos are interesting there i have no idea. is there a guide/howto?

btw we enable sip on all machines we mostly run mac os and on those which run windows most of the time we disable because this enables us to do remote backups of the windows partition via winclone. otherwise we had to first disable sip everytime.

 

[LightBulbFun]

View attachment 947662

View attachment 947664

View attachment 947663

Everymac states that the introduction date of this model was 2007-04-04. I'm not sure how to read the CPU-Z date but I'm assuming it's the Eurasian standard, so mainboard manufacture date is 2007-06-09. That means this particular unit likely wouldn't have been available at retail until summer or fall 2007.

ROM dump to follow.

Personally I always write my numerical format dates as YYYY-MM-DD so everyone in the world can understand.

MP21.007F.B08 firmware microcodes:

View attachment 947725

Is it correct to assume that last 6FB entry isn't there in the B06 firmware?

BTW, do you usually run these machines with SIP enabled or disabled? I only realized my install of El Capitan had SIP enabled when ROMTool told me to disable it.

very interesting stuff the date under BIOS is the date the BIOS was compiled/made, not when the computer was made, that will be the first 3 numbers of your serial number

indeed the 6FB entry is not there in the B06 firmware

it is very interesting and cool to see that there was a later MP2,1 firmware that added official support for G0 CPUs

I wonder how hard it would be to sanitise your BootROM of all the personal info and use it to update the MP1,1-2,1 tool and for others to update their Mac Pro 1,1/2,1's with

one for @tsialex perhaps?

 

[EugW]

very interesting stuff the date under BIOS is the date the BIOS was compiled/made, not when the computer was made, that will be the first 3 numbers of your serial number

indeed the 6FB entry is not there in the B06 firmware

it is very interesting and cool to see that there was a later MP2,1 firmware that added official support for G0 CPUs

I wonder how hard it would be to sanitise your BootROM of all the personal info and use it to update the MP1,1-2,1 tool and for others to update their Mac Pro 1,1/2,1's with

one for @tsialex perhaps?

1st 5 digits 40752, which I understand to mean the last week of 2007. 40 is the location, no?

How to sanitize?

 

[LightBulbFun]

1st 5 digits 40752, which I understand to mean the last week of 2007. 40 is the location, no?

How to sanitize?

yeah 40 is the factory code, its not always numbers however, it can be a set of letters or a letter/number or number/letter

indeed from 52nd week 2007

sanitise would be like removing your serial numbers from it and the such like, so its a blank BootROM, that can be used to update other machines

 

[EugW]

With that manufacture date, that means the Mac Pro 3,1 only came out a week or two later. That must have hurt. Or maybe they got a discount, after 3,1 was released.

Which hex editor is recommended for this, how much sanitization should I expect to do, and any tricks to maintain integrity of the file?

I haven’t hex edited anything in decades.

 

[tsialex]

With that manufacture date, that means the Mac Pro 3,1 only came out a week or two later. That must have hurt. Or maybe they got a discount, after 3,1 was released.

Which hex editor is recommended for this, how much sanitization should I expect to do, and any tricks to maintain integrity of the file?

I haven’t hex edited anything in decades.

There are several checksums that need to be valid after removing the hardwareIDs from your own Mac Pro. There is one for the NVRAM volume, a free space indicator and another one for each of the Fsys and Gaid stores. It's not just to open your dump and overwrite the hardwareIDs with an hex editor. since a wrong checksum frequently makes a brick.

I'll send you a PM later tonight or tomorrow.

 

[EugW]

very interesting stuff the date under BIOS is the date the BIOS was compiled/made, not when the computer was made, that will be the first 3 numbers of your serial number

indeed the 6FB entry is not there in the B06 firmware

it is very interesting and cool to see that there was a later MP2,1 firmware that added official support for G0 CPUs

I see that the Mac Pro EFI updater that updates MacPro2,1 to B06 is from 2007-09-27.

Mac Pro EFI Firmware Update 1.2 - Apple Support (CA)

This EFI Firmware Update fixes several fan behavior and Boot Camp issues on Mac Pro computers.

support.apple.com

However, the B08 firmware in my EMC 2138 machine was built 2007-06-09, which is 3.5 months before that B06 firmware updater was released.

I wonder when Apple started shipping G0 in their 8-core models. Maybe that didn't happen until fall. There is a mention of a 2007-08-13 release date for this chip at CPU-World, but the entry is a bit confusing.

Intel Xeon X5365 - HH80563KJ0808MP (BX80563X5365A / BX80563X5365P)

Intel Xeon X5365 server CPU: latest news, detailed specifications, side by side comparison, FAQ, pictures and more from CPU-World

www.cpu-world.com

Regardless though, assuming that the hardware hadn't changed (aside from the CPU stepping), there would be no need for the B08 firmware to be released. All the G0 X5365 machines would already have the B08 firmware, and the B06 firmware would be fine for everything else.

 

[LightBulbFun]

I see that the Mac Pro EFI updater that updates MacPro2,1 to B06 is from 2007-09-27.

Mac Pro EFI Firmware Update 1.2 - Apple Support (CA)

This EFI Firmware Update fixes several fan behavior and Boot Camp issues on Mac Pro computers.

support.apple.com

However, the B08 firmware in my EMC 2138 machine was built 2007-06-09, which is 3.5 months before that B06 firmware updater was released.

I wonder when Apple started shipping G0 in their 8-core models. Maybe that didn't happen until fall. There is a mention of a 2007-08-13 release date for this chip at CPU-World, but the entry is a bit confusing.

Intel Xeon X5365 - HH80563KJ0808MP (BX80563X5365A / BX80563X5365P)

Intel Xeon X5365 server CPU: latest news, detailed specifications, side by side comparison, FAQ, pictures and more from CPU-World

www.cpu-world.com

Regardless though, assuming that the hardware hadn't changed (aside from the CPU stepping), there would be no need for the B08 firmware to be released. All the G0 X5365 machines would already have the B08 firmware, and the B06 firmware would be fine for everything else.

no the B06 firmware is from the 2nd of June 2007 http://valid.x86.fr/3gn7hb

your special B08 firmware is from the 6th of September 2007

(the BIOS dates is in the confusing US format)

 

[GreppMichaels]

Hey LightBulbFun, been following your posts on several threads over the last several weeks and was wondering if you think any ingenious hackers such as yourself will be able to, or if it's even possible, for a microcode update to cMP 5,1 along with a bios mod to adjust the base frequency multiplier so that we can get some better CPUgains with the X5698 or any of the higher clock chips compared to the x5690 to help future proof us a bit more. Personally I have a feeling the next gen of video Cards are really going to start bottlenecking us if we do a little gaming on the side, I primarily do professional video/music/editing on my 5,1 but do have a win10 drive that I occasionally game on and the cMP 5,1 with a vega 56 flashed to 64 performs pretty well, but I'm wondering how the future looks.

 

[tsialex]

Hey LightBulbFun, been following your posts on several threads over the last several weeks and was wondering if you think any ingenious hackers such as yourself will be able to, or if it's even possible, for a microcode update to cMP 5,1 so that we can get some better CPUgains with the X5698 or any of the higher clock chips compared to the x5690 to help future proof us a bit more. Personally I have a feeling the next gen of video Cards are really going to start bottlenecking us if we do a little gaming on the side, I primarily do professional video/music/editing on my 5,1 but do have a win10 drive that I occasionally game on and the cMP 5,1 with a vega 56 flashed to 64 performs pretty well, but I'm wondering how the future looks.

Not possible, X5698 incompability is not related to microcode but to the unsupported base frequency multiplier needed for the two models with higher frequency than X5690.

X5690 is the fastest possible Xeon with MP5,1 forget about X5698/5687.

 

[GreppMichaels]

Not possible, X5698 incompability is not related to microcode but to the unsupported base frequency multiplier needed for the two models with higher frequency than X5690.

X5690 is the fastest possible Xeon with MP5,1 forget about X5698/5687.

Thank you, I'm aware the X5690 is currently the fastest, however there have been instances in the past related to bios mods that address base frequency multipliers where CPU's either don't work or problems with clockspeed because of the frequency, I believe even earlier Mac Pro's addressed overclocking via the base frequency multiplier. There was a time when 4,1 Mac Pro's couldn't run Westmere,and while I understand that was a simpler solution that revolved around hacking what Apple had already done with the 5,1, I don't see the problem in hoping someone will figure something not. If we know what the problem is and people have dumped bios's and bootrom's, there's no reason we can't find a solution.

It sounds like it would entail modifying the bios to support the base frequency multiplier and then add the micrcode to the rom, no?

 

[tsialex]

Thank you, I'm aware the X5690 is currently the fastest, however there have been instances in the past related to bios mods that address base frequency multipliers where CPU's either don't work or problems with clockspeed because of the frequency, I believe even earlier Mac Pro's addressed overclocking via the base frequency multiplier. There was a time when 4,1 Mac Pro's couldn't run Westmere,and while I understand that was a simpler solution that revolved around hacking what Apple had already done with the 5,1, I don't see the problem in hoping someone will figure something not. If we know what the problem is and people have dumped bios's and bootrom's, there's no reason we can't find a solution.

It sounds like it would entail modifying the bios to support the base frequency multiplier and then add the micrcode to the rom, no?

Again, this has nothing to do with microcode, the maximum supported multiplier is 26 and this is set on stone.

Btw, why do this at all? X5698 is dual-core, X5687 is quad-core.

 

[GreppMichaels]

Again, this has nothing to do with microcode, the maximum supported multiplier is 26 and this is set on stone.

Btw, why do this at all? X5698 is dual-core, X5687 is quad-core.

My apologies if I wasn't clearer in my original post, I was addressing the CPU bottleneck in terms of single core performance when gaming or anything that utilizes single core. Not sure if you're a gamer but I'm sure you're aware that single core clockspeed is the most important thing in a pc or mac to keep up with high powered GPU's so the CPU can handle all the informatin it's throwing out. So going from 3.6ghz at turbo to 4.4 ghz would be substantial, and in a dual CPU setup, i'd have to think losing 4 cores but gaining roughly 1ghz in clock speed per CPU would be a relative wash. I guess I'm not understanding why you're shooting this down so much especially when LightBulbFan was the one who was originally postulating if this was even possible based on similar issues that were overcame on other platforms.

I guess I just don't understand the point of your comments as you're not telling me anything we don't already know, yes X5690 is fastest CPU at the moment, and telling me the multiplier is "set in stone" doesn't tell me anything especially when I know for an absolute fact unsupported multipliers in bioses have been adjusted in the past to support higher clocked CPU's, have you tried to modify a cMP bios to address the multiplier? I don't mind being wrong or being shot down, but you're not giving me helpful information to understand why it isn't possible, when so many similar modifications have been done on both PC's and Mac's.

 

[tsialex]

My apologies if I wasn't clearer in my original post, I was addressing the CPU bottleneck in terms of single core performance when gaming or anything that utilizes single core. Not sure if you're a gamer but I'm sure you're aware that single core clockspeed is the most important thing in a pc or mac to keep up with high powered GPU's so the CPU can handle all the informatin it's throwing out. So going from 3.6ghz at turbo to 4.4 ghz would be substantial, and in a dual CPU setup, i'd have to think losing 4 cores but gaining roughly 1ghz in clock speed per CPU would be a relative wash. I guess I'm not understanding why you're shooting this down so much especially when LightBulbFan was the one who was originally postulating if this was even possible based on similar issues that were overcame on other platforms.

I guess I just don't understand the point of your comments as you're not telling me anything we don't already know, yes X5690 is fastest CPU at the moment, and telling me the multiplier is "set in stone" doesn't tell me anything especially when I know for an absolute fact unsupported multipliers in bioses have been adjusted in the past to support higher clocked CPU's, have you tried to modify a cMP bios to address the multiplier? I don't mind being wrong or being shot down, but you're not giving me helpful information to understand why it isn't possible, when so many similar modifications have been done on both PC's and Mac's.

It's not possible since Apple designed the multiplier circuit and firmware code to support up to 26x1333, the maximum possible multiplier with a MP5,1.

You can disassemble the EFI module yourself and check it, all CPU related PEI and EFI modules are inside 7A9354D9-0468-444A-81CE-0BF617D890DF, if I remember correctly, it's the CPUInitPei and CPUInitDxe.

 

[GreppMichaels]

It's not possible since Apple designed the multiplier circuit and firmware code to support up to 26x1333, the maximum possible multiplier with a MP5,1.

You can disassemble the EFI module yourself and check it, all CPU related PEI and EFI modules are inside 7A9354D9-0468-444A-81CE-0BF617D890DF, if I remember correctly, it's the CPUInitPei and CPUInitDxe.

Ok I appreciate the more detailed explanation, so even if we were able to rewrite the firmware code, the real issue here is the actual hardware itself, IE the multiplier circuit cannot support it regardless of what the bios says? If that's the case thank you for the explanation, I really appreciate it. I was coming from the position of with the right knowhow nearly anything is possible, I had a 2010 imac that I flashed an nvidia quadro mxm card with a modified mac bios which both allowed me a boot screen and to run significantly better video rendering and gaming software on my mac, albeit with extreme hardware and software modifications, so I'm always hopeful, but I understand there are concrete limitations of these machines.

 

[tsialex]

Ok I appreciate the more detailed explanation, so even if we were able to rewrite the firmware code, the real issue here is the actual hardware itself, IE the multiplier circuit cannot support it regardless of what the bios says? If that's the case thank you for the explanation, I really appreciate it. I was coming from the position of with the right knowhow nearly anything is possible, I had a 2010 imac that I flashed an nvidia quadro mxm card with a modified mac bios which both allowed me a boot screen and to run significantly better video rendering and gaming software on my mac, albeit with extreme hardware and software modifications, so I'm always hopeful, but I understand there are concrete limitations of these machines.

Macs don't have BIOS or anything similar, the CPU initialisation at PEI it's not even related to the way PC motherboards do the configuration with the UEFI SetUp/Configuration. Anyway, you can always rewrite the CPUInit PEI/DXE and modify the backplane for a more versatile clock CI, the cost of developing this will be at insane levels, but can be done.

 

[AlexMaximus]

My apologies if I wasn't clearer in my original post, I was addressing the CPU bottleneck in terms of single core performance when gaming or anything that utilizes single core. Not sure if you're a gamer but I'm sure you're aware that single core clockspeed is the most important thing in a pc or mac to keep up with high powered GPU's so the CPU can handle all the informatin it's throwing out. So going from 3.6ghz at turbo to 4.4 ghz would be substantial, and in a dual CPU setup, i'd have to think losing 4 cores but gaining roughly 1ghz in clock speed per CPU would be a relative wash. I guess I'm not understanding why you're shooting this down so much especially when LightBulbFan was the one who was originally postulating if this was even possible based on similar issues that were overcame on other platforms.

I guess I just don't understand the point of your comments as you're not telling me anything we don't already know, yes X5690 is fastest CPU at the moment, and telling me the multiplier is "set in stone" doesn't tell me anything especially when I know for an absolute fact unsupported multipliers in bioses have been adjusted in the past to support higher clocked CPU's, have you tried to modify a cMP bios to address the multiplier? I don't mind being wrong or being shot down, but you're not giving me helpful information to understand why it isn't possible, when so many similar modifications have been done on both PC's and Mac's.

Hi Michael,

You are not at the end yet. Replace your Vega 64 for a Vega VII with 16GB VRam. That's what I did and it works really great for years now. I used to have the Vega64, there is a noticeable difference.

 

[GreppMichaels]

Hi Michael,

You are not at the end yet. Replace your Vega 64 for a Vega VII with 16GB VRam. That's what I did and it works really great for years now. I used to have the Vega64, there is a noticeable difference.

Thanks appreciate the comment, the VII's are way too pricey right now (no idea why) however I just picked up a Radeon Frontier with an aftermarket waterblock already installed for 275 off facebook marketplace (been hunting one forever and this was the best deal by far) my next step is figure out where to squeeze a radiator in but it looks like it's been done before in a 5,1. Eventually i'll go to the vii or big navi, but this should be a good stopgap.

 

[koreda]

Mac Pro 5,1 and 4,1
Xserve 3,1

• 4,1 dual-processorMac Pro uses a special, "lidless" CPUs (the single-processor models use normal CPUs). This has a different height than the normal CPUs, so if you upgrade a dual-processor 4,1 Mac Pro with different processors, you need to deal with the height difference. There are several ways to do this:
  • Buy CPUs that someone has already de-lidded.
  • Buy normal CPUs and de-lid them yourself. Here is thread about delidding yourself and here is a video. (And here is a post warning against delidding yourself.)
  • Buy normal CPUs and pay for de-lidding service.
  • Keep the lids on, but be extremely careful about not tightening down the heat sink too much. (If you don't remove the lids it is very easy to tighten down the heat sink screws too much on a CPU that's too large, resulting in damage to the socket, CPU, and/or board.)
  • Other people have kept the lids on, but added washer stacks (of equivalent height to the lid difference) to where the tightening screws go--this is intended to prevent the heat sink from overtightening and causing damage. Here is an informational washer stack thread, and here is a really good post with pictures and a step-by-step procedure.
  • If you keep the lids on, the existing heat pad won't reach the heat sink due to the extra height of the lids. You need to make up for this gap by replacing the heat pad with a thicker pad.

Just to confirm, All of this [above quoted text] applies to MacPro4,1, and not 5,1?

I upgraded my cmp5,1 (mid-2012) to 3.46 over a year ago and more recently, confirmed, CPU heat infrequently approaches approx 136-139°F

I try to re-grease my CPUs every 18 - 36 months, regardless. But after reviewing this thread, specifically regarding (CPU upgrade), I’m wondering if more immediate care is necessary, i.e. delidding or absorption pad?

Mac Pro (mid-2012) 128GB 1333Mhz RX 580 16TB RAID 5/NWR TECH mini-SAS controller​

 

[GreppMichaels]

Just to confirm, All of this [quoted text specifically] applies to MacPro4,1 (and not 5,1)?

I upgraded my cmp5,1 (mid-2012) to 3.46 over a year ago and more recently, confirmed, CPU heat infrequently approaches approx 136-139°F

I try to re-grease my CPUs every 18 - 36 months, regardless. But after reviewing this thread, specifically regarding (CPU upgrade), I’m wondering if more immediate care is necessary, i.e. delidding or absorption pad?

Mac Pro (mid-2012) 128GB 1333Mhz RX 580 16TB RAID 5/NWR TECH mini-SAS controller​

Dual delidded 4,1 x5690's here. Are you running a fan manager? Apple is notorious for not ramping up fans until your CPU is near max operating temp. I was a bit worried about my idle temps, my apartment is generally pretty warm and I wanted my CPU's to run cooler and both cpu's now run around 40c at idle (CPU B always a bit cooler.) I'm not sure what they used to run at but running mac fan control and just upping the RPM's for intake, exhaust,and the CPU fans to maintain closer to 1k-1.5k at idle, also helps keep the ram cooler which can help in throttling as well. Those fan RPM's are pretty quiet too, I have the temp settings setup to do max cooling at minimum RPM, I'd say experiment with fan management first, unless you already have.

 

[mav2287]

@ActionableMango for the Xserve 3,1 it lists max ram as 96Gb. Isn't the limit actually 128GB using 16GB sticks?

 

[LightBulbFun]

@ActionableMango for the Xserve 3,1 it lists max ram as 96Gb. Isn't the limit actually 128GB using 16GB sticks?

192GB actually, using 12 16GB RDIMMs

 

[ActionableMango]

@ActionableMango for the Xserve 3,1 it lists max ram as 96Gb. Isn't the limit actually 128GB using 16GB sticks?

I wouldn't know. Feel free to edit the list with corrections.

 

[sablackman]

Has anyone followed this set of procedures from iFixit? i’m particularly curious about the steps around 20 that start guidance on cleaning the northbridge heat sink. The whole part about pulling it off and using a new set of screws and washers that also includes drilling a hole in the tray. Is this really necessary? It seems like a risky maneuver. i’m not suggesting that it shouldn’t be cleaned. It just seems like an unusual set of steps to put it back in through drilling a hole in the pan that holds a daughterboard.

 

[saulinpa]

Has anyone followed this set of procedures from iFixit? i’m particularly curious about the steps around 20 that start guidance on cleaning the northbridge heat sink. The whole part about pulling it off and using a new set of screws and washers that also includes drilling a hole in the tray. Is this really necessary? It seems like a risky maneuver. i’m not suggesting that it shouldn’t be cleaned. It just seems like an unusual set of steps to put it back in through drilling a hole in the pan that holds a daughterboard.

They wanted to do a thorough cleanings but got in over their head. Can only assume they broke the original plastic stays (retainers) during the disassembly and rather than track down replacements they jury rigged a solution with metal nuts and bolts. Yes, it will be stronger but plastic stays work just fine. The same issue was seen with replacing the CPU in the old Mac mini. Nut and bolt kits were sold to to those that broke the plastic ones.