MP 7,1 CPU upgrade in 2019 Mac Pro

[tsialex]

Not really . The Gold 6212U I tested had a compatible CPUID ( as we found out later , but I deduced this at first ) and Stepping version ( which I knew from day one , which is critically important ) in the firmware ; two things very much in its favor .

Where the non-authorized processor upgrade failed was precisely in the total number of PCIe lanes provided to the System and maybe how they are organized ( since not all PCIe lanes are generated by the CPU ) .

The big question now is , can and how this limitation be patched away ? And what would be the limitations , if possible ? The unauthorized processor might survive all the tests , but we could wind up with a situation where not all the lanes of a PCIe slot are available , maybe we'll lose a slot even , or gosh forbid if the missing lanes are tied into the factory NAND drive ? ( another failed boot situation , since that's where the key is that starts up the System , metaphorically speaking ) .

Apple could enable a Cascade Lake mode with just 8x for the MPX slots, for example. We can’t patch it, since T2 validation of the firmware will reject it.

Anyway, T2 is connected to the C620 PCH then via DMI3.0 link to the CPU (just 4 lanes to the CPU, T2 alone probably can saturate it).

 

[Softco]

Why climb only half way up Mt. Everest ? ? Someday , I'm gonna test 2TB in one of these rigs ...

Annoying end-user answer to your rhetorical question:
"Because it's $3K cheaper half-way up Everest, and your machine makes you money just as fast there." ?

 

[Snow Tiger]

We can’t patch it, since T2 validation of the firmware will reject it.

Alex , did you ever think of spending some of your talents on PC workstations ? My C.L. Xeon Supermicro System makes me feel like its 1977 all over again with that good old fashioned home brew fun . We could do some really cool things and I'm certain it would be intellectually rewarding . Apple is forcing us to go full dinosaur and I'm not ready for extinction . They won't be happy until they've turned every last aspect of the Macintosh line into a horribly expensive and disposable appliance . I've been doing Apple System CPU Upgrades for the last 35 years ...

A client from California performs some pretty sophisticated environmental simulations with Macs and one of his fully decked out cMPs just died so hard he literally nearly had a panic attack . So he sent it to me because nobody else repairs these systems anymore . His local Mac tech retired and Apple won't touch it . Guess what it will take to get that puppy back online again and purring along at continuous load ? A fifty cent spring , two dollars of quality thermal paste , a good cleaning and a twenty five dollar chassis filter for preventative maintenance ( his local environment is very dusty ) . Apple would recommend he fill his lab with brand spanking new $6000 MP7,1s or iMac Pros . This gear is way out of his budget ... Apple is all about money and control . I know I sound like a broken record , but Apple keeps singing the same ... exact ... song .

 

[Snow Tiger]

Annoying end-user answer to your rhetorical question:
"Because it's $3K cheaper half-way up Everest, and your machine makes you money just as fast there." ?

Wait until you find out I saved 2,800 bucks doing a self install of the processor I picked up used from China . ? . So , I'll just have to suffer and plant my pirate flag on the tippy top of the mountain , there . I just hate paying full price for anything ! I just hope they send me a production version and not some sneaky rebadged ES or QS silicon . It's not like I won't be able to figure it out or anything . In fact , the very first person who performed the W-3275M upgrade did exactly that from Taobao , I think . It's not without risk , though . Helps to be a native speaker on these markets ...

 

[Rehtori]

So earlier today Xeon 3265 arrived and off we went. It jumped in and stock cpu jumped out. Installation was pretty straight forward, but ofc needed a great care and attention not to drop or tilt the cpu while unsocketing / re-socketing. No need to tear down anything except the heatsink ofc. The system booted up as expected. But..... there's was a quirk. After few minutes of Catalina running idle the mac rebooted unexpectedly, then booted up into catalina. After this, all seemed to be normal. I then ran Extre check. All checked out ok. After this I turned of the mac and ran Apple's Diagnostic test. All checked out fine there as well. No issues. The system is now booting up normally. No SMC reset / pram reset done yet.. Temps normal, turboboost seem to be working as expected. Will report here if something unexpected happens again in coming days.

 

[Snow Tiger]

So earlier today Xeon 3265 arrived and off we went. It jumped in and stock cpu jumped out. Installation was pretty straight forward, but ofc needed a great care and attention not to drop or tilt the cpu while unsocketing / re-socketing. No need to tear down anything except the heatsink ofc. The system booted up as expected. But..... there's was a quirk. After few minutes of Catalina running idle the mac rebooted unexpectedly, then booted up into catalina. After this, all seemed to be normal. I then ran Extre check. All checked out ok. After this I turned of the mac and ran Apple's Diagnostic test. All checked out fine there as well. No issues. The system is now booting up normally. No SMC reset / pram reset done yet.. Temps normal, turboboost seem to be working as expected. Will report here if something unexpected happens again in coming days.

Do the NVRAM reset dance please You should not have a problem , but ...
[automerge]1579300077[/automerge]


If your Mac is unexpectedly restarting , maybe you didn't tighten the heatsink enough .

 

[Rehtori]

Under full load with cinebench R20 and temps are hovering around 70 celsius, and when idling, between 39 - 45 celsius. Fans seems to be idle or at least inaudbile under cpu load, so I do think heatsink should be ok.. I made nine point thermal pasting with Noctua Nt1. inserted the heatsink gently down, tightened those screws just to their nice limits where they started to say It hurts, we don't wanna to go any further and brake anything.

 

[DoofenshmirtzEI]

Under full load with cinebench R20 and temps are hovering around 70 celsius, and when idling, between 39 - 45 celsius. Fans seems to be idle or at least inaudbile under cpu load, so I do think heatsink should be ok.. I made nine point thermal pasting with Noctua Nt1. inserted the heatsink gently down, tightened those screws just to their nice limits where they started to say It hurts, we don't wanna to go any further and brake anything.

What were the temps before the surgery?

 

[Rehtori]

What were the temps before the surgery?

I think it was pretty similar figures, maybe a bit less with the stock cpu...

I wonder what are good temps under Cinebench load with this.
I could try to fasten those screws just a tad more.. but I wonder where is the safe limit... don't want to brake the chip or mobo.

 

[Rehtori]

I think it was pretty similar figures, maybe a bit less with the stock cpu...

I wonder what are good temps under Cinebench load with this.
I could try to fasten those screws just a tad more.. but I wonder where is the safe limit... don't want to brake the chip or mobo.

*Update* Fastened screws few mm. No change in temps. Temps actually seem quite high when running cinebench in loop... around 70-75 and peaks hitting even 80 celsius pkg according to intel power gadget, Fans idling. I havent done pram / smc reset as I want first confirm hardware working 100% before taking the last step with T2 chip..

 

[LightBulbFun]

if its just the one off Reboot, then id write it off as just a fluke or such then worry about the heatsink (don't want to tighten it too much or you can risk crushing the socket)

your temps look good to me for a 24 core CPU

 

[Snow Tiger]

*Update* Fastened screws few mm. No change in temps. Temps actually seem quite high when running cinebench in loop... around 70-75 and peaks hitting even 80 celsius pkg according to intel power gadget, Fans idling. I havent done pram / smc reset as I want first confirm hardware working 100% before taking the last step with T2 chip..

Tcase ( a thermal sensor just under the IHS ) of a W-3265 ( 24 Core at 2.7 GHz base ) is 78 degree C . For durability reasons , I'd get that temp down a little lower if you are pushing your Mac at continuous load .

 

[LightBulbFun]

Tcase ( a thermal sensor just under the IHS ) of a W-3265 ( 24 Core at 2.7 GHz base ) is 78 degree C . For durability reasons , I'd get that temp down a little lower if you are pushing your Mac at continuous load .

its important to note that tCase and TJunction temps are 2 different things, intel power gadget etc all read off tJunction which is usually a bit higher then actual tCase temp

I dont think theres any way to read the tCase temp of a CPU not without embedding an external thermocouple into the IHS of the CPU

 

[Snow Tiger]

its important to note that tCase and TJunction temps are 2 different things, intel power gadget etc all read off tJunction which is usually a bit higher then actual tCase temp

I know the difference . Tjunctions are deeper in the CPU package ( they should be next to the cores ) and are historically notoriously difficult to accurately measure , since direct sensors are not used . I think some sort of mathematical formula is used to determine an estimated temp .

If Intel's Power Gadget is utilizing those sensors , then it should also tell us what the current and maximum permitted package temp is for the CPU in question . Could Rehtori post some pictures , please ?

It would not be hard at all to embed a thermocouple onto or under the IHS of a Cascade Lake Xeon . When I started working with T2 Macs , I thought we'd lose access to all System sensor info . So , I planned for exactly this contingency .

TC Direct of the UK has some in house branded thermocouples that should do the job nicely :

0.13 mm thick Cement On Polyimide Thin Film Thermocouple - Aerostructure Grade

Cement On Polyimide Thin Film Thermocouple - Aerostructure Grade - TC Direct

Cement On Polyimide Thin Film Thermocouple - Aerostructure Grade These versatile cement on foil thermocouples give exceptional performance for exceptional applications. Although used widely in the aerospace industry for dynamic testing of composite structures, they can be used on any...

www.tcdirect.co.uk

 

[Rehtori]

I know the difference . Tjunctions are deeper in the CPU package ( they should be next to the cores ) and are historically notoriously difficult to accurately measure , since direct sensors are not used . I think some sort of mathematical formula is used to determine an estimated temp .

If Intel's Power Gadget is utilizing those sensors , then it should also tell us what the current and maximum permitted package temp is for the CPU in question . Could Rehtori post some pictures , please ?

It would not be hard at all to embed a thermocouple onto or under the IHS of a Cascade Lake Xeon . When I started working with T2 Macs , I thought we'd lose access to all System sensor info . So , I planned for exactly this contingency .

TC Direct of the UK has some in house branded thermocouples that should do the job nicely :

0.13 mm thick Cement On Polyimide Thin Film Thermocouple - Aerostructure Grade

Cement On Polyimide Thin Film Thermocouple - Aerostructure Grade - TC Direct

Cement On Polyimide Thin Film Thermocouple - Aerostructure Grade These versatile cement on foil thermocouples give exceptional performance for exceptional applications. Although used widely in the aerospace industry for dynamic testing of composite structures, they can be used on any...

www.tcdirect.co.uk

Here's a picture from cinebench testing session.

 

[Snow Tiger]

Here's a picture from cinebench testing session.

Thanks . The Tjunction ( Core Max ) is permitted to have only a slightly higher maximum temp ( 80 degree C ) than the Tcase maximum ( 78 degree C ) . Well , now you know what temps to avoid . Keep that chip cool and it'll stay online longer before you need to re-thermal paste the HS . ( You should not place paste directly on this large a processor ) .

I wonder why the power consumption for the PKG is so high ?

It displays 230 W and yet the TDP of a W-3265 is 205 W ?

 

[Rehtori]

That's interesting indeed. Just got power peaking at 240W with Cinebench according both Intel Power Gadget 3.6 and 3.7...

 

[LightBulbFun]

It displays 230 W and yet the TDP of a W-3265 is 205 W ?

That's interesting indeed. Just got power peaking at 240W with Cinebench according both Intel Power Gadget 3.6 and 3.7...

thats because the TDP figure given by intel is for the base clock, when a CPU turbo boosts it can exceed that figure (especially in more modern intel CPUs)

remember that Turbo boost is technically just a nice "up to" extra given by intel its not guaranteed in any much way

so its not counted in official TDP figures etc

 

[Snow Tiger]

thats because the TDP figure given by intel is for the base clock, when a CPU turbo boosts it can exceed that figure (especially in more modern intel CPUs)

remember that Turbo boost is technically just a nice "up to" extra given by intel its not guaranteed in any much way

so its not counted in official TDP figures etc

Interesting ... with the X5690 upgraded SP MP5,1's , the processor at System load ( when all major components are concurrently at load ) consumes only 95 W while it's TDP is rated by Intel for 130 W . Sounds undervolted ...

In case anyone needs help understanding the above screen , it's sensor information of a 24 hour burn in with all major System components ( single processor , memory , graphics card , NVMe M.2 SSD ) at load at the same time .

The configuration is :

SP Mac Pro 5,1 ( factory 2009 ) .
Boot ROM version 144 .
1 x X5690 Intel Xeon CPU ( 6 Core @ 3.46 GHz ) .
48 GB ( 3 x 16GB ) 1333 MHz DDR3 ECC memory .
MSi RX 560 4GB non-EFI graphics card , rebuilt .
2 TB Flash SSD , Intel 660p NVMe PCIe M.2 w/ adapter .
macOS Mojave 10.14.6 installation , PCIe Slot 2 .
3 x OWC High Capacity HDD Bay Carriers .
14x Blu-ray burner LG WH14NS40 .
Custom Wireless Subsystem 802.11ac / BT 4.0 Apple Brand Card , 2 - 6 dBi PCIe RP SMA external Antennas , Continuity Support .
USB 3.1 Gen 2 ( 10 Gb/s ) PCIe Card , Dual Ports .

Geekbench 3 was used to push the CPU at load , memtest for Mac pushed the main system memory at load , Luxmark 3.1 pushed the GPU at load and BlackMagic Disk Speed Test pushed the NVMe M.2 SSD at load . Hardware Monitor by Bresink gave most of the sensor info . SMART Utility ( a third party app ) gave individual SMART test results on the SSD . A terminal command gave GPU sensors . System profile provided live ECC memory results . Menu Meters displays some real time activity on the menu bar . Eagle eyed users can see I am running MFC with accelerated constant system fan rotationals just to be on the durability safe side . I like my rigs to stay online a very long time ...

 

[Rehtori]

I decided to have a re-check and removed the fan unit from the case to have better look... Indeed there was some room for tightening those pesky heatsink screws. All nice and tight now, but no change in temps..in fact after 6 consecutive Cinebench runs, power gadget is hovering around 80 and peaks at 90 celsius.. No thermal throttling, Fans idle at 500rpm (except GPU fan which revs up occasionally when playing games). Also SMC reset done.

 

[Snow Tiger]

I decided to have a re-check and removed the fan unit from the case to have better look... Indeed there was some room for fastening those pesky heatsink screws. All nice and tight now, but no change in temps..in fact after 6 consecutive Cinebench runs, power gadget is hovering around 80 and peaks at 90 celsius.. No thermal throttling, Fans idle at 500rpm (except GPU fan which revs up occasionally when playing games). Also SMC reset done.

It's easy to verify the Blower Side's clearance between the CPU heatsink and the standoff . The Fan Side's requires removal of the Axial Fan Assembly to see this . Just another thing to remove ...

Apple service manuals sometimes say a tech should loosen the fastener by a quarter or half turn ( once installed completely ) , as you want to avoid over-tightening the heastsinks if the System gets shipped or moved roughly . I can't think of many techs that use torque drivers , so it's probably good advice . I use ordinary manual drivers . No power tools for this operation and no torque drivers .

Did you perform the NVRAM reset ? It's the old PRAM keys combo pressed at Start Up . ( You can release the keys after the Apple logo appears and disappears for the second time ) . It's tricky since modern Macs no longer chime at start up .

Reset NVRAM on your Mac - Apple Support

If one or more settings stored in NVRAM don't work as expected on an Intel-based Mac, resetting NVRAM might help.

support.apple.com

SMC ( logic board ) reset happens when you unplug the power cord for at least 15 seconds , then reattach the plug and wait 5 seconds before you press the power button .

Reset the SMC of your Mac - Apple Support

The SMC is responsible for managing power on your Mac. Resetting it can resolve certain uncommon issues related to power or thermal management.

support.apple.com

 

[Snow Tiger]

Well , my used 28 Core W-3275M Xeon was finally delivered and installed in my factory stock 8 Core MP7,1 .

It was successful , of course , and also survived a NVRAM reset . It also passed the built in Apple Diagnostics ( D key pressed at start up ) with flying colors .

Geekbench 5 score is a little low ( which is as expected ) because only 4 memory modules are installed . To get the highest GB CPU score for this model you need to install the memory in matching sets of 6 or 12 modules .

 

[majus]

Well , my used 28 Core W-3275M Xeon was finally delivered...

How is it that these fairly new processors are already turning up on the used market unless, as you said, they are rebadged? Are they maybe somehow defective?

 

[Snow Tiger]

How is it that these fairly new processors are already turning up on the used market unless, as you said, they are rebadged? Are they maybe somehow defective?

I hope my chip was not stolen . The build quality is not the best I've ever seen from Intel . I took an USB microscope and examined the entire land grid array side of the silicon , including the SMCs because I worried I was receiving a broken item . One or two of the resistors / capacitors seemed soldered on at a funny angle , which annoyed me as I expect perfection with a piece of silicon that has a MSRP of 7 grand . I wonder if Intel has factory seconds ? ? . Well , of course they don't . Right ? Can any one in Shenzhen verify there is no Intel Factory Outlet Center there ? ?

The IHS markings indicate it is a production unit and it is reported as genuine using Intel's MacCPUID checking utility . There is a serial number in QR code format .

I will examine it further in a Windows environment in order to run the latest version of CPU-Z , which should verify if it is really a production , ES or QS chip .

All the same , the chip seems to be behaving as expected . I am pretty happy , considering how every time I swap out the silicon in these machines I get another grey hair .

 

[MacFlaX]

Well , my used 28 Core W-3275M Xeon was finally delivered and installed in my factory stock 8 Core MP7,1 .

It was successful , of course , and also survived a NVRAM reset

These are really good news for buyers of the base model with intention to potentially upgrade CPU in future (like me). It's a confirmation that Apple security/T2 doesn't block CPU upgrade, at least for not for today's BTO options.
Interesting to know would be if this is valid for the W-3265/75 non-M version as well, which may be good enough for many of us.