Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
MP 7,1 CPU upgrade in 2019 Mac Pro
- Thread starter crystalidea
- Start date
- Sort by reaction score
As a System Builder , I always do this for a client's Mac after performing any hardware upgrade .
Also, an SMC reset may do more than resetting the NVRAM, tho most people unplug their Mac's when they add upgrades, so in the case of the older Mac Pro's this will reset the SMC.
I'm not sure the 2019 MP has an SMC, but if it does, there should be some documentation on how to reset it.
There are no reset buttons or diagnostic LEDs on the logic board of the Mac Pro 7,1 . So , I guess the way to reset SMC is to simply unplug the power cord for about 15 seconds and then reattach . Then wait a few seconds before turning on the Mac again . I haven't obtained the manual for this Mac yet .
It's foolish not to unplug a computer before adding or removing components , but with the MP7,1 it is a requirement that the power cord is unplugged before the user has the ability to remove the housing . Without removing the housing , access to the internal components is impossible .
Last edited:
I don’t enable any encryption that I cannot control.
If you have a couple minutes can you please check core speed running Intel power gadget? You can use Cinebench to maximize CPU usage.
must not have been many that fell off the back of the truck ?
seriously though that’s not just a good price, that’s an unexplainable low price.
Well , four things to consider here .
The chip just might be an engineering sample . For production level Macs , this would be a horrible upgrade since there might be a lot of undocumented errata . The chip would simply make a System unstable , defeating the whole purpose of a genuine workstation ( high performance , reliability and accuracy ) . Best to verify it with the latest version of an utility like CPU-Z . The heatspreader should also indicate if its an ES or you could always check it's S-spec number . The S-spec numbers are different for samples and production chips .
If someone were truly evil ( maybe from Dogbert's Discount CPU Emporium ) , they could remove the heatspreader from an engineering sample and replace it with one from a production version . There are heatspreader removal devices that aren't too difficult to use and how many end users would detect this 'ol switch-a-roo until its too late ?
The chip might be a genuine production version , but be physically or electrically damaged in some fashion . Would you install a processor that had it's substrate damaged or scratches across the lands ( those gold circles ) into the socket of a thousand dollar logic board ? When Xeons get too hot , I've had select features temporarily go offline ( e.g. hyper-threading , ECC memory ) in machines I was building . So, a chip might seem to work until you examine it closer , if it has had its internal circuitry damaged . At the very least get some actual high res front and back pictures of the very item you're considering purchasing .
The chip might be hot and I don't mean thermally .
The chip might be missing from a Terminator and his friends will come looking ... People always overlook this for some reason .
Last edited:
Thanks for reminding, I did that after cpu replace, but not yet for ram, will do it later,
[automerge]1577309534[/automerge]
is it possible to check the chip via cpuz without using bootcam on mac? is run on parallel destop will see the actual cpu?
I have using a lot of terminal command yes > /dev/null
to fully load the mac cpu, and run it for 6 hours until now, works fine, is the 6 hours long enough to test?
is this command give enough stress to the cpu?
in the meantime i also runs blenderbenchmark on top of yes command
still works fine
but there is one shut down when I run gpu test via blenderbenchmark on top of the fully load mac with yes command. my gpu is 580x
in this case will it consider cpu unstable or just too much stress on cpu and gpu make it shut down? it does pass the gpu test when not underfully load cpu with yes command.
[automerge]1577309614[/automerge]
Last edited:
"Thanks for reminding, I did that after cpu replace, but not yet for ram, will do it later,
[automerge]1577309534[/automerge]"
This is critically important since I was the first person to install an unauthorized processor ( Gold 6212U ) in the Mac Pro 7,1 and everything worked fine until I reset the NVRAM . She went dead after that ... fortunately , nothing was permanently damaged so it was either a procedural error on my part or a firmware incompatibility .
"I have using a lot of terminal command yes > /dev/null
to fully load the mac cpu, and run it for 6 hours until now, works fine, is the 6 hours long enough to test?
is this command give enough stress to the cpu?
in the meantime i also runs blenderbenchmark on top of yes command
still works fine"
Download a copy of the modern macOS version of Menu Meters available here :
This utility will give you a visual feedback of thread and core activity of your processor in the menu bar of your screen . When all the thermometers light up , you know you are pushing your processor to the max .
Right now , I am in the process of replacing all my 32 bit apps I use as a Builder , since Catalina will not run them ( 64 bit apps only permitted now ! ) . One of those apps is a CPU stress test . I think I have found a replacement program called GIMPS Prime Stress Test . Will run it soon ...
Before I release a System to client , it runs at full System load for at least 12 hours and preferably 24+ hours . Full System load means all the major components ( CPU , GPU , Boot drive , Memory , etc ) run concurrently at 100% load . That is a synthetic test and an extremely stressful one . If the test does not pass , I discover reason why .
Your Mac should never shut down for hardware reasons . That is always unacceptable . And macOS is a very stable OS , compared to Windows . So a software error is less likely with macOS . A badly written program also should not shut down a macOS session , due to preemptive multitasking technology .
A blender benchmark running a test with a GPU should also not shut down a System . Something is wrong . Your graphics card is brand new and so is your computer . The worst that should happen is a poor score , not a System shut down .
Reset your NVRAM and see if that makes your Mac perfectly stable .
You should also reset your SMC , by unplugging your Mac's power cord for 15 seconds , plugging it back in again and wait several seconds . Then start up your Mac . This should reset your logic board .
Movie reference. Come with me if you want to live.
Well, this didn't take long. A company in Milwaukee on eBay is offering a 7,1 CPU upgrade service for the 16-core W-3245 and 24/28-core W-3265/75 (not the M variant). You have to send them your machine, price includes return shipping. After some shipping estimating on my part it appears they are tacking on around $1,100-1,200 markup for their service. Pretty profitable for them if anyone bites.
Who would bite though? if I’m risking my system while under warranty I might as well do it myself and save even more money.
(I get while Apple decided to keep it simple, but thanks to intel its such a massive expense to get more cores without paying an arm and a leg for a higher RAM ceiling you may not need. Aftermarket upgrades down the line to non M high core Ws will I imagine be fairly popular.)
SO someone tell me if I am seeing this correctly. I, like many of you, had the same idea that if this machine is upgradable then why not get the base model and upgrade everything myself for less than Apple is charging, including the CPU. SO then I started thinking, what chip would I upgrade to? I settled on the 16 Core chip as a decent upgrade. I found on Every Mac that the 16 core Xeon used is a W-3245 which I can find for sale at more than Apple charges (2000 USD) for the upgrade from any reputable seller I would want to do business with.
SO, if thats the case, then why bother? I should just get the 16 core from the get go.
SO, if thats the case, then why bother? I should just get the 16 core from the get go.
At the moment, there's no financial gain to upgrading the CPU on your own outside of Apple, unless you opt for the non-M 24 or 28-core chip. If you do, you'll save upwards of $3-4K because the chip is actually a lot less expensive than the M version, but you'll be "limited" (heh) to 1TB of RAM.
In the future, that may not be the case. The Intel chips may and probably will come way down in price. But at the moment, not so much.
On Jan 2 2020, I'll be purchasing the 16 Core.
1) I don't need to immediately update upon arrival
2) I have the option to upgrade to a better chip later when necessary
Some of the issues which have been brought up previously about not being able to remove the Apple installed SSD drive due to issues with T2 or the OS.. I wonder if it would be acceptable to use the merging of a PCI-e card and the propriety apple SSD installed using the make a larger fusion drive. Ive done such things in my hackintosh in the past or my imac.. which has been a great way to make use of small SSD's.
Well , since the MP7,1 is a LGA 3647 socketed workstation and the Cascade Lake Xeon is the last silicon from Intel that will be designed for this socket , whatever W series chips that are currently available should be the most powerful verified chips we can use to upgrade this Mac with .
Intel might add to the catalog of W series Xeon in the future and we might just possibly get an interesting OEM chip not intended for retail release ( but will eventually show up as a rarity on eBay in the future ) . But I think the most powerful chip likely has already been released ( W-3275M ) . Intel is not going to go beyond 28 Cores with this series and it's really hard to bump up the frequency with so many cores .
The wildcard is the rest of the Cascade Lake catalog - the non W series silicon . If they are compatible . We , again , won't be able to go beyond 28 Cores so don't be so eager to look for magical performance beyond that which Apple is already delivering . The big advantage here is price and convenience . There are some genuine goodies Intel is shipping now because of the heated competition AMD is giving them at the low end single processor server space .
The more of these chips we can install as upgrades in the MP7,1 , the cheaper and easier it will be for us in the future . Wait for a larger number of these chips to come up as used on the market some three years after they are originally deployed .
I'm working hard to verify these non W Series Xeons are truly compatible with the MP7,1 . It's a matter of firmware ( the boot ROM and the iBridge versions ) . I can get these chips running in the Systems , but they don't survive a NVRAM reset . Since that's an operation that most Mac users will experience at some time in the life of their computers , the installed silicon has to pass that test or the upgrade is just an academic experience .
Another issue we face with upgrading the processors in this Mac is this : the retail availability of the W Series Xeons is real thin . Try buying a brand spanking new W-3275M . The officially reported Tray price is $7,453.00 . You'll pay a bit more since you're just buying one . Try finding one available today without having to special order it ( which takes a month or so for delivery in the USA , from Intel ) . You won't at that price . That's also true with most of the line . Nobody has these available at retail - you'll find just one today or three today available . And then they're gone . So , the prices are artificially high at the moment . You might just want to wait until the market settles down and supplies are more reliable . That'll reintroduce competition and lower prices .
Intel might add to the catalog of W series Xeon in the future and we might just possibly get an interesting OEM chip not intended for retail release ( but will eventually show up as a rarity on eBay in the future ) . But I think the most powerful chip likely has already been released ( W-3275M ) . Intel is not going to go beyond 28 Cores with this series and it's really hard to bump up the frequency with so many cores .
The wildcard is the rest of the Cascade Lake catalog - the non W series silicon . If they are compatible . We , again , won't be able to go beyond 28 Cores so don't be so eager to look for magical performance beyond that which Apple is already delivering . The big advantage here is price and convenience . There are some genuine goodies Intel is shipping now because of the heated competition AMD is giving them at the low end single processor server space .
The more of these chips we can install as upgrades in the MP7,1 , the cheaper and easier it will be for us in the future . Wait for a larger number of these chips to come up as used on the market some three years after they are originally deployed .
I'm working hard to verify these non W Series Xeons are truly compatible with the MP7,1 . It's a matter of firmware ( the boot ROM and the iBridge versions ) . I can get these chips running in the Systems , but they don't survive a NVRAM reset . Since that's an operation that most Mac users will experience at some time in the life of their computers , the installed silicon has to pass that test or the upgrade is just an academic experience .
Another issue we face with upgrading the processors in this Mac is this : the retail availability of the W Series Xeons is real thin . Try buying a brand spanking new W-3275M . The officially reported Tray price is $7,453.00 . You'll pay a bit more since you're just buying one . Try finding one available today without having to special order it ( which takes a month or so for delivery in the USA , from Intel ) . You won't at that price . That's also true with most of the line . Nobody has these available at retail - you'll find just one today or three today available . And then they're gone . So , the prices are artificially high at the moment . You might just want to wait until the market settles down and supplies are more reliable . That'll reintroduce competition and lower prices .
Last edited:
Is it possible the 6212U failed the NVRAM reset because you had two DIMMs per channel installed? The 6212U does not appear to support that configuration.
You're confusing two different types of memory products .
NVRAM is a small amount of fast ram on the logic board that controls certain functions ( display settings , time zone , etc ) whose settings remain even if the Mac loses power .
DIMM is main system memory and is much larger . It's where your data and machine code are temporarily stored during an OS session . This information is lost when your Mac loses power .
This Mac has a Cascade Lake Xeon . The LGA 3647 Cascades have two memory controllers for a total of six memory channels that support up to 12 DIMMs . Two DIMMs per channel . We're still trying to verify memory configuration optimization , but it appears six matching modules will provide 97 % max memory bandwidth and twelve matching modules will provide 100 % max memory bandwidth .
My Mac came from the factory with an unoptimized DIMM memory configuration ( four 8 GB modules ) , but that would not prevent a NVRAM reset .
I'm still looking into the NVRAM reset failure . Either its a failed procedure on my part , bad luck or the firmware does not fully support the processor .
Yep, I am ordering the 16 core from Apple. in the future, well, we'll see what develops.