MP All Models Mac Pro CPU Compatibility List

[h9826790]

Just saw this delid tool which is specially made for 1366 Xeon

https://rockitcool.myshopify.com/products/lga-1366-delid-tool?variant=2388025671693

 

[ActionableMango]

Nice!

Back in Feb 2017 I emailed Rockitcool to let him know about demand for delidding the older Xeon processors suitable for Mac Pro. He was already aware of the demand and had a prototype. Nice to see it available now, and apparently it is successful since it is currently sold out.

 

[Vondust]

I'm sure you've already made the purchase of the X5690, but if not, take a look at the X5680. It's roughly half the price of the X5690 (roughly, and you might find a rare good deal on a X5690), but is just a teeny bit slower. About 3-4% slower. You won't notice it, and you can save money.

Also get the X-series over the W-series (again, unless you find a rare ultra-low price on a W chip) because you never know if you might get a dual CPU tray or dual CPU MP in the future, and you can use your existing X-series and would only need to buy one more. The W series cannot be used in a dual config.

Have fun!

Well I was weighing the option of getting the X5680 instead but it was "only" 24$ I would save on it.
I have not received my X5690 yet, but I'm very excited and when I have that installed I will restart my trial and error on my gpu bios flashing (cant get more then one port working with boot screen).

 

[Jayson A]

Would I notice much of a difference between my quad-core W3565 and the 6-core W3670?? Should I even bother?

 

[h9826790]

Would I notice much of a difference between my quad-core W3565 and the 6-core W3670?? Should I even bother?

For what usage?

 

[kohlson]

Would I notice much of a difference between my quad-core W3565 and the 6-core W3670??

In all likelihood, nope. Same single core performance. In most uses I saw essentially no difference between my 4x2.93 and the upgrade 6x3.33. Except video. Glad I did it since that is my main use on this system.

 

[Jayson A]

For what usage?

Gaming. I was wondering if it would tide me over until I can build a legit rig.

 

[h9826790]

Gaming. I was wondering if it would tide me over until I can build a legit rig.

99% no difference, only very very few games can utilise more than 4 cores.

 

[Jayson A]

99% no difference, only very very few games can utilise more than 4 cores.

Thanks. I won’t bother then.

 

[macnu]

In case it helps someone:

In January of this year 2018 I decided to update my macpro 5.1 (mid 2010) 2x 2.4GHZ (E5620), I bought 2x 3.46GHZ (x5690) processors on ebay (Matched pair) and I am delighted with the result, not bad for an "old rocker" ...
No problems since I installed them with high sierra or now with mojave.

 

[SamPotts]

Recently upgraded my refurbed quad-core 6,1 trash can to a 2667 v2. The swap went relatively well other than it having the tamper evident screws. I've ordered some replacements from a supplier in case I need to return the Mac Pro for service - hopefully they don't make a note of which systems had them?! CPU temps seem to have decreased around 5-10c too with the new thermal paste. The stock paste wasn't very well applied and there was a big spot in the middle with little to no paste. Replaced it with Gelid paste.

Anyway, the scores are 4055 for single core and 26060 for multi score which I thought was pretty decent! My 2018 MacBook Pro 15" six-core outperforms it by a fair margin in single core but still fairly decent. Results are here: http://browser.geekbench.com/user/100016

 

[tpivette89]

Recently upgraded my refurbed quad-core 6,1 trash can to a 2667 v2. The swap went relatively well other than it having the tamper evident screws. I've ordered some replacements from a supplier in case I need to return the Mac Pro for service - hopefully they don't make a note of which systems had them?! CPU temps seem to have decreased around 5-10c too with the new thermal paste. The stock paste wasn't very well applied and there was a big spot in the middle with little to no paste. Replaced it with Gelid paste.

Anyway, the scores are 4055 for single core and 26060 for multi score which I thought was pretty decent! My 2018 MacBook Pro 15" six-core outperforms it by a fair margin in single core but still fairly decent. Results are here: http://browser.geekbench.com/user/100016

Nice scores! It's cool that the 8-core trashcans are getting equal multi-core scores as the 12-core 3.46ghz cheesegraters.

Any chance of running cinebench?

 

[SamPotts]

Sure. It seems to get around 1225. Had one 1233 and one 1222.

My 5960x Windows machine usually gets ~1700 but then it's overclocked and newer architecture. I'm pretty happy with how the Mac Pro is performing for what I need it to do.

 

[FlorisVN]

Am I correct that the Mac Pro 4,1 (2009) single cpu model does NOT need a Delided IHS CPU ?

Which makes upgrading the single 2009 Mac Pro version, a lot easier than the dual cpu 2009 MacPro.. ?
So single cpu 2009 would be same like upgrading the 2010-2012 Mac Pro versions.. ?

 

[tsialex]

Am I correct that the Mac Pro 4,1 (2009) single cpu model does NOT need a Delided IHS CPU ?

Which makes upgrading the single 2009 Mac Pro version, a lot easier than the dual cpu 2009 MacPro.. ?
So single cpu 2009 would be same like upgrading the 2010-2012 Mac Pro versions.. ?

Yes for all.

 

[FlorisVN]

Yes for all.

ok thx for your answer !

 

[FlorisVN]

@tsialex

One more question btw.
Tighning the CPU heatsink screws on a single CPU 2009, is also the same method I presume as the 2010's ??
So exactly counting the screw cycle times is not necessary , which makes it also a lot easier. ?

As far as I have seen, with delided xeon's putting the cpu heatsink bank on is very painfull process.
but then only for the dual cpu 2009 version I presume.. ?

So just tighten those cpu screws like you can not go further until it stops, and breaking the socket is them nearly impossible.. ?

 

[tsialex]

@tsialex

One more question btw.
Tighning the CPU heatsink screws on a single CPU 2009, is also the same method I presume as the 2010's ??
So exactly counting the screw cycle times is not necessary , which makes it also a lot easier. ?

As far as I have seen, with delided xeon's putting the cpu heatsink bank on is very painfull process.
but then only for the dual cpu 2009 version I presume.. ?

So just tighten those cpu screws like you can not go further until it stops, and breaking the socket is them nearly impossible.. ?

2009 single CPU Mac Pros backplanes and CPU trays are identical to the 2010/2012 ones in everything except the SMC firmware version.

Every nut and bolt, every component, is the same.

 

[h9826790]

@tsialex

One more question btw.
Tighning the CPU heatsink screws on a single CPU 2009, is also the same method I presume as the 2010's ??
So exactly counting the screw cycle times is not necessary , which makes it also a lot easier. ?

As far as I have seen, with delided xeon's putting the cpu heatsink bank on is very painfull process.
but then only for the dual cpu 2009 version I presume.. ?

So just tighten those cpu screws like you can not go further until it stops, and breaking the socket is them nearly impossible.. ?

Correct, Apple technician manual suggest "finger tight + 1/4 turn", that's it.

 

[FlorisVN]

thanks guys for your info !

 

[razorfsh]

I'm totally confused about all this ********...

after upgrade macpro 2010 to 2X 5690 3.47GHZ have not seen much of an increase in speed
ram 16GB 1333->96GB 1600 (1333)

my "old" cpu 2.33 & RAYDON top of the line were giving better "benchmark" results than my nvidea GTX1070

 

[h9826790]

I'm totally confused about all this ********...

after upgrade macpro 2010 to 2X 5690 3.47GHZ have not seen much of an increase in speed
ram 16GB 1333->96GB 1600 (1333)

my "old" cpu 2.33 & RAYDON top of the line were giving better "benchmark" results than my nvidea GTX1070

Thats becasue you use a wrong benchmarks.

E.g. Cinebench OpenGL test

 

[razorfsh]

Thats becasue you use a wrong benchmarks.

E.g. Cinebench OpenGL test

I started with cinebench and thats where i got the info

5870 1G + 12c 2.66 (57.49)
is beating a
1070 8G + 12c 3.46 (56.12)

Then i moved to other test tools
CUDA-Z /ocean wave and the results are not consistent

 

[h9826790]

I started with cinebench and thats where i got the info

5870 1G + 12c 2.66 (57.49)
is beating a
1070 8G + 12c 3.46 (56.12)

Then i moved to other test tools
CUDA-Z /ocean wave and the results are not consistent

That's exactly what I said. You use the wrong tool to measure the performance.

CineBench OpenGL test is extremely CPU single core performance limiting. And the GPU driver performance is also CPU single thread limiting.

So, you upgraded the CPU from 2.66 to 3.46. If you keep the 5870, the performance will increase to 70FPS range.

But if you swap to 1070, due to the higher overhead of the Nvidia web driver. The performance increment is completely offset by the driver efficiency decrement. So, roughly the same result.

1070 is a better card, but it's not a tool for everything. You still have to pick the correct tool for the correct job. For some simple stuff, 5870 can do better because has better driver support in macOS.

 

[razorfsh]

I don't normally play games....
Just wonder how the F** a 2009 GPU can out perform a card available in 2018 on any test.