World’s Fastest Hackintosh Ever! 32/64 Core - Mac Pro Killer! Morgonaut’s Mac Pro annihilator

[sakagura]

check this

World’s Fastest Hackintosh Ever! 32/64 Core - Mac Pro Killer!

Not sure why she has 32/64 in title, leads me to think 32 core 64 threads but she shows 64 cores in System Profiler

but there’s no EPYC at 3.7ghz with 64 cores

So maybe it is a TR3 3970X (a $2000 CPU) and System Profiler she shows is incorrectly misrepresenting threads as cores?

wow, smokin

But what's he doing with the machine? All those cores to make youtube videos?
[automerge]1590841646[/automerge]

View attachment 918960

That's not too great for 12 cores. I can OC a 9900K to 5.2ghz and get 1450 single core and 10,000 multicore. The overall performance is much slicker with 4 cores less.

 

[mmomega]

That's not too great for 12 cores. I can OC a 9900K to 5.2ghz and get 1450 single core and 10,000 multicore. The overall performance is much slicker with 4 cores less.

It is ok for a 12 core device.
It isn't overclocked. There is a significant difference between 5.2 Overclocked single core and a 4.4 to 4.5 non overclocked single core.
Without even sitting there to try and dial in perfect settings. So this is completely stock, on an operating system that is hacked to run that CPU, definitely not close to being optimized.

In Windows on the exact same system, again not overclocked, default bios settings, microatx board, not fast RAM and on an air cooler. That CPU got over 13,000.

So now if we go back to a Mac for a reference point to the only CPU that I have with more cores. Is the Intel Xeon W 2195 18 Core. And that machine got less than 14,000.

Then if we are talking about a non-overclocked 9900K in macOS, then we are looking at the 8,000 point area.

So if we only check the difference between the stock 9900K on macOS vs the 3900X on macos (hacked) then there is a 32% increase.
9900K at $530
3900X at $418
So a 26% higher cost, only for the CPU, and a 32% decrease in performance. That is just whatever that means for any particular person.
And from overclocked a 10,000 Geekbench to the 11190 Geekbench, that is an 11% increase in score, for 26% less cash on 1 component with no effort outside of pressing the test button.

Then if we just want to show 8Core vs 8Core, I have a 3800X on stock air cooler. And it benchmarks higher than the 9900K. No, everything was not controlled exactly to the same specs and probably never will. This is a Plex server that I'm never going to touch again outside of cleaning dust out or adding additional drives occasionally.

I have a machine that is full custom looped EKWB and I use to chase the highest benchmark number I could afford, then for myself the amount of time I put in to playing, did not net me any meaningful results.
That machine is now my sons racing sim PC.

I am just not at a point any longer, and many are and that is ok, that I do the overclock benchmark game. I'll benchmark some systems just to show a measurable number that does not take too much time to perform. Benchmark same systems in between build numbers to see if performance significantly increases or decreases but that is pretty much the extent of what I feel like doing.

Congratulations on your accomplishment and that you have a system you are happy with.

 

[handheldgames]

But what's he doing with the machine? All those cores to make youtube videos?
[automerge]1590841646[/automerge]

That's not too great for 12 cores. I can OC a 9900K to 5.2ghz and get 1450 single core and 10,000 multicore. The overall performance is much slicker with 4 cores less.

@skagura - Nice to hear your 9900k can keep up. How much voltage / heat is required to achieve that score. I'm also wondering, if the performance you are getting is before or after the multiple microcode mitigation's implemented by Intel.

Based on my own Geekbench Scores, the 3900x @mmomega used for Geekbench does not appear to be overclocked, nor does it need it to get a great score at a $419 msrp.

The lack of performance impacting security mitigations and the price comparison for an out of the box 3900x against the 12-core MacPro 7,1 is where the AMD CPU hits the ball out of the park and the main reason I went with a Ryzen based hack.

 

[Tiem]

Love this thread, so much disinformation and useless benchmarks.

First off, Geekbench is garbage. It does not even remotely reflect real world usage. One of THE worst benchmark tests that can be manipulated with Framebuffer entries to almost double. Totally artificial with zero actual performance gains.

Moreover, run GB 10x. How much variance do you see? Because many people make tweaks and report modest boosts. Going from say 54k to 64k. Then they run it again and see 56k. The reliability is one of THE worst. So if that's your basis for "beating" Apple... LOL

AMD processors are getting all the attention right now, but they can't reliably run Adobe apps. In fact, lots of production quality apps are just not meant to handle AMD and there's no amount of flags you can set in OC or Clover to fix that. Everything on macOS is meant for Intel and you get the best performance and reliability running the right silicon.

The MacPro7,1 SMBIOS too is not without it's fair share of headaches. No Sidecar. And that's only possible with iMac19,x. Which runs macOS like a bag of sand (re-encode a video using QuickTime and tell me what you see). You get Sidecar at the cost of quite a bit of performance. So yay? You could run iMacPro1,1 but that has a T2 chip, which Hacks won't ever support, so again, you lose Sidecar and you lose Hey Siri (not really possible to hack around since they changed dictation to Voice Assist and that has a little toggle overlay. While MacPro7,1 (no idea why she's running 1,1) is ideal to run if you have a beefy system, it doesn't come at a cost. Also the new Memory Manager needs its own kext and that hasn't seen much development. Nor will you ever get a proper working PCI tab, just saying.

It's absolutely smarmy af to claim this is a real Mac Pro killer when it's not just about the cores or stupid Geekbench numbers. It's about architecture and bus speeds and many many other components that Apple has built and optimized to death. Something even the most arduous Hackintosher could just never achieve.

Her and Snazzy are in the same league tbh. Padding their numbers selling snake oil. When the reality is that Hacks aren't perfect and won't ever kill a real Mac if you actually make valid tests.

But you know, keep drooling over those GB numbers folks...

 

[handheldgames]

Love this thread, so much disinformation and useless benchmarks.

First off, Geekbench is garbage. It does not even remotely reflect real world usage. One of THE worst benchmark tests that can be manipulated with Framebuffer entries to almost double. Totally artificial with zero actual performance gains.

Moreover, run GB 10x. How much variance do you see? Because many people make tweaks and report modest boosts. Going from say 54k to 64k. Then they run it again and see 56k. The reliability is one of THE worst. So if that's your basis for "beating" Apple... LOL

AMD processors are getting all the attention right now, but they can't reliably run Adobe apps. In fact, lots of production quality apps are just not meant to handle AMD and there's no amount of flags you can set in OC or Clover to fix that. Everything on macOS is meant for Intel and you get the best performance and reliability running the right silicon.

The MacPro7,1 SMBIOS too is not without it's fair share of headaches. No Sidecar. And that's only possible with iMac19,x. Which runs macOS like a bag of sand (re-encode a video using QuickTime and tell me what you see). You get Sidecar at the cost of quite a bit of performance. So yay? You could run iMacPro1,1 but that has a T2 chip, which Hacks won't ever support, so again, you lose Sidecar and you lose Hey Siri (not really possible to hack around since they changed dictation to Voice Assist and that has a little toggle overlay. While MacPro7,1 (no idea why she's running 1,1) is ideal to run if you have a beefy system, it doesn't come at a cost. Also the new Memory Manager needs its own kext and that hasn't seen much development. Nor will you ever get a proper working PCI tab, just saying.

It's absolutely smarmy af to claim this is a real Mac Pro killer when it's not just about the cores or stupid Geekbench numbers. It's about architecture and bus speeds and many many other components that Apple has built and optimized to death. Something even the most arduous Hackintosher could just never achieve.

Her and Snazzy are in the same league tbh. Padding their numbers selling snake oil. When the reality is that Hacks aren't perfect and won't ever kill a real Mac if you actually make valid tests.

But you know, keep drooling over those GB numbers folks...

Interesting... Sidecar has no value for a lot of users. Personally, I really don't have a value to use a spare iPad Pro as a mac display when I have multiple Display Port ports outputs available on a 5700xt as well as spare display port monitors available at my disposal if I needed them.

Seriously.. None of the issues you mention are must haves and I'd rather have the performance of a $8,000 7,1 with a sub $2000 3900x system. FWIW... the t2 is a disaster in the 7,1 as well as a iMac pro. Skipping Apple's custom - error prone silicon is one of the factors that makes the Hackintosh a plus. Side stepping Intel's ongoing performance / security mitigation's is another benefit of going AMD.

Is an AMD mac 100% reliable as a software engineer?? YES.
Was this a surprise?? Absolutely.

 

[sakagura]

@skagura - Nice to hear your 9900k can keep up. How much voltage / heat is required to achieve that score.

HWmonitor app says in Cinebench it can go around 1.37 - 1.38v on CPU and about 88-90c. But this is an extreme case. In real world apps the CPU will rarely touch max frequency and voltage. Even gaming on max settings won't push that temperature. In any case I don't keep the system overclocked if I don't need it. It's a one click overclock in the BIOS when I want it.

Geekbench rarely pushes any CPU to max clock, voltage or temp. It is very sensitive to overclocked XMP memory though. Faster and quad channel ram will boost those scores giving a false impression of better CPU performance.