2019 Mac Pro is Now Available at Apple.com

[D-One]

The fact that a laptop can often times outperform a Mac Pro with 50% more cores, goes to show that even apps that Pros use, can favor low core-count but higher frequency CPUs. Also the new Navi GPU in the MacBook Pro seems to work very well.

Imagine if, without testing, he had bought a maxed out Mac Pro, thinking that for his specific Pro tools, more cores/more expensive = better investment/productivity. He would have wasted his money and gotten worse performance.

Arent the adobe apps (specifically premiere and after effects) notoriously known for being optimized for NVIDIA thus not performing so good in MacOS? At least that's my understanding.

Folks also mentioned that GPU utilization is still low and should improve within the next os versions which explains low performance in some scenarios, so I am not sure those early results are representative enought to take conclusions as of now... IMHO, could be wrong tho...

 

[Gigamaxx]

Arent the adobe apps (specifically premiere and after effects) notoriously known for being optimized for NVIDIA thus not performing so good in MacOS? At least that's my understanding.

Folks also mentioned that GPU utilization is still low and should improve within the next os versions which explains low performance in some scenarios, so I am not sure those early results are representative enought to take conclusions as of now... IMHO, could be wrong tho...

The Mac Pro has one well priced option the $2000 Afterburner card which works as a graphics cruncher for video rendering. I’ve seen videos of it at work and the CPU and GPU are basically idle while it plows through video rendering tasks.

It’s supposed to be for sale on Apples site but it’s not available yet, only with the MacPro when ordered.

Making a 16K Video on the 2019 Mac Pro

 

[avkills]

The Mac Pro has one well priced option the $2000 Afterburner card which works as a graphics cruncher for video rendering. I’ve seen videos of it at work and the CPU and GPU are basically idle while it plows through video rendering tasks.

It’s supposed to be for sale on Apples site but it’s not available yet, only with the MacPro when ordered.

Making a 16K Video on the 2019 Mac Pro

Adobe apps support CUDA, OpenCL and Metal acceleration. The Afterburner card only aids in decoding ProRes files (at least that is what Apple's website says); it does not speed up encoding. If applications like Millumin or Mitti can tap into that card it is going to be a game changer for playing tons of media back to LED walls with massive pixel counts.

Afterburner is good because if the CPU and GPU do not have to decode then both can be used for effects and filters.

 

[avkills]

The whole point of this or any hackintosh community is to experiment with going off the straight and narrow path of doing things only the way Apple does them. It's all about research, posting the results of that research and then sharing that with the community. If no one ever tries that 16 core Xeon in a Mac Pro how will we know if it works or doesn't ? So if anyone doesn't want to take the chance that it might not work they're not obligated to. Someone probably will and then share what they find here. If they do buy the 16 core Xeon Gold from Amazon and for whatever reason they need to return it, they'll get a full refund with no questions asked. It's quite easy to return a CPU in my experience with them.

Any of the possible hardware configs for the Mac Pro I've posted are based primarily on saving as much money as possible to keep the total cost of ownership lower. I'm fully aware that there may be better choices that can be made by spending more. I'm in no way saying that the lower cost options are the only way or the best way to go. Someone with a $20,000 new Mac Pro budget is probably going to choose only what Apple offers from the factory and not make any hardware upgrades themselves.

I hope that you can understand what I'm doing here. What's the point in having a Mac Pro thread and saying "Spec it out with all the Apple factory upgrades you can afford" and leave it at that ?

No one can argue that Apple's hardware upgrade prices are up to 5x what they should be. Especially for ram, SSD storage and graphics. In my opinion even someone flush with cash should still do the upgrades themselves rather than throw more money at Apple's 230 billion dollar cash hoard.

Actually we do not know if the GPU prices are out of line because you can't buy either of the top tier options directly from AMD; but if you have priced any of the Radeon Instinct or WX Workstation cards, part of me can not help but think that the card options for the Mac Pro are actually not out of line; they are not gaming cards first and foremost. AMD charges a mint for those Instinct cards; which is the closest thing to what the Vega II/Vegas II Duos are compute wise.

 

[trs96]

AMD charges a mint for those Instinct cards; which is the closest thing to what the Vega II/Vegas II Duos are compute wise.

Yes, much of the cost of workstation cards are the drivers that they have to develop to make them work properly. Gaming graphics cards have no where near as much time and money invested in drivers for gaming purposes.

 

[trs96]

Here are the read and write speeds you'll get when combining four 1TB Samsung 970 Evo Plus NVME drives in software raid. The Sonnet M.2 4x4 PCIe card was used. The total cost for the drives and the Sonnet comes to less than what you'd pay for the Apple 4TB SSD pre-installed at the factory. 7.2 GB/s read speeds must be some kind of speed record. Apple NVME drives will get you up to about 3.0 GB/s.

 

[Sofronis]

In multithreaded workloads, an overclocked 10980XE 18-core, would come very close (if not surpass) the low frequency 28-core (W-3275) that Apple uses, and would easily best it in single/lightly-threaded workloads.

Mac Pro benchmarks had started popping up in Geekbench charts, so I decided to test this hypothesis; although we probably shouldn't draw too many conclusions (Geekbench is a synthetic benchmark and systems/scores vary).

I selected two Geekbench 5 CPU scores that are closest to the median for an 18-core (9980XE) hackintosh and a 28-core Mac Pro (removed any very low score outliers from both sets).

So compared to the 28-core Mac Pro, an 18-core hackintosh is:

• ~20% faster in sigle-core
• ~9% slower in multi-core.

A couple of notes:

1. The 28-core Mac Pro scores, vary by up to 30% and I think the culprit is the memory configuration (or misconfiguration for that matter). For optimal performance (at least under Geekbench) you should correctly populate all 6 channels (i.e. 6 or 12 DIMMS), ideally with identical DIMMs. You can find more technical info here from Lenovo (Intel Xeon Balanced Memory Configurations).
2. The hackintosh chart surely contains overclocked as well as unoptimized systems. If you overclock (the newer 10980XE also clocks higher) you'd probably get a higher score than 1324/16966. If you keep it stock it might be lower.

Here are the read and write speeds you'll get when combining four 1TB Samsung 970 Evo Plus NVME drives in software raid.

I'd expect higher throughput from a RAID 0 array, since each 970 Evo+ has a read speed of ~3 GB/s. There must be a bottleneck somewhere (perhaps it being software RAID?)

 

[trs96]

I'm hoping that @AnAppleADay or anyone else that has received their Late '19 Mac Pro recently will post the config of their hardware and run some benchmark tests to share here. Then we'll get a better idea what performance you're really getting by spending over 10K on the new Mac Pro.

 

[AnAppleADay]

More to come.

 

[AnAppleADay]

More to come.

First benchmark. Today I'll be switching over to a 970 Evo Plus NVMe M.2 2 TB. Assuming all goes well I'm going to leave the Apple proprietary 4 TB internal disk unused for now.

Apple SSD 4 TB (w/encryption):


970 Evo Plus 2 TB (w/o encryption):