New Mac Pro!


#41

I’m just curious why they went to AMD for GPU but stick with Intel Xeons for CPUs when they don’t seem to offer dual CPU options (or so I’ve read) and AMDs new chips compare more than favourably pricewise and in performance (Threadripper or Epyc?) especially the upcoming 7nm ones.

Apparently the GPU and CPUs are passive cooling only, relying on the front ventirads to cool everything inside. That means probably quieter at first, but also no watercooling option, and with such power sucking cards as dual vegas, throttling is a concern.

As for the monitor stand, there isn’t a basic one included apparently. You either buy the vesa mount (and ergotron arm) or go for the $1000 one.

They’re just being cynical about the whole thing: people interested about it will buy it at almost any cost, anyway, just like the rest of the system. That’s what they did with the iphone.

It has become a luxury item, like a birkin hermes bag. It’s not worth $20.000 (used!) in material or craftsmanship, it’s for the status it brings.


#42

[quote=“EricM, post:41, topic:2052024”]
I’m just curious why they went to AMD for GPU but stick with Intel Xeons for CPUs when they don’t seem to offer dual CPU options (or so I’ve read) and AMDs new chips compare more than favourably pricewise and in performance (Threadripper or Epyc?) especially the upcoming 7nm ones.[/quote]

Top of the line TR maybe a great rendering CPU but it’s far from Intel performance for single threaded tasks(edit, I mean specifically the C4D viewport), for example(not coming from synthetic benchmark) the 32core is basically as fast as my old 2013 8core MP in Cinema4D wiewport, that’s not too bad but according to benchmark made using a real 2GB scene the new Xeon 28core should be more than 3 times faster than my old MP, now do your math;)
Yes, the upcoming 3gen Threadripper may change that and still cost a little less, but until you can get one you never know.
Epyc is not a workstation CPU, costs more than TR and single threaded tasks will be even worst, you can say the same about the dual Xeon as well.

[quote=“EricM, post:41, topic:2052024”]
Apparently the GPU and CPUs are passive cooling only, relying on the front ventirads to cool everything inside. That means probably quieter at first, but also no watercooling option, and with such power sucking cards as dual vegas, throttling is a concern.[/quote]

Well, not really, did you notice the massive heatsink on the GPU module? that alone has more TDP capacity of the entire 2013 MP(and no, despite the common mith coming from disinformed people the 2013 does not throttle even under full load).

[quote=“EricM, post:41, topic:2052024”]
As for the monitor stand, there isn’t a basic one included apparently. You either buy the vesa mount (and ergotron arm) or go for the $1000 one.[/quote]

AFAIK there is a basic stand included. Realistically the average price of the 2019MP plus new display will be around 20k, so investing a 5% more for the stand will not make a great difference.

[quote=“EricM, post:41, topic:2052024”]
people interested about it will buy it at almost any cost, anyway, just like the rest of the system. That’s what they did with the iphone.[/quote]

Not necessarily, I will probably buy the 2019 MP but I’m still happily using a 5 years old iPhone;)


#43

threadripper 2950x for example has a R20 single core cinebench of 425.
While a Intel Xeon W-3175X, the 28 core apple is talking about has a r20 single core speed of 432

sources:


I can’t see a big difference.


#44

First, that’s not the CPU Apple will use, it’s the old generation. Then, synthetic benchmark are one thing, real world it’s quite often a very different thing. Don’t know about the 2950(16core) but the 2990(32core) with a 1080Ti using the same scene runs the viewport more or less like my old Intel 8core, using my old FirePro 700.
I’m a big fan of AMD CPU when it comes to multithreading, I’m using one in my render nodes and it’s great, but it’s not a mystery that Intel it’s quite better at single threaded tasks, no matter if it’s on Windows or MacOS, that’s why AMD has worked hard to improve the IPC on the next generation(not yet announced).


#45

I beleve you. but no one of us knows how fast the mac pro will be. you clamed it will be 3 Times as fast and I doubt it.
the fastest single core speeds have i9 cores right now. The 2xxx threadripper is around 15% to 20% slower than theese. xeons are slower then I9 as they normaly have slower clock speeds but more cores. Therefore the Threadripper is quite competitive. the already anouced Rizen 3000 series has singlecore speeds on paar with the I9. therefor the Threadripper 3xxx are expected to be as fast as any Intel CPU.

best regards
Jops


#46

To be honest I think it won’t be 3 times faster on average, that’s probably the best case scenario(but since 2GB is my average scene size it will be great at least for me;) ) non the less I’m expecting it to be far ahead at least compared to the current generation of TR or my old 8core Xeon.
I should be more clear about my previous posts, I do not mean it will be 3( or 2 or wathever) times faster in every single threaded tasks, the difference normally should be 15-20% like you noted, I mean specifically for Cinema4D viewport editing. While benchmark may give Intel only a little advantage in single core, even Cinema4D OpenGL benchmark(not available anymore in version 20) seems to support what I’m talking about: https://www.kitguru.net/components/james-morris/armari-amd-ryzen-threadripper-2990wx-32-core-threadripper-2-workstation-review/all/1/
Also, while its true that Xeons usually have slower turbo clock, the W line is specifically conceived for Workstation and basically brings all of the advantage of the Xeon core count and features(beside the scaling to 2-4-8 way option) and most of the i9 single core speed, the 3175X have also unlocked multipliers(the MP will use the next generation), the i9 can OC easily though since core count it’s usually lower.


#47

I didn#t have time to read it compleetely, but the benchmark you linked seems compares amd to nvidia grafics cards:
“Cinebench R15’s OpenGL test is not a strong point for current AMD graphics. The Radeon Pro WX 9100 managed better than the AMD graphics in the V25R and S16T, but wasn’t a match for the NVIDIA graphics of the Scan or S18X.”

It would be nice If we could expect to gain a serious speedup in editor speed. but I have little hope.
I wouldn’t put to much hope into the new Mac in this respect.


#48

Guess what? Cinema4D OpenGL score(and viewport editing) it’s mostly dependent on the CPU turbo speed rather than on the GPU itself, that’s why if you install a modern RTX 8000 card(about 10.000 $) on an old PC or Mac(with working drivers), it will not improve the editing;) as already said the 2990 I compared my old Xeon to was running a 1080Ti so it doesn’t matter that much if you run Nvidia or AMD GPU
The question it’s not about having hope in the upcoming machine just because it’s a Mac, if you run a recent Intel with fast turbo then it won’t improve that much the editing, but coming from old machine like mine or a 32core AMD you will fell a large difference, BTW you can experience similar performance on a Windows machine as well and right now(mostly), it will cost a lot though…
-3175 costs 3000$ and the upcoming version will cost significantly more unfortunately: https://ark.intel.com/content/www/us/en/ark/products/series/125035/intel-xeon-w-processor.html
-dedicated 3175 PC Mobo costs about 2000$ and still lack many features present on the MP(seems folks here forget that little detail when pricing what’s included in the basic machine)
-add ECC RAM, large PSU ecc and you will end up on a 10k $ system even if you build it for Windows by your self, of course if you get the same specs from HP or similar it will cost you about the same as the MP since usually R&D have a cost.


#49

I would expect that Cinema 4D would improve the viewport to not be so dependent on single core speed.


#50

Its interesting to see these developments at WWDC. A few thoughts:

  • Charging $1000 ea. for a stand and for a matte upgrade on a $5000 monitor is a dick move, clearly. On the other hand, this monitor is meant for a very rarified audience: film studios, vfx facilities, post houses, and the highest end photo shops. It really isn’t a commodity monitor, and the vast majority of Mac Pro users will be using third party displays.

  • The Mac Pro is also super-expensive, but it is arguably a decent value compared with (truly) comparable systems. And again, it will do things that are simply not available now for the price (though mostly in the area of editing/grading/post work).

  • The Platform State of the Union (which is the dev focussed keynote at WWDC) revealed some very cool things about Metal 3. Many more things moved to the GPU (the bounding volume hierarchy, denoising), and they showed a test app running a scene doing realtime raytracing (AO, reflections, diffuse light, denoising) at 4K at 30fps (see here at around 1:47…). This shows that the sophistication of Metal Compute shaders has reached what is necessary for Otoy and Redshift to implement their render engines on Mac OS.


#51

I guess if you are not happy with your TR editor speed, you will not be happy with the editor speed of the mac pro.
But as you seem to plan to buy one we will see and you can prove me wrong afterwards.

till then, best regards
Jops


#52

Well it will be likely twice as fast or more for editing than my old MP or my TR, and it will also have 5 times faster render times compared to my actual editing machine, that will avoid using my Windows farm(both Intel and AMD CPU nodes) most of the time, I think I’ll be quite happy;)
No reason to wait to prove you “wrong” :slight_smile: there are tons of benchmark on the 3175 right now and it beat the TR 2990 in 90% of the benchmark, sometime by very little margin, other times by a significant margin like in Corona and Vray, and in specific tasks like C4D viewport editing by a large margin, this is proved by both synthetic and real world tests.
Of course the upcoming version used in the 2019MP will be even faster.
On the other end the TR it’s a lot cheaper and it’s my number one choice for building a new Windows node, if and when the TR gen 3 will match or exceed the Intel counterpart on IPC, then it will be a great CPU for fast WS usage too, and if cheaper even better;)


#53

No NVIDIA, ridiculous prices, promises of maybe some day being able to run Octane.

Same exact reasons why I was forced to switch to Windows three years ago.

Great job, Apple.


#54

I wish you luck. if the mac pro was more then lets say 30% faster then the current TR generation in editor speed, that would be a fantastic outcome for all of us.


#55

Agreed:)


#56

I’m not a fan of GPU renderers but people interested may find this useful: https://home.otoy.com/octane-x-wwdc2019/
Apparently Otoy is giving a full Octane commercial license for free for 2019 MP owners.


#57

More info here. Looks like a lot of companies are getting to grips with Metal. Exciting times!


#58

Both Redshift and Otoy say their respective GPU render engines are coming to metal at the end of the year. I guess I can wait and see what happens.

Go to 1:49:49 to see Otoy strutting their stuff with Metal 3 (and no doubt some expensive AMD hardware):


#59

That’s not Otoy, it’s an Apple dev. Their team knocked together a hybrid raytracing engine with Metal’s new APIs.


#60

My mistake, I thought I saw an OTOY logo on the train in there.

In my defense, I am actually having severe eye problems this week. :wink: