MP 14,8 2023 Mac Pro is very useful

[Ethosik]

The Extreme would likely cost at least $10000, have an excessive number of CPU cores, and only have the GPU power of a PC with a single $1500 graphics card. It would certainly be powerful, but not especially cost effective. And still not upgradeable in any meaningful way. Though its role in the range would be a lot clearer.

It depends on what your criteria is. The $50,000 Mac Pro wouldn't be able to compete with a $5,000 Windows computer if you need an NVIDIA graphics card for CUDA workflow.

My M1 Ultra Mac Studio is better than my 13900k/4090 config in some workflows.

But this is why I have 7 computers in my workflow all with configs that maximize their performance for the task they are assigned.

 

[quarkysg]

That's not the only way to do it, and Intel makes it work. Their Xeon Max processors have 64GB of fast HBM2e that can be used exclusively as only fast RAM, as cache for off-package DDR5 memory which is transparent to software, or in addition to the external memory in which case the software must distribute between faster and slower RAM.

Intel Xeon Max 9480/9468 Show Significant Uplift In HPC & AI Workloads With HBM2e Review - Phoronix

www.phoronix.com

Technically it is possible to build all the circuitry to support such a deployment.

I guess it would be easier for the Xeon, since it probably have a single memory controller controlling the HBM and DDR5 DRAM. The Mx Ultra have 8 memory controllers each managing 128-bits of DRAM across two Mx Max dies via UltraFusion. Probably a lot harder for Apple to sync the memory access across all the memory controllers.

Using a 64GB cache for a 4K page table is also a challenge for the memory controller I would think. Quick calculation shows that the memory controller needs about 16m cache index entries to track 64GB of cache.

I wonder tho. how the real world performance of the Xeon would be like in cache mode when the dataset to be processed is a lot more than 64GB where random access to the dataset is required. It would have trashed like mad.

 

[urbanman2004]

I've been seeing a lot of negativity about the Mac Pro 2023 being "useless".

Majority of people who need the 2023 MP do not post on forums here or anywhere.

These are production companies that have PCIe cards that need to be put into the new Mac Pro. Music, Video, Etc. Colorists, Music producers, editors, and so on.

A lot of these production companies "lease" Macs and they swap it every 12-36 months. They have endless supply of money so the $3,000 that's extra on top of a similar performing Mac Studio M2 is not a big deal for them.

To me, being in this field, $3,000 is better than the $10,000+ that the 2019 Mac Pro cost.

Now is this a niche market? You bet it is. Are you part of it? Most likely not. Are you even aware of this field and how it fully functions? Nope.

To me, this is a good way of Apple to get rid of Intel overall. They are going to go to 3nm with the M3 and most likely we will see PCIe 5.0, increased RAM/GPU performance, and possibly dGPU support. The Mac Pro 2019 case has A LOT of head room for a higher clocked SoC, so don't be surprised if they do a M3 Extreme or some new line just for the Mac Pro down the line. They are not trying to kill the Mac Pro, they understand that certain conditions require a desktop/PCIe expansion slots.

So take a chill pill.

MKBHD just posted a video today a/b how the Mac Pro is useful based on what you said:

 

[Smartuser]

#2 yeah I think that is part of it too. Like what I said about shareholders. Putting more money in to Vision Pro because "we need a new product line" instead of you know being good to your consumer base. I have a large issue with public companies for this very reason. You must do EVERYTHING possible to maximize your profits. It is actually a requirement if you are a public company. Apple cannot get away from that. So to maximize profits, they focus more on iPhones and try to get in a new market. Which LOL I don't believe AR/VR is the end all be all market everyone else thinks it is. It hasn't taken off enough for YEARS.

Unfortunately, no matter what Apple could do I think GPUs are lost though which is unfortunate. But hopefully an Extreme SOC will make things a bit better.

Surely a company like Apple can only focus on one thing, and adding new products is a zero-sum game in that each new product automatically takes away from existing ones.

 

[iPadified]

There are two basic customers:
Those who configure, use and dispose in a three years cycle and those who buy what they need at the moment and upgrade along the way. I do wonder if larger companies are mainly in the former camp (SDAVE describe it rather well) and one man/woman bands in the latter. Apple clearly do not address the upgraders.

Apple has a very narrow focus on video and music as their "high end market" and has had that for a very long time. They basically missed the boat on anything GPU heavy such as cutting edge 3D (rendering) and ML/AI. Good or bad choice? Apple seem to have found niche and seem to survive rather well without supporting these high end applications.

The 2023 Mac Pro is boring and useful for some, but it seems like something they could have released with the M1 based Mac Studio. I think something did not pan out as expected.

 

[Allen_Wentz]

I understand your point and respectfully disagree 100%.

I have owned ever Mac tower since the G4 and am very disappointed with this iteration. Apple took away functionality yet raised the base-line price.

Yes, the top of the line 2023 MacPro is far cheaper than even a mid-grade 2019, but that’s not the big issue. The issue is what to you get for it? A Studio with PCIe? That’s not what we want. We want a Mac Pro in the truest sense of the name. We want:

1. A customizable machine that can be outfitted based on individual needs. You can’t do that without ram and GPU options.

2. A machine that has a universally better / faster chipset than available anywhere else. The MacPro was once advertised as the fastest computer on the market. Now it is just the fastest MacPro.

3. Prior to 2019, the MacPro was reasonably priced as a high end Mac. Now it is priced to push people away from it.

4. There are other issues too: the displays cannot be connected to a PCIe card. So no matter how you configure this, your stuck using thunderbolt for displays.

There are benefits to the Mac Pro such as two additional thunderbolt ports and the ability to add USB 3 / 4 data ports via PCIe. And, unlike the 2019, each thunderbolt port has its own controller, so, in essence, while you have a few less thunderbolt ports than in 2019, your getting nearly double the controllers.

Also, the media encoder, Afterburner, is a separate section of the SOC and therefore doesn’t use (much) processing power allowing for significant multi-tasking.
My take is that Apple was taking crap for the delays and came out with an M2 Mac Pro to satisfy that. I hope that the M3 MacPro will fix, or at least address, the very understandable criticism.

I liked on your comment, but #4 confuses me. Why care if the displays cannot be connected to a PCIe card when there are plenty of TB ports?

 

[Ethosik]

Surely a company like Apple can only focus on one thing, and adding new products is a zero-sum game in that each new product automatically takes away from existing ones.

Apparently not. Whenever you make iPhones incredible, Mac suffers (2016-2020). Whenever they make Macs great (M1 era), iPhones haven't been too great. And rumors are people got pulled away to work on Vision Pro.

 

[Smartuser]

Apparently not. Whenever you make iPhones incredible, Mac suffers (2016-2020). Whenever they make Macs great (M1 era), iPhones haven't been too great. And rumors are people got pulled away to work on Vision Pro.

Baseless.

 

[MayaUser]

since M3 ultra is probably 2 years away...there are still 50% chances that Mac Pro will add an M2 extreme down the path...the space is there, the cooling is defiantly there, Mac pro is not Apples profit margins so there is a chance

 

[rm5]

I agree with @jimmy_john. In fact, I was just talking to someone who does REALLY heavy 3D work, and they have a ridiculous AMD EPYC system with 2x NVIDIA A6000s. People who do that kind of work and need that kind of power know they need it, and thus, they go for other options (Windows).

Here's the thing I want to reiterate—if the company (or individual) wants to get 2023 Mac Pros, they can do that. idrk why people are saying, "stop getting these machines, they're useless." People know what's best for them, so they're fully entitled to buy whatever it is that will suit THEIR needs. No need to judge people's workflow or choices regarding what machine they buy...

 

[mode11]

Baseless.

The Accidental Tech Podcast have discussed this before. Apple's engineering teams are smaller than one might assume given Apple's size and wealth. Whilst they could afford to have huge teams, they prefer to keep things more manageable and prioritise quality personnel over quantity. So it's entirely possible that if it's 'all hands to the pump' to e.g. get the Vision Pro ready, other areas could take a back seat.

 

[mode11]

Apple has a very narrow focus on video and music as their "high end market" and has had that for a very long time. They basically missed the boat on anything GPU heavy such as cutting edge 3D (rendering) and ML/AI. Good or bad choice? Apple seem to have found niche and seem to survive rather well without supporting these high end applications.

Part of the issue is the refusal to allow Nvidia into Macs. I suspect this is to keep CUDA out of macOS. If they allowed it on the platform, many cross-platform pro apps would default to using it, as it would be less effort than re-writing them for Metal. Then Nvidia permanently have a seat at the table and have leverage over Apple - if Apple ever released a pro machine without an Nvidia GPU, it would bomb.

The other part is that Apple likes slim laptops and AIOs, which preclude big honking GPUs. The Mac Pro is the only machine they sell that could possibly take them, and that's too small a market on its own. This is the problem with Apple's long-term refusal to make a machine equivalent to a typical desktop PC (i7 / i9 tower).

So if your machines can't have big GPUs, where does that leave you? Audio, video, graphic design and so on. Also, bear in mind that Apple's profitability mostly comes from the iPhone, iPad, services etc. The Mac platform pulls in less cash than AirPods, so whatever choices they make with it doesn't affect their survivability much.

The 2023 Mac Pro is boring and useful for some, but it seems like something they could have released with the M1 based Mac Studio. I think something did not pan out as expected.

It would seem so. But perhaps they just focussed on their mainstream models first, before getting to the Mac Pro. They only have so many resources, and the 2019 Pro was still relatively new. Perhaps they wanted to get the Studio established in its own right? Perhaps the M1 Ultra wasn't a sufficient step up from the 2019 MP, so they waited for the M2 (where it seems to have found its stride)? Perhaps 128GB was less than the typical load-out of a Mac Pro, but 192GB would be enough for most customers?

We'll find out in the long run if the Extreme is something Apple's interested in. Various people have pointed out that although it makes intuitive sense to the layman, it's not obvious how it could actually be made. The Max only has one UF connector, and to access any other die in single hop (to keep the latency low), it would need three. The edges of the Max are already filled with memory interfaces and so on.

 

[NEPOBABY]

The anger in these threads all boil down to Apple's use of SoC in the Mac Pro because they want to show off 'UMA'.

But UMA can be done without an SoC. Silicon Graphics did it in their Wintel and MIPs systems over 20 years ago.

In their Wintel based UMA system if a user installed a third party GPU then the cards memory would be used instead of system RAM, and the memory was upgradable!

The Mac Pro shouldn't have been SoC based. It could have done UMA while still allowing upgrades by using proprietary buses and interfaces.

 

[gpat]

"I work in an admittedly niche field, and since this new computer works better for my chosen field than the previous one, everyone else should sit calmly while this new computer addresses only my needs. I also have the audacity to tell everyone to chill solely based on the fact that this new computer directly addresses my needs and loses functionality the previous computer had."

The fun thing is that this was also Apple of doing things in the 90's before going bankrupt.
The only difference is that now they have a multi-billion-dollar iToy business to keep them afloat while they squeeze niche professionals for every last drop of blood they can, and they do that just as a side project instead of their main business.

 

[quarkysg]

The fun thing is that this was also Apple of doing things in the 90's before going bankrupt.
The only difference is that now they have a multi-billion-dollar iToy business to keep them afloat while they squeeze niche professionals for every last drop of blood they can, and they do that just as a side project instead of their main business.

If you think Apple today is equivalent to Apple then, I would say you don’t understand Apple at all.

 

[gpat]

If you think Apple today is equivalent to Apple then, I would say you don’t understand Apple at all.

Actually I've stated the exact opposite.
Apple can afford to put out this kind of product at this kind of price, because they care little enough about it, and they didn't have the same luxury back then.

Go back 30 years and the Macintosh Quadra is just as invaluable as the Mac Pro is today for its designed niche.
Also still just as overpriced.

 

[0339327]

I liked on your comment, but #4 confuses me. Why care if the displays cannot be connected to a PCIe card when there are plenty of TB ports?

I have lots of TB connectivity and 4 monitors (3 for me and one for clients).

My current setup uses 9 TB.

 

[Smartuser]

The Accidental Tech Podcast have discussed this before. Apple's engineering teams are smaller than one might assume given Apple's size and wealth. Whilst they could afford to have huge teams, they prefer to keep things more manageable and prioritise quality personnel over quantity. So it's entirely possible that if it's 'all hands to the pump' to e.g. get the Vision Pro ready, other areas could take a back seat.

I've been listening to ATP since about 2015 (with breaks of a few months after some of Marco's most awful and off-putting hour-long rants(*), which haven't occurred on that level for the last few years), and while it's possible they said what you mention, it doesn't mean what you think it means. Using relatively small teams does not mean they can't have several good small teams. The qualifications needed on the Vision Pro are not the same as the iPhone team (as an example), so you can't just pull away all the good people from one product and move them to any other. Many of these projects also take a lot more years than they seem to from the outside (the Vision Pro apparently started in 2015), which puts a further dent in that theory.

(*) I fondly remember his one hour long rant about the 2015 MacBook keyboard, which I had no problem with whatsoever. The butterfly keyboard gave all podcasters endless filler for every single episode, they were probably secretly grateful for it.

 

[mode11]

I haven't listened to ATP much in the last few years; I just return for the WWDC episodes these days. It's possible I heard this elsewhere, but ATP seemed the most likely source. I'm sure certain products are very specialised (the VP in particular), but with every product running ASi and being based on the same macOS-y underpinnings, I would have thought people could move between projects reasonably easily. But I'm not a developer.

 

[Ethosik]

Baseless.

No it’s not. iPhone 14 was the worst launch of any iPhone period. M1 Macs are the best macs have been in a very long time. macOS and iOS are getting buggier because people got pulled off to work on vision pro from a report months ago.

 

[Ethosik]

I agree with @jimmy_john. In fact, I was just talking to someone who does REALLY heavy 3D work, and they have a ridiculous AMD EPYC system with 2x NVIDIA A6000s. People who do that kind of work and need that kind of power know they need it, and thus, they go for other options (Windows).

Here's the thing I want to reiterate—if the company (or individual) wants to get 2023 Mac Pros, they can do that. idrk why people are saying, "stop getting these machines, they're useless." People know what's best for them, so they're fully entitled to buy whatever it is that will suit THEIR needs. No need to judge people's workflow or choices regarding what machine they buy...

Agreed. And no matter what Apple did or what AMD comes out with, to some nothing beats NVIDIA due to their proprietary CUDA.

 

[Basic75]

but with every product running ASi and being based on the same macOS-y underpinnings

It doesn't need the same processor architecture to use the same basic operating system!

The first iPhone was presented as "running Mac OS X" because iPhone OS 1.0 was based on Darwin.

Just like Mac OS X.

Even though the iPhone was ARM and the Mac was Intel at the time.

And Mac OS X and Darwin still supported PowerPC Macs in 2007.

 

[Smartuser]

No it’s not. iPhone 14 was the worst launch of any iPhone period. M1 Macs are the best macs have been in a very long time. macOS and iOS are getting buggier because people got pulled off to work on vision pro from a report months ago.

No, it wasn’t, period.

 

[DrEGPU]

The anger in these threads all boil down to Apple's use of SoC in the Mac Pro because they want to show off 'UMA'.

But UMA can be done without an SoC. Silicon Graphics did it in their Wintel and MIPs systems over 20 years ago.

In their Wintel based UMA system if a user installed a third party GPU then the cards memory would be used instead of system RAM, and the memory was upgradable!

The Mac Pro shouldn't have been SoC based. It could have done UMA while still allowing upgrades by using proprietary buses and interfaces.

I think the anger could be further distilled into paying more for fewer features. Yes, it’s faster with ProRes/FCP than the 2019 MP, but that’s at the expense of more ram and powerful GPU’s.

 

[Ethosik]

No, it wasn’t, period.

Yes. The camera flaw was the worst issue imaginable. There were other bugs. It cause a major uproar in these forums because of “WFH quality”. macOS had some big bugs too.

Heck the recent bug with the camera adapter was a big issue for me.

iPhone and iOS bugs were enough where Craig responded to a question on quality.