No proofs, no thanks. If you ever tried Blender or other 3D software, it's way slower than you can imagine.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Mac Pro 2023 = Another Mac Pro 2013 crisis and failure.
- Thread starter sunny5
- Start date
- Sort by reaction score
We actually have proofs:
Apple M2 Max 38-Core GPU - Benchmarks and Specs
Specifications and benchmarks of the Apple M2 Max 38-Core GPU GPU.
Benchmark results doesn't really reflect actual uses with software. Even Apple themselves used mediocre benchmark software that M1 Ultra is as good as RTX 3090. In fact, they're aren'y many GPU intensive software on macOS so it's totally pointless since Apple GPU's performance lack so much compared to AMD and Nvidia in real life.We actually have proofs:
Apple M2 Max 38-Core GPU - Benchmarks and Specs
Specifications and benchmarks of the Apple M2 Max 38-Core GPU GPU.www.notebookcheck.net
GFXBench is a great example as M1 Ultra is close to RTX 3090 but did it really close to it with actual software or game? No.
Maybe start demanding better optimization of software, if software that is optimized runs competitively against everything that is current gen tech on this hardware?
P.S. "The more the data shows how wrong I am, the worse for the data."
Who would even want that as Mac Pro 2023 came out in a terrible way? 3D/AI software aren't interested in Mac because of poor GPU performance for a long time and Apple Silicon has finally killed it.
It appears you want. Just because you use rubbish software that doesn't make Mac Pro a failure.
Hahahahaha you first say there is no proof it is that fast at synthetics and then move the goalposts when the proof is provided.
Next:
You’re simply wrong to claim that in Blender it is slower than others
Here is Open Data Blender results:
Open Data
The M2 Max has a score around 2000, if we double that for the M2 Ultra (assuming M2 Ultra scaling matches M2 Pro -> M2 Max scaling) that is 4000, which is right in line with the 3090 in Compute capability.
Apple’s GPU cores are very very fast for compute - the fact that Games aren’t well optimized doesn’t change this fact.
I give you NVIDIAs SoC like design:
Grace Hopper
NVIDIA is moving away from discrete CPUs and GPUs and is moving towards ever tighter integration just like Apple. Grace Hopper still has them running separate physical memory but in their press briefs they talk about how tighter coupling is required moving forward.
Intel is also beginning to move this direction:
Falcon Shores
I haven't asserted that the Mac Studio has a thermal problem (though it apparently has a noise issue with the cooling system). I don't believe you have a Mac Studio with an M2 Ultra though because Apple doesn't start shipping it until next week.
Memory and storage. Enough memory to run the HPC workloads you are developing locally. Enough fast storage to keep to all the intermediate files around, allowing you to resume working quickly. Memory in particular is a less scarce resource in the HPC world than in the Apple world, and you can often make things faster with less effort by assuming that you have enough RAM.
Not praising it. It sucks. But the reality is there is a tiny tiny tiny market for HEDT, and the market for HEDT on mac is a tiny tiny fraction of that tiny tiny tiny market. There is no financial or business case to make a great mac pro. There's honestly a bit of a case that the studio should have been a bit bigger and had a 2-4 internal PCIe slots. But beyond that, even AMD doesn't bother to release HEDT parts regularly anymore and intel just put out their first major competitive update in years. The market between the MxUltra/x950x/xx900k and the datacenter/server is tiny and not worth investing in. if 24 cores of M2, 16 of Zen 4, or 24 of Raptor Lake isn't enough for you, chances are you're buying a server or cloud computing, not a higher end workstation
Everyone is using SoC. Phones, most laptops, all game consoles, every camera, etc. Even at the high end, look at every development. Intel has Ponte Vecchio with up to 47 tiles glommed together on one socket. AMD has been building more and more complex chiplet and stacked desgins for over 5 years. HBM or super fast an wide on package LPDDR massively faster ands wider than DDR or even GDDR.
There was a time we built computers out of vaccum tubes, and for the last 70 years the march has been towards more and more tighter integrtation of all hardware components because it allows for faster and faster speeds, lower latencies, better signal integrity, etc. The only advantage discrete componenets have is packaging and thermal.
As 100+ BN transistors can now be packed onto one SoC and be managed by a 200W thermal solution, the end of discrete componets is nigh. A large, properly powered and cooled AMD SoC with 16 zen 4 cores, 96 RDNA 3 compute units, and 192GB of shared GDDR6x or LPDDR5x on a 384 or 512 bit bus would run circles around a 7900x + 7900xtx system. So much overhead would be gone, so much wasted time moving data around, etc. Not understanding that a SoC is better is like not seeing the advantage a transistor had over a vacuum tube.
Apple's fault. Intentionally though, and they see it as a good thing. Because a few guys at Disney don't pay the bills to develop hardware and software that will be used by hundreds or thousands of people when those resources could go to solutions like a better A series chip or iOS apps that sell in the 100s of millions per year.
Workstations don't exist anymore. Threadripper hasn't had a new release in years and the TR 7000 series keeps slipping. Intel only put out Sapphire rapids because they are desperate to find any niche they beat AMD in, and they can beat 4 year old TR3000. Again, there are not enough people who want/need more than the MxUltra/x950X/xx900k but don't use server/datacenter hardware to make entering that market worthwhile. It sucks to be one of the people still stuck in the middle, who maybe could have used it. But your use case is extremely rare, and the cost of developing a new chip is extremely high. As a HEDT platform, the MacPro is DOA. But so is the whole HEDT market.
Not excuses. Facts. Apple knows they will sell thousands of MacPros, while they sell millions of MacBook Pros, 10s of millions of macbooks and ipads, and hundreds of millions of iPhones. They must recoup the RnD costs of the platform to justify making it. That means each buyer is covering many orders of magnitude more RnD compared to even a MacBook Pro.
The MacPro is at the pointest end of Apple's most Niche product line. Stats I can find lump " All-in-one, blade, mobile, rack, and traditional desktop workstation" together with no further detail, so no doubt include a lot of consumer and server hardware as well, but they state that in 2022 only 7.7mm "Workstations" shipped globally. If even 25% of that was actual HEDT towers, that's under 2mm global units. Assuming ~8% Mac market share, that's ~150k potential MacPro customers. I feel this is a very generous estimate for total MacPro customers.
Even if Apple expects to sell those 150k MacPro towers, at an average transaction price of 9000, their 40% target margin leaves ~$540mm in income from the mac pro line. That has to cover development of the hardware, marketing, factory costs, mananging the supply chain for the many unique components, special software development for the only PCIe equipped AS mac, etc.
Meanwhile, let's assume just 10% of Mac Sales are MBP/Studio (so using dies above the M1/M2), so about 3mm units in 2022. At an average of probably $3500 per unit, that leaves about $1500 in gross margin per unit or $4.5 BILLION to cover all development and other costs plus profit. Apple would need to sell well over 1 million $9000 MacPros per year for it to be as profitable as the MacBook Pro line likely is. That would mean apple would need the vast majority of the world's HEDT market to make the MacPro even as profitable as the MacBook Pro/Mac Studio, let alone something like the iPhone.
The market to take MacPro seriously just does not exist. Apple threw something out for the die hards, priced it so even selling 10k a month would cover costs, and will keep doing this until the category dries up completeltey. i don't like that. But pretending the dozens of MacPro customers are a market force apple cares about is delusional.
If by "Computers" you mean ATX towers. A tiny fraction of the global computer market. Even if you're generous and don't count phones and tablets as computers, "Desktops" are ~20% of the PC market, and desktops with high end Power supplies and x16 PCIe slots needed for GPUs are less than that, maybe 10% of the PC market. The reality is 9 out of 10 computers and more like 49 out of 50 computing devices can't use dedicated GPUs or any PCIe slots. The use case and world you live in is a tiny, dying niche. I hate that. I love a good tower. I love filling in PCIe slots. I love waking up an old system with a new GPU or even CPU, but we're a dying breed and the hardware we use or love is dying with us.
nVidia sells to datacenters, not workstations. They are a hype company pivoting their offerings towards whatever buzzword has piqued the anxiety uninformed FOMO executives and investors. "Buy our accelerators for Crypto! Wait, we meant for NFTs! Oh, no, it's 'AI' that's gonna change the world, powered by nVidia, of course!" Quardo is dead. Titan is dead. GeForce is dying to SoCs. GPUs or compute accelerators outside of datacenter are a very rare, niche product.Mac Pro 2019 prove your theory wrong. When Mac Pro 2013 was a huge failure, they admitted it's their fault to ditch the pro market. If they didnt care, how come they made a better workstation in 2019? Why not keep using Mac Pro 2013? Loosing the pro market is making a bad decision again just like they did from 2013 to 2019. All you are saying is that workstation is not profitable and yet, many companies such as Nvidia proves you wrong. Now, do you even wish FCPX editors to ditch Mac system?
Apple "messed up" (i still argue it was intentional, they wanted to kill off the not very profitable pro desktop end of their offering) in 2013 because the MacPro did not have PCIe, not becasue it didn't have GPU support. That's what they admitted. They realized that the audio pro and tiny boutque video shop might still use a mac from time to time and built a 2019 and 2023 mac pro that allows those users to stay on mac if they want to pay for the privilege. The truth is, a MacBook Pro or Ultra Studio can handle multicam 8k ProRes editing live, and any workflow beyond that in the AV space is being done by server racks running custom Linux. As for Science, especially Data Science, Apple prioved they were willing and ready to shun those customers when they abandoned nVidia over a decade ago.
Apple picks markets to be in they feel can be most profitable and fit their core competices the most. 20 years ago one of those was PC towers. And despite becoming a phone company that makes a few PCs, they know they have a few loyalists who love a big tower, so they have a tower out there again. They tried to scare you off in 2013. You didn't listen. So now they will make a tower as long as they can make a little money and mostly ignore it.
It's sad and lame, but the era of the tower is over. Like VCRs, CRTs, Camcorders, Firewire, consumer SCSI, and so many other once great products. Time marches on and tech improves. Every time it does, someone gets left behind at the expense of 20 others being carried forward. The blacksmith/horseshoer lost our to the automobile, but for every blacksmith out of work, dozens of people could travel more safely cheaply and quikcly than ever before. The PCIe/GPU die hard is losing out in this transition, but for each of us who lose the ability to add USB 2.0 to a 90's machine, or an RDNA 2 GPU to one from 2009, dozens more people have access to better, faster, and cheaper computing solutions than ever before.
In short, it's not about you. The MacPro mutating on the way to an all but certain death sucks for you, but it's good for Apple and 95% of consumers that development has shifted from ATX towers to super fast SoCs.
I'm not justifying anything. Just laying out the facts behind the decision. Apple is doing better than ever. Apple exists to drive wealth to their stockholders. Not make weridos like us who like tower PCs in the 2020s happy. If they could make billions on a cheap tower, they'd sell one. But there's no way to make iPhone or services products marketing a tower to a couple hundred thousand people worldwide.
The workstation/server/data center markets are basically the same. They all use similar hardware, configured in slightly different ways. (There are also workstations using a mix of consumer and workstation/server/data center hardware.)
AMD launched the Threadripper 3000 series in 2019-2020 and the 5000 series in 2022. The 7000 series is expected in 2023-2024, which doesn't sound that late. The 5000 series only had Threadripper PRO chips, because it didn't make sense for AMD to sell separate consumer, kind-of-workstation, and workstation chips. At least at that time.
The Mac Pro is in trouble, because it's Apple's highest-end product, achieved by scaling up their consumer hardware. For everyone else, workstations are the low end of the market, with products that are scaled-down versions of mainstream hardware. That's why Intel/AMD workstations still have a future, and we are probably going to see some ARM-powered competitors.
MacPro is not in trouble. It's a product that has no meaningful market that apple puts out as a charity to a few loud die hard fans. If the MacPro shipped with full GPU support, a line of Apple GPUs that wiped the floor with AMD and nVidia, 128 core configurations, 1TB+ of 1TB/s+ RAM, and SATA+NVMe for storage, all for the same $7000 price, the number sold wouldn't change meaningfully. I'd put my life savings on it. On a global computer hardware scale, workstation towers are dead.
I'd get the Pro over the Studio because it has space to store snacks:
It's likely more than a "drop in the bucket" (and the atom thing is just hyperbole). Just for fun:
One drop in a 5-gallon bucket is ~1/400,000*, and Apple's annual revenue is ~$400B, so a "drop in a bucket" out of Apple's total revenue would be $1M. Assuming an average sales price of $10,000, that would be 100 Mac Pros/year (which would be, say, an average of one per state in the US plus another 50 worldwide). Certainly they sell more than that (not necessarily including this past year), so it's likely more than just a "drop in the bucket".
The 6% is an interesting statistic. What's your source?
20 drops/mL x 3785 mL/gallon x 5 gallons/bucket = 378,500 drops/bucket
Last edited:
If Apple started selling workstations like that, they would take over the server/data center market. Much like cheap x86 hardware took it over decades ago.
To Apple, workstations are a separate market segment that requires separate R&D spending. To everyone else, workstations are just an easy way to make more money from the R&D spending they had to do anyway. The Mac Pro is a niche product, because Apple is deliberately ignoring the mainstream market for that kind of hardware.
You're right, I don't have an M2 ultra, nor will I. The noise issue is also real for the M1 based Studios, and no way will I ever chance that again without a big redesign. I know most don't have the problem, but I'm "lucky" that way. I just traded my Studio Max in for a Mini Pro. The Mini will have to do.
As for having a thermal problem, you seemed to be falling into that side of the argument questioning those that said it doesn't have a problem. Remember you're talking about basically a mobile chip in a desktop with desktop level cooling.
It's just you dude. There is no workstation market. It doesn't exist on a meaningful scale. Apple wants to enter consumer facing market niches that have 100s of millions of customers. They've never been able or interested in selling to corporate customers like server hardware would require. As stated in my longer post, a generous estimate puts the theoretical MacPro market at 150k units per year. This market is not something apple wants to be in, even if that bums out a few of us nerds on this forum.
It's a Mac Studio with more ports and 7 times less RAM capacity than the previous Intel Mac Pro. It's a joke.
There is no workstation market, because it makes no sense to talk about workstations alone. The real market segment has been workstations / servers / data center for a long time. Apple is still trying to stay in that market, perhaps for prestige or to keep their future options open. (Maybe they believe they will need that kind of hardware if Apple Car ever becomes a thing.) It's not a significant market for them, but only because they are choosing to keep it that way. For Intel / AMD / Nvidia, workstations remain significant, because they offer an easy way to make more money without any major investments.
One observation from a dual-platform guy. If you go into a computer store (the kind that sells parts to build a Windows box) today and look at what's happening in the market (build-it-yourself types) that should value modularity most, what you'll see is this:
- most enthusiast/gamers cases no longer have any drive bays. No 5.25" bays, no 3.5" bays. All your internal storage is NVMe on the motherboard. It's not just that people don't seem to want any internal devices currently that would require those drive bays, but they seem confident that nothing will come along during the life of the case that requires such a bay
- enthusiast motherboards are seriously losing the PCI-E slots. The lanes are being reallocated to NVMe storage. The newest innovation seems to be placing your GPU sideways... which happens to block off all the space for PCI-E cards on an ATX motherboard.
As I said to a dude at the store, at least with a Mac studio, you get a built-in card reader. Most of the enthusiast cases don't give you a place to put one anymore. The idea that you want internal expansion because some new thing of some sort will come along and you will want to add it to your modular Windows desktop seems... gone. A quaint relic of earlier times.
The other observation I would make is that non-gamer non-factory-built desktop boxes are gone. 15-20 years ago, in the days of affordable Power Macs, you could walk into those stores and get them to build you a boring desktop machine, complete with a boring beige or black case. There are almost no boring beige or black cases anymore. No boring, lower-end motherboards. It's all about the gamery stuff with the RGB, the AIO/liquid cooling, etc. In Windowsland, that and some business machines seem to be the main market left for desktops. Laptops, also with ever less expansion (e.g. Lenovo ThinkPads now have soldered RAM), are where things have gone.
This makes me think Apple has been on to something in the days since they gave up on a mainstreamish desktop with internal expansion. Other than some very pro cards for some very pro users (which they gambled were no longer needed in the 2013 Mac Pro, only to reverse course in 2019), modularity and internal expansion and desktop machines just... seem on the way out. Same with removable storage - as much as it seemed crazy when Apple dropped optical drives and refused to support blu-ray, it seems like removable media is dead in Windowsland too - external HDDs offer a cost-per-GB that just crushes blu-ray.
The modern computer seems to be embracing the paradigm of the retina MacBook Pro or the Mac studio - something that remains in the same configuration it was when you bought it, that has built-in storage, maybe a teeny bit of USB and good, likely wireless, networking connectivity, and that's it. Something that is overwhelmingly and increasingly likely to be a laptop.
(Do I like this trend? No... but I'm a guy who would probably still have a floppy drive in his Windows desktop if they hadn't removed the floppy controller from the motherboards.)
- most enthusiast/gamers cases no longer have any drive bays. No 5.25" bays, no 3.5" bays. All your internal storage is NVMe on the motherboard. It's not just that people don't seem to want any internal devices currently that would require those drive bays, but they seem confident that nothing will come along during the life of the case that requires such a bay
- enthusiast motherboards are seriously losing the PCI-E slots. The lanes are being reallocated to NVMe storage. The newest innovation seems to be placing your GPU sideways... which happens to block off all the space for PCI-E cards on an ATX motherboard.
As I said to a dude at the store, at least with a Mac studio, you get a built-in card reader. Most of the enthusiast cases don't give you a place to put one anymore. The idea that you want internal expansion because some new thing of some sort will come along and you will want to add it to your modular Windows desktop seems... gone. A quaint relic of earlier times.
The other observation I would make is that non-gamer non-factory-built desktop boxes are gone. 15-20 years ago, in the days of affordable Power Macs, you could walk into those stores and get them to build you a boring desktop machine, complete with a boring beige or black case. There are almost no boring beige or black cases anymore. No boring, lower-end motherboards. It's all about the gamery stuff with the RGB, the AIO/liquid cooling, etc. In Windowsland, that and some business machines seem to be the main market left for desktops. Laptops, also with ever less expansion (e.g. Lenovo ThinkPads now have soldered RAM), are where things have gone.
This makes me think Apple has been on to something in the days since they gave up on a mainstreamish desktop with internal expansion. Other than some very pro cards for some very pro users (which they gambled were no longer needed in the 2013 Mac Pro, only to reverse course in 2019), modularity and internal expansion and desktop machines just... seem on the way out. Same with removable storage - as much as it seemed crazy when Apple dropped optical drives and refused to support blu-ray, it seems like removable media is dead in Windowsland too - external HDDs offer a cost-per-GB that just crushes blu-ray.
The modern computer seems to be embracing the paradigm of the retina MacBook Pro or the Mac studio - something that remains in the same configuration it was when you bought it, that has built-in storage, maybe a teeny bit of USB and good, likely wireless, networking connectivity, and that's it. Something that is overwhelmingly and increasingly likely to be a laptop.
(Do I like this trend? No... but I'm a guy who would probably still have a floppy drive in his Windows desktop if they hadn't removed the floppy controller from the motherboards.)
You can’t compare the previous Intel having 28 core to this one say “it only has 24 cores”.
There is a big, big difference in those chips and cores.
They don’t directly correlate and compare to a 128 core chip that puts out so much heat that the system can’t manage either.
The 2013 lost out huge in 2 Key areas:
1. Expandability - they were really pushing to go toward Modular systems with Thunderbolt. This did not serve Pros well.
2. GPU - the dual GPU workflow didn’t materialize the way they thought it would. And Thunderbolt wasn’t a great way to expand.
Apple is poised well for the future.
They will hopefully introduce chips even more powerful than the Ultra, or expansion modules with more CPU and GPU cores.
But the expansion capability is there, and it’s very different than 2013 trash can MP.
There is a big, big difference in those chips and cores.
They don’t directly correlate and compare to a 128 core chip that puts out so much heat that the system can’t manage either.
The 2013 lost out huge in 2 Key areas:
1. Expandability - they were really pushing to go toward Modular systems with Thunderbolt. This did not serve Pros well.
2. GPU - the dual GPU workflow didn’t materialize the way they thought it would. And Thunderbolt wasn’t a great way to expand.
Apple is poised well for the future.
They will hopefully introduce chips even more powerful than the Ultra, or expansion modules with more CPU and GPU cores.
But the expansion capability is there, and it’s very different than 2013 trash can MP.
not really true when some of the leading software have ported over to apple silicone and metal
Wrong. Apple wants no part of the server/datacenter market. They've never been in it and they don't want to be. What makes you think Apple, the iPhone company, wants to build the skills, staff, products, and infrastructure to compete in a completely different business (large corporate B2B server sales). Would you assume your favorite donut shop was in the commercial flour milling business if they made a few extra bucks with overpriced novelty "make your own donut" kits?
More importantly, the server/datacenter market is... concentrating. Organizations that might have bought a couple Xserves or similar boxes from Dell or HP or whoever a decade ago are migrating a lot of workloads to solutions on public clouds. Even bigger businesses/governments/etc are moving a lot of things to things like Office 365. On-prem servers can't be a growth business these days.
And I suspect selling servers to AWS, Azure, and the other big players in public cloud is a perfectly good business, but that's a business that Apple isn't interested in and has no chance to compete in... and frankly, a business that likely doesn't exist as those guys probably design their own customized x64 (or now ARM) boxes and just have manufacturers compete for the privilege to build those boxes.