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.
Apple Silicon Mac Pro Said to Feature Same Design as 2019 Model, No User-Upgradable RAM
- Thread starter MacRumors
- Start date
- Sort by reaction score
So what I'm hearing is you don't do the kind of pro work that most people in the market for these machines do and are happy to proclaim this isn't a problem for you who likely doesn't even need this machine....
In other news, the people who are editing pro video and rendering special effects and would be the folks shelling out $10-$50K for a set up probably really do need more RAM than what Apple is going to be capable of offering.
You would think Apple could engineer a means of letting a user put additional RAM into a slot and MacOS would be able to use that ram in conjunction with the soldered ram on the SoC. You'd think..... sounds like they found a way to do it for graphics cards unless that is BS
On the advice of counsel, I will have no comment…
I guess the idea is if you shell out $10-50K you should shell it out to a maximised Pro model from the get go.
Side question if you are in the know. Do movie makers and tv series creators use the regular software available for everyone else like FCP, AVID, Resolve...etc? I always thought they had more specialized software.
Exactly what I was thinking reading this article: savvy businesses are going to buy windows.
And so Apple is only going to cater to the artistic types who don’t think to hard about the bad deal they’re getting.
And the ones who DO think about their purchases know full well and accept the higher price of Apple products and will buy the upgraded models… and live a good life.
Sad, but it honestly makes sense considering the new arm-based architecture. As long as we still have PCIe and the ability to internally add SSDs/HDDs, then we’re golden.
Because you can have "internal" PCIe slots so you don't have to have 100 cables and power adopters around your Mac Studio and pretend it to be "small" form factor. Also better cooling and quieter, the only noise-less fan of Macs is the ones in 2019 Mac Pros. I can have it running 100% for a month without any fan noise, but can't say the same with others.
Because it is cheaper. R&D costs $0.00.
Look at the M1 Mini, MBA , MPB 13" ..... exact same case/chassis.
The M1 Mini uses up what 1/2" - 3/4" of the internals. Doesn't matter it is cheap. Even more so if going to sell the Intel 2018 model side-by-side for over two years. One case, two systems ... two-birds-one-stone.
Apple might be shipping the assembly process over to Vietnam but the exact same case subcontractor in China can keep churn out the exactly the same thing on exactly the same jigs/tools are they have for the last 3 years. ( all the R&D for the case is highly likely paid for last year if not even farther back in the past. ). If they make enough of the other parts in Vietnam it gets a "made in Vietnam " label slappled on it. And Apple has spent the minimal amount possible to move the production out of China.
Unless Apple has stripped 100% of the Thunderbolt controllers out of their updated SoC, MPX is a solution in search of a problem. It is primarily there so that discrete TB controllers deployed at the far outer edges of the internal system get provisioned with the correct inputs. If the TB controllers are inside the SoC the need for that drops to about zero. The TB controller are deployed at about the complete opposite end of the internal layout; toward the middle not the outer edges.
The height of the Mac Pro 2019 largely driven the the slot width requirements for the 8 PCI-e slots. There are 4 doubles wides and 4 single wide PCI-e standard slots. The two MPX bays intersect with each of the first two double wides but the double wides themselves drive the width.
There are hundreds of more PCI-e cards that can go into a PCI-e slot than just a GPU one. Even if the GPUs are out there is certainly still usages for several single width slots ( networking , storage , etc. ). There are I/O cards that don't consume 300+ W but do have several connectors on the back that only fit over a double wide backplane.
There are non GPU card that can take > 75W power the standard bus provides so 6/8 pin aux power connectors can be used on non GPU cards too. Not a pervasive as GPUs, but not zero either.
AMD just proviewed their MI300. The normal version is 600W so not fitting onto an ordinary PCI-e card any time soon, but they could also produce MI305 that is scaled back to 12 CPUs and fewer GPU dies that runs in a 280W envelope. If the card runs its own self contained version of Linux there is no good reason to block that kind of "compute accelerator on a card" from being placed in a Mac Pro.
A DPU/IPU (data/infrastructure processor unit) cloud SAN/NAS card (e.g.,
Amazon Nitro
AWS Nitro v5 Ups the Cloud DPU Game Again
AWS Nitro v5 extends an already enormous lead in the DPU space with a new generation of card designed to deliver more speed and services
NVIDIA BlueField-3 400Gbps DPU Exposed
We found a NVIDIA BlueField-3 400Gbps DPU while checking out upcoming Supermicro servers at an undisclosed location
or
https://www.servethehome.com/marvel...dpu-in-the-wild-rivaling-2017-era-intel-xeon/
) If Macs are going to continue to be placed in Amazon/Azure/etc cloud services nodes then having those kinds of cards in a Mac would be useful ... although the Mac Pro is rather 'painful' to rack up in a space efficient fashion. But can just charge more. LOL.
Over time there are likely going to be specific workload accelerators that are not retail, off-the-shelf, commodity gaming GPU cards. The whole world doesn't revolve around just those cards.
There could be a way that Apple opens the door for using those cards a "Compute GPGPU" cards. In that case not really directly competing with Apple GPU for display GPU work. But it is a way to throw another 250-300W at new chunk of TFLOPS capacity to chew through some data crunching workloads.
You really think so? I am not saying you're wrong. I just think it would be needlessly expensive and drawn-out to slowly kill of the Mac Pro by deliberately sabotaging it just to save face for a small minority of Apple's client base. It is my opinion that whoever is making the calls on the Mac Pro is really just that out-of-touch. 😆
No - it is worse than that. The first R2D2 (trashcan) Mac Pro was meant to be small compact and yet powerful ... but without an expandable path. The new 2023 Mac Pro (according to this article) is NOT SMALL, is not compact - the whole point of which is expansion ... and yet apparently does not offer the most basic element of expandability.
Apple has been patiently waiting (as have I) to take revenge on all the 6,1 haters! That's right, people, get ready for Trashcan 2.0! Speaking of, my 6,1 is working flawlessly and looks fabulous. Let the hate fly all you slot jockeys!
It is extremely unlikely that Apple is going to back to one, and only one, internal drive for the Mac Pro. Even if there is zero 3rd Party display GPU support shows up leaning too hard on Thunderbolt is a bad idea (Apple has already explicitly admitted that). Once in the > 16TB internal capacity range Apple has no solution even if want to pay the BTO pricing. A single SSD is only going to be so big.
"Thunderbolt Next" really isn't going to solve the problem. There are PCI-e v5 x4 SSD being shown at CES right now. Cobble 4 of those together and that will hobble "Thunderbolt Next" also.
The slots are not likely to disappear again. Even if Apple had to do something extremely cheesy like map "extra" TB controller to an internal discrete TB controller and back out to PCI-e v3 switch.
All they really need is something not optimized to be a MBP 14" SoC. Just a simple set of desktop chiplets with just a narrow one or two x16 PCI-e v4 bundle allocation. That could be used across the STudio version (where ignore the PCI-e x16 ) , Mac Pro , and any large screen performance iMac they wanted to do. Aggregate all of those together and probably have enough volume to support the work.
People may try to declare this "Trashcan 2.0" but if Apple uses the old MP 2019 chassis that says more about dubious descriptive commentary than insightful commentary. The current chassis looks nothing like the trashcan.
What is going to matter is how much 'backhaul' bandwidth is behind all the slots in the new system. If it is comically low then folks will come up with a new name. If it is about the same ... how throw rotten tomatoes at that and not at the original one is a bit of hypocrisy.
For the Pro market it’s designed for, those folks will be delighted. Mainly because Apple’s in direct contact with those few thousand that Apple invites to their campus to discuss their workflows. There are several million that do NOT have that direct contact with Apple that absolutely will be disappointed.
I seriously hope this is in the “flat Apple Watch” category of Apple scoop.
This would be a terrible product with almost zero flexibility. Take a quick look at Apple’s Refurbished Store online and you’ll see page after page of Mac Pro’s for sale. I’d imagine it would be many times worse with this glorified Mac Studio.
This would be a terrible product with almost zero flexibility. Take a quick look at Apple’s Refurbished Store online and you’ll see page after page of Mac Pro’s for sale. I’d imagine it would be many times worse with this glorified Mac Studio.
While it’s possible he’s simply wrong, it’s not that unusual for rumors to change as Apple can change their plans internally.
His older tumors may have been true at the time, but as things change within the company/market/supply chain their plans can change.
It could be 'off die' but it would still be inside of Apple's package. Amazon's Graviton 3 doesn't put the PCI-e or Memory controllers on the "compute" die.
https://www.nextplatform.com/2022/01/04/inside-amazons-graviton3-arm-server-processor/
AWS Graviton3 Hits GA with 3 Sockets Per Motherboard Designs
Amazon AWS Graviton3 C7g instances are now in GA and are built upon a 3 CPU per motherboard design which is very uniquewww.servethehome.com
AMD did a different disaggregation where they took the memory and L3 cache off and left the PCIe and displayPort output on the main die.
IMHO, Apple does need a better chiplet design strategy for desktop ( minimally the Studio , Mac Pro and perhaps a large screen iMac with performance ) that laptops probably won't want to use. A 'desktop' Max , Ultra (which only will ever be desktop) , and a > 2 compute chiplet "extreme".
Does Apple have a chiplet design strategy at all?!
This is eye opening.
“Chiplets” seems like a very smart way of being able to improve “blocks” of an SoC in-line with the most recent technology advancements, without waiting for a years-long complete redesign of the entire die.
I agree that Apple shouldn’t decouple the CPU, NPU, ML, GPU cores from the die unless — UNLESS — the performance and clock speeds of, say, GPU cores are being held back by the much slower clock speeds that CPU cores can only run at.
If Apple could design a GPU-only IC that could run at 10 GHz — and the required I/O memory bus/data bus is sufficiently fast, I say, Go for it! Especially if more and more General Purpose instructions can be performed on a GPU instead of the CPU. (Apple needs to strive much harder to find more and more GPGPU optimizations. Linux is way ahead of Apple in this pursuit.)
But I suspect Apple doesn’t want a situation where an Apple Silicon SoC design changes every six months, much to the confusion/frustration of developers and even Apple’s own OS/SDK engineering teams. It’s already the case that two years after the M1s release, MacOS software that claims it’s optimized to run on the M1 isn’t reeeeeeally as optimized to run on the M1 as it could be. (Including even Apple’s own core apps like Final Cut Pro.)
Frequent changes to Apple Silicon via regular “chiplet” improvements might present an ever ”moving target” that developers will be demotivated to code specifically for, knowing that a fundamental part of the architecture might change in 6 months.
That’s why the burden on Apple’s own OS software engineers should be so high. If hardware abstraction is strictly adhered to, Apple’s OS (and SDK) software engineers can make changes to the underlying OS — in-line with changes to the underlying Silicon — such that existing codebases can simply inherent Silicon and OS performance improvements automatically, without the need for developers to so much as recompile existing apps.
I do realize that this is how it works already; I’m calling for even greater dedication on Apple’s part. Rosetta 2, for example, was Apple’s “Moon Shot” that Microsoft can only DREAM of ever accomplishing.
(As I understand it, Apple Rosetta 2 engineers even found plenty of Intel instructions that needed no translation at all to run on ARM.)
Rosetta 2 might have led to a lot of Apple engineer burnout, but Apple needs to find a way to motivate engineers to be as devoted and dedicated to “impossible” feats like this again.
Apple isn't out of touch. The Mac Pro doesn't strategically matter for the Mac business. It is effectively a hobby product. Apple does what they feel like doing. It doesn't make a difference to the overall business line.
The whole time Apple was supposedly 'failing' at the Mac Pro 2013 the number of Mac sold went up. The overall Mac revenues went up. How a business makes substantially more money and sells millions more systems than they did over the previous 10 years and 'fails' is an extremely obtuse use of the adjective 'fail'.
The didn't fail financially. And that is what mainly matters.
Macstadium was buying pallets of new MP 2013 systems on the exact same week that the MP 2019 shipped. The extremely warped narrative her is largely "because Apple didn't sell a Mac Pro to me , they failed. ". No. You just didn't buy a Mac Pro. Apple may not have necessarily have needed your money. If they got money from someone else that pays for MP 2013 components and R&D also. Some folks liked the Mac Pro and others didn't. Failure is more a matter of whether the first group is big enough to pay the bills for making the system or not.
Apple has about 10% of general PC market and Windows has about 90%. Apple didn't sell as many Macs as they could have ... They failed. Really? Not. Apple doesn't have to sell more PC units than Dell to be successful.
Apple isn't out of touch as much as they have a different weighting on priorities. Building a container that can accept the most affordable commodity parts possible. I don't doubt they heard that. But I doubt they put a high priority on that. Apple's priority was getting better at buidling more integrated designs. The high growth segments of their overall business needed more folks with that skill. Apple went about building that with success. Pragmatically that put the Mac Pro at the end of the line for updates. So nothing happened for a long time.
Apple also saw in 2015-2016 that Intel would doing more goofy stuff and AMD was struggling to keep the lights on. As Intel got more stuck the performance was cranked up by just burning more power. Nvidia and AMD were in a suck power out of the wall contest too. Apple watched for while and then went to a design to which pulls the max power to code (well USA code) out of the wall while they spend tons more money to get off that max power consumption crazy train. Then cranked up the entry price 100%. That isn't "out of touch". They are deliberately leaving some folks behind. Extremely unlikley that they are thinking the exact same set of folks who bought those MP 2008-2012 systems are going to buy something 100% higher in price. They aimed for a subset (that they probably felt was going to give them better margins. It is not a strategic/critical product. All it has to do is make enough to pay for the work at some minimal return rate of investment. Do that and it is a 'winner'; gravy revenues on top for the core Mac business. )
Could have gone Zen 5 with TBs of RAM using EPYC processors or Threadripper Pro. Instead, you expect people to dump > $15k down at once or more with soldered on RAM. DOA.
Shouldn't be a problem as long as these machines come with 6TB or RAM.