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.
MP 7,1 Would the Mac Pro 2019 (7.1) capable to upgrade from Intel CPU to the new Apple Silicon CPU??
- Thread starter Kentrent
- Start date
- Sort by reaction score
I definitely see Apple making a new Mx motherboard that we would just replace the old one with.
WHEN HELL FREEZES OVER!
Wow - what a deal! You must be really happy with it. I just hope, much like every other Intel Mac, that Apple gives them proper support/updates/etc for as long as they would have in the past.
What happened to all those Intel Ice Lake talks that were going on a year ago?I definitely see Apple making a new Mx motherboard that we would just replace the old one with.
I also hope that Apple continues to support the 2019 Mac Pro for some time, but I'm not certain that they will. During the last transition, the Power Mac G5 got three years of support after it was discontinued. My guess, and this is purely a guess on my part, is that there will be two new x86 releases of macOS after Ventura, then an additional two years of security support, and that will be it. I've heard some folks thinking ten years, but I believe that is too optimistic.
From what I can tell, most of that talk has gone away. I'm sure Apple had another Xeon Mac Pro sitting in their labs, at some point, but Apple prototypes a lot of things that never materialize. They may not have released it for performance reasons, marketing reasons, or any other factor. I think they are going to announce the Apple Silicon Mac Pro as soon as possible, and the 7,1 will be the last of the Intel line.
Apple has never been afraid to cut loose support for old models with impunity. The transition from PowerPC to Intel was rapid, and Apple's judgement was swift and final. I expect the same to happen this time around. I don't say this with any joy, like I said I just bought a Mac Pro, but I think this is the most realistic outcome.
Totally agree. The 2019 mac pros are like the latest G5s: high consumption and performance not at the level of their intel competitors at that time.
The support will be short-lived, and mostly it will be a comparison to show how much better the Mac Pro silicon is in all aspects.
Another point which also seems almost certain to me: the Mac Pro silicon may be expensive but will be less expensive than the xeons.
The support will be short-lived, and mostly it will be a comparison to show how much better the Mac Pro silicon is in all aspects.
Another point which also seems almost certain to me: the Mac Pro silicon may be expensive but will be less expensive than the xeons.
Yes in those days I had a SE and put in a SE30 motherboardI had a performa 6300 .... I put in a 6400 motherboard and a POWERPC upgrade card plugged in the PDS (processor direct slot). Those were the days ... sigh
Although unlike the G5, the 2019 Mac Pro utterly demolishes Apple Silicon in the heavy lifting and GPU intensive tasks, for which it's intended.
The G5 in its day was outclassed in every way by consumer-grade Intel machines of the time.
The Apple Silicon Mac Pro is going to start with an entry price above the most expensive Mac Studio config, and there is no chance the price is going to be less than the current Mac Pro. None whatsoever.
You say that as if Apple released the 8,1 AS Mac Pro. So far Apple never released workstation class SoCs nor a proper Apple Sillicon workstation. I would not call the Mac Studio a workstation nor the M1 Ultra is a workstation chip(which is just two laptop SoCs 'glued' together)
Look I get its 'cool' to crap on Apple's latest tech but comparing workstation class parts to mobile laptop parts is funny to me.
If Apple intended the Mac Studio to be it's halo workstation it would have stopped selling the 2019 Mac Pro and a replacement to that machine would not be in the works.
The 2019 Mac Pro is outclassed in CPU pref by the M1 Ultra which are just 2 laptop SoCs put together. It's not even a proper dekstop workstation class chip.
Why would you assume that Apple will build a "workstation" class chip, rather than just kludging together more laptop components, and expecting workstation users to eat a generational pause in improvements?
oh they have to build a workstation/server class chip! How else do they even come close to the 1.5TB RAM limit in the current Mac Pro.
Apple better do what Nvidia did and provide a large pool of 1TB+ RAM. That's when you know it's a workstation/server chip. You don't get that RAM amount by gluing down laptop parts, you get there by building a workstation chip from the ground up.
If Apple don't do that then its time to avoid the 8,1.
What makes you think they'll do that, vs using influencer marketing to push a narrative that you don't need 1.5tb of ram, the vast majority of customers don't need 1.5tb of ram, and that unified memory and fast storage means virtual memory is fast enough?
Nvidia did it by allowing GPUs to access system RAM & networked storage directly - providing more external resources to the GPU is more or less the opposite path to that pursued by Apple.
If it's the same philosophical path as the 6,1 and Mac Studio, yes.
I mean the Nvidia server ARM chip which has 1TB of LPDRR5 RAM on die.
They could have stopped at the Mac Studio if that was case and said that Mac Pro was no more. Like you don't need PCIe slots, you don't need more than 20 cores. That's a dumb arugument if that really was Apple's reasoning for not providing the RAM capacity.
Why even make the Mac Pro if thats Apple's thought process?
Yep, it better have lots of display support, 1TB+ RAM support and at least 4 PCIe support for me even consider it.
Actually, it's LPDDR5X SDRAM; which Apple could also use, placing up to 1TB of RAM in the ASi Mac Pro...
It would be interesting to see where the economics of that end up - would they make every chip with 1TB of ram, and then use binning and software disabling to create chips with "less" memory. In which case, every chip eats the cost of manufacturing a 1TB version.
Or, would they manufacture a very, very small quantity of chips with 1TB.
Apple's MO has generally been to convince people to change they way they do things, so as to suit the products Apple wants to make.
The purpose of vaporware is to prevent customers from buying a competitors products, e.g. talking about the 2019 Mac Pro when the iMac Pro was still unreleased and had been set to replace the Mac Pro as the strategic direction for high end Macs.
If it doesn't, the Grubers of the world will be well compensated to convince you of why you no longer need those things any more.
Gruber recently got very mad at Apple's new System Settings app in macOS Ventura. I have never seen him so.
One other thing even if the hardware is good enough. Apple's software quality these days makes Microsoft's look like a beautiful functional artwork. iOS 16 - buggy, iPhone 14 Pro - buggy, iPadOS - buggy, macOS Ventura- VERY buggy.
Apple needs to fix their software. They are releasing Vista in every product OS category.
Gruber is allowed to get mad at things Apple can fix - System Settings is early developmental software, it can be changed, and updated. If you look at the new responsive System Settings app in Elementary:
https://blog.elementary.io/updates-for-august-2022/
what's wrong with System Settings is solvable, and I'm sure Apple's goal is to solve it. They're trying to dogfood an immature API in their own products, so they can convince 3rd party developers to use it. That's understandable.
What Gruber, and anyone wanting to keep access don't criticise, is The Plan.
Personally, I think the hardware is no less misguided than the software. Unnecessary notches on laptop screens because decorators are obsessed with having equal top and side bezels (the influencers really shot themselves in the foot complaining about the LG5k's "big forehead" on that one), killing off eGPU support after only one generation of Thunderbolt enabling it, supplanting TouchID with FaceID etc...
No disagreement, though I suspect it's largely impossible with the current corporate structure. If they get rid of the money guy, and put a product person in charge, a person whose goal for every product is how good that product is, regardless of larger company strategies, that might change. I think they're aiming towards a world in which software is continuously updated, and there's no such thing as a "version", and then eventually Apple as a utility, like power and water.
"Remember, buying a computer isn't something you should be afraid of. Even if you make a mistake and get the wrong one, it'll be obsolete so fast, it just won't matter."
Yeah I know but it irks me that Apple can't get the simple things right. Oh a translation layer between ARM and x86, Apple is the best industry there or making the best ARM chips best there too.
You know, I always thought of Apple as company that can do the hard things with perfection but Apple stumbles at the simple things.
Alan Dye.
I want Johny Srouji to become CEO. He one of the potential candidates to become the CEO of Intel.
Tim is lucky to have him around. Without him and his team Apple Sillicon wouldn't be as good as it is today.
we will laugh if the Apple Silicon Mac Pro is over 50000$ for his best configuration !😂🤣😂🤣😂🤣😂🤣😂🤣😂
That's exactly what they did with the Trashcan - so it wouldn't be unprecedented - and they stuck by the trashcan for years before relenting. They could be testing the reaction to and uptake of the Studio before committing to a Mac Pro.
So I wouldn't rule the "no Mac Pro" possibility out just yet - we only have one fairly ambiguous hint from Apple that there really is an Apple Silicon Mac Pro coming and no clue as to whether it's going to be a big box 'o' slots or a "Studio Extreme". I'd put a long shot on the "Mac Pro" just being the Studio Ultra innards in a 1U Rackmount form with a (probably third party) Thunderbolt-driven storage, and PCIe enclosures for those who need PCIe for specialist interfaces, AV cards etc.
They won't manufacture it at all unless they identify a market that needs & can afford it - probably something like Big Data/Machine Learning - and I bet systems won't be starting at 4-digit prices! I can't imagine it turning up in anything that could remotely be described as a "personal computer" - it's a beast for high-density computing in data centres and cloud computing providers, and the sort of specialist.
Whereas the current Mac Pro is still, really, a personal computer workstation (the W in Xeon W) - albeit a very high-end one - primarily aimed at media production housed with an existing buy-in to MacOS software. It's not particularly compelling if you don't have some sort of commitment to MacOS when there are cheaper AMD-based systems and more powerful Xeon systems with Xeon Scalable processors and/or 10x NVIDIA GPUs. The question is whether there's a big enough market for $20k-$50k Mac Pro systems to cover the cost of Apple developing a die just for those systems. Even with the Intel Mac Pro, Apple are just one of many customers using the Xeon-W.
The real joke, is that there's no reason for Apple to have a single CPU architecture - moving entirely to ARM is politics. Like refusing to allow Nvidia to release GPU drivers, unless they gave up CUDA was politics, Tim's Apple has its goal of never being reliant on anyone for anything, and never allowing anyone else's standards or technology to be the primary customer attachment.
When you put a box decorator in a position of design, decoration supplants design.
Honestly, there's a danger in allowing any singular technologist to be a manager, because then their personal technlogy fetishes become the focus of the company.
Steve Jobs cared about what technologies could do, in terms of the products they could make, a lot of the Swift (and frankly ARM) crowd seem to be more into technological purity for its own sake, and it's a little creepy.
Imagine if Chris Lattner had never worked at Apple - we might be on another evolution of ObjectiveC, using mature libraries that only experienced developers like, rather than operating systems that feel like they're made by college kids using experimental javascript libraries, clogged up with "anyone can code" junkware applications.
MacOS X used to be marketed for its UNIX credentials - the fact that your laptop could run the same industrial strength system that ran traffic lights, and other critical infrastructure. We measured uptmies in multiple months, and rebooted only for hardware changes. I don't think anyone would suggest you run even a basic webserver on macOS now. It's become a lightweight system that "needs" regular periodic reboots.
Last edited:
I remember in the early days of 2019 Mac Pro, someone (reviewer?) discovered CPUs on it might just be replaceable or upgradeable, certainly not as modular as MPXs though. That being said, I wonder if Apple would drop x86 lines entirely; one of the reasons I hopped on board for Mac Pro was because there might not be another Intel Mac again. Looking back, I'm still not sure if prospect Mac Pro buyers can make the jump to Apple Silicon as easily.
Side note, I wonder what this means for the futures of MPX, and possibly just discrete GPUs for Mac desktop lines.
Side note, I wonder what this means for the futures of MPX, and possibly just discrete GPUs for Mac desktop lines.
Is Nvidia doing 1TB of usable RAM ?
From Nvidia's recent more detailed blog on Grace.
Inside NVIDIA Grace CPU: NVIDIA Amps Up Superchip Engineering for HPC and AI | NVIDIA Technical Blog
Discover the key features and benefits of NVIDIA Grace CPU, the first data center CPU developed by NVIDIA. It has been built from the ground up to create the world’s first superchips.
Note the "Up to 512GB" fo the LPDDR5x . Later in the blog article
" ... With 144 cores and 1TB/s of memory bandwidth, ... "
Also post Hot Chips 2022 presentation this year. In the Next Platform's write up
" ...
That fat pipe between Grace and Hopper means that the CPU and the GPU can use the other’s memory almost directly. So in a sense, the 80 GB of HBM3 memory on Hopper is a high bandwidth memory for the CPU and the 512 GB of LPDDR5 memory on Grace is a fat auxiliary memory for the GPU. ... "
Details Emerge On Nvidia’s “Grace” Arm CPU
Imagine, if you will, that Nvidia had launched its forthcoming “Grace” Arm server CPU three years ago instead of early next year. How much better would
www.nextplatform.com
Graphic from the same article ( lifted from Nvidia's slides )
similar ... from Sever the Home overview
"... While there is a penalty for going to the Grace CPU’s larger 512GB memory footprint, NVIDIA seeks to minimize the impact. ..."
New NVIDIA Grace Arm CPU Details Ahead of HC34
Ahead of HC34, NVIDIA is allowing us to share details of the new NVIDIA Grace Arm server CPU and thoughts behind why NVIDIA needs Grace
There is decent chance there is 1TB of 'raw' LPDDR5x there. However, because LPDDR5x doesn't directly implement ECC for data at rest in the RAM module and the need for sever RAS ( reliability , availability , serviceability ) requirements, it would not be surprising if half of that raw capacity disappeared into overhead and bandwidth service agreement needs.
In the extended long term some denser LPDDR5x options may appear. But pragmatically even Nvidia isn't doing it.
I suspect those early 'sneak peaks' at Grace last year were more so about the "super chip" Grace ( 2 x 512GB ). Which goes substantially past "one die". And there is a limited "remote" tax for the other half that capacity ( similar t tax to get to Hopper's RAM ) . Layer on NUMA capable server Linux on that and it shouldn't be a problem.
Nvidia doesn't run a Uniform/Unified Memory connection over their C-to-C link off to Hopper GPUs. In terms of uniform/unified memory access Nvidia's Sever cap is 512GB capacity.
The Utlra already has 32 memory channels. Apple has different issues than this Nvidia Server chip. Apple has more concurrent accesses going on inside the semi-custom memory packages. They run 4 connections to a memory stack ( looks like Nvidia is running less. closer to 2 I think ). Apple has different thermal constraints. And a pretty good chance they don't want to do ECC if they can avoid it. ( not useful elsewhere in the product line right now. )