Mac Pro Enthusiasts Raise Concerns Over Upgrade Limitations of Apple Silicon

[Unregistered 4U]

The only thing that shows is that Apple does not serve the low cost and embeded markets. And your claim that they were "outperforming a larger percentage of them every year" seems to imply that Apple was somehow pulling away. But I don't think that's the case at all.

Apple’s processors keep performing better and better, such that the lowest end shipping M series chip is performing even BETTER than the lowest end shipping Intel and AMD chips. Math says that, as they continue to market those low performance chips year after year and Apple improving year after year, MORE of what they make will be below the lowest of what Apple makes. You’re right, they don’t serve those markets, BUT that doesn’t change the fact that those that do intentionally ship VERY underpowered solutions. It’s not said to be controversial, just a realization of the factors at play indicate that Apple’s lowest in processors will stick near the top of performance curves while most of everything else will be FAR below it.

Huh? Apple doesn't make any low end processors. As a result, you can buy x86 laptops for less than half the price of the cheapest Macbook. By all estimates the M-series CPUs are very expensive since they use TSMC's cutting edge manufacturing and packaging technologies and are made in relatively small numbers.

Right, and that’s why ALL of them are near the top of the performance curves, not the bottom where most of what Intel and AMD ships to customers is.

There is thriving competition between manufacturers in the PC market, as well as between Intel and AMD. We currently see in their financial results what happens if they let themselves fall behind.

If Intel ships a processor line that runs hotter than they said it would, the vendors who were in line to buy it are STILL going to buy it! They’ll just beef up their temperature control hardware to compensate. Intel missing their performance targets will still yield their expected sales because it’s not like those customers are going anywhere else. AMD STILL can’t meet Intel’s production volumes.

I think that's exactly what they are aiming to do with products like "Lunar Lake".

Yes, and what Intel “aims” and what they “do” don’t have to align as long as they can physically make more chips than AMD. What else are folks that need x86 solutions going to buy? I can predict now that they won’t be as performant and won’t run as cool as expected, but Intel will still ship a lot of them because ANY vendor that says, “No, this doesn’t suit me, I’m going to wait it out” is going to have marketshare taken by those that just alter their hardware to suit and ship it anyway.

 

[bradman83]

There are still very, very few apps that make use of al the special silicon in the new chips, like the 'neural engine'.

Any app that uses CoreML will be accelerated by the Neural Engine. Just in the photography arena apps like Adobe Photoshop, DxO PhotoLab & PureRAW, the Topaz AI suite, the Affinity Suite, Pixelmator's products, Luminar, and more all use CoreML and in turn the Neural Engine.

 

[Unregistered 4U]

Yes, I want to know the technical reason.

The technical reason since the advent of Apple Silicon (and the reason why folks have been saying “likely no memory configuration or external GPU going forward on any Mac”) is the way the system is built, all of the RAM used by the system AND the GPU is on the SoC. There is no allowance or any capability for the current SoC’s to have RAM or a GPU that’s not ON the SoC. The system doesn’t even “look” for it there as all the connectors related to RAM/GPU are right there on the SoC. It’s like if you took an entire Intel MB, shrunk it down to the size of an Apple Silicon SoC. The RAM is inside the resulting “Intel SoC”, there is no “installing RAM” outside of it because all the traces and the teeny tiny slots if there are any are shrunk in there.

The reason for THAT is both “that’s the way SoC’s are built” and “that’s the way Apple built their SoC”, either of which are just as valid. Apple COULD have designed their solution more like what folks are used to when they think about a “proper” PC CPU, but, as Apple’s experience with building their Apple Silicon solutions comes from the work they did on the iPhone, any variation to something entirely different that hadn’t been vetted for literally for years would have likely been a lot more risk that what they actually did. I think there’s some merit to this because if you look at the number of companies making fairly high performance CPU’s there HAS to be a reason why, for the way Intel and AMD makes their chips, they’re the only two.

COULD Apple be successful making a computer like the others do? COULD means a lot They COULD… make a computer that allows external RAM and Nvidia GPU’s and stackable processors with 4 times the performance of the fastest current Intel chip without any cooling required… given enough desire, money, time, and expertise. So, while I feel it’s unlikely, this is not MacTOTALtruth, it’s MacRumors, so folks are free to wish all sorts of exotic solutions. Imagination is a cool thing, it is it is.

and Mac users don't even know what the hell any of that is.

A lot DO know what it is. And, for that want a piece of that action, they just buy a PC to get it.

 

[deconstruct60]

iGPU in SoC handles display output, utilizing UMA RAM...

ASi GPGPU(s) handles compute/render tasks, using on-card RAM...

Two problems there. First, so far Apple hasn't provisioned any reasonable I/O channel to communicate to a relatively far away card ( 4-8 x1 PCI-e v4 lanes isn't going to cut it. )

Second, if going to copy and send to do "extremely embarrassing compute" on a remote card ... there are already other players with even bigger TFLOP budgets. It is a non-unified/uniform memory hand off so have to change code anyway.

The AMD MI210 is already a "send me datat to run TFLOPs over" card. How many are they gong to sell? Some narrow subset of Mac Pro users and a smattering of eGPU users? 10-20K/yr ? Is that worth forking off a whole another die for? ( even more so when barely keeping up with the regular's; if indeed 'dropped the ball' on M2 Extreme).

There is a growing class of 'compute' GPGPU implementations out there ( 'data center' GPUs : Intel Max GPU (Ponte Vecchio ) , MIxxx , Nvidia high end Axxx that are just focused on compute).

MI300 is somewhat putting a SuperCompute node on a card. Toned dow a bit ( off 500-600W budget) could have an x86 Linux node with substantive offload compute running inside the Mac Pro box. Only driver needed is a virtual Ethernet over x16 PCI-e v4 connection on both sides. That is not that invasive to the bullk of either side's OS image and extremely fast. And for apps that have 'ship batch job off to the cloud' there is really no huge feature difference.)

The disconnect here is where folks want to rip either 100% of the CPU cores off the die or rip 100% of the GPU cores off the die and return to the 'purity' of the old days. That is a highly dubious spend for Apple. It blows up lots of software that Apple has beat the drums on to get developers to write and extremely optimize over the last 3 years and almost useless for the rest of the Mac product line up ( which is mostly paying the R&D bills. )

 

[deconstruct60]

Yes, I want to know the technical reason. So far it seemed to me that a computer was made of a CPU, a GPU and RAM, among other things, and that the RAM and GPU could be swapped if they weren't soldered in.

How many mid level to high end GPUs have socketed RAM? AMD's , Nvidia's , Intel's.

Same baseline techincal reasons. What you are ignoring in your examples is the soldered VRAM on the GPU cards being cast aside. The hyper modularity folks don't have hissy fits over that. The technical needs are for

1. relatively very wide memory bus width. The bus width on a high end GPU is likely going to wider than most mainstream CPU only packages.

The bus with on Intel and AMD iGPUs is the same as the mainstream. But are any of them mid-high end GPUs ? Nope. If you cannot get the data to the arithmatic units fast enough then they starve. Hence the wider than 'normal' buses. Don't have the space for the bulkiness of DIMMs busses. ( and that new connector for laptops is an even bigger space waster. )


2. caches are not going to get you 98% hit rates all the time. Caches are effective over a percentage of the larger capacity they cover. 10MB cache on a 1,000,000MB storage drive isn't going to be as comprehensively effective as a 1,000MB cache over the same amount. The lower the percentage cover the more likley the hit rate will be lower. (even more so as throwing multiple concurrent workloads grabbing at different ranges of the primary source being cached. Get more mapping conflicts and data being evicted prematurely )

Joe Blow putting in random RAM capacities means can't have a predictable hit rate over a range of configurations.
(don't tend to see high performance GPUs with extremely large variance in configurable RAM. It is bad for optimized turning. )

3. power. The farther away the memory is the more power inefficient it is to get to.


All of those lead to why have dense package GDDR VRAM packed around the perimeter of a GPU package on a discrete card. And also why in the even higher performance zone will see stuff like HBM memory where stacking RAM dies even taller and with wider busses.


The other issue progressing over last 5-6 years is better packaging technology. Which is stacking things higher and closer at lower power consumption. Pointing back at the technology limits of the 1980's isn't going to give much insight into what can do now.

I get that AMD, Intel, and Apple make different types of CPUs. But both AMD and Intel allow you to have whatever RAM and GPU you like. Why not Apple? I suspect it's just "we want to force you to pay 5x price for RAM and we want you to get the M2 Pro Max Ultra just to get an extra GPU core".

Apple isn't mostly doing a CPU. Intel was Apple's largest GPU vendor ( not AMD or Nvidia. ). That's is a primary 'dump' target here. Lots of folks have spun this change as Apple dumping Intel ment that Apple was obssesed with just dumping Intel CPU cores. It was both.

THe Mx Max die is a GPU with some CPU/NPU elements sprikled around it. Apple is primarily building a mid-large size GPU so it really shouldn't that it looks like the mid-large size GPUs that other folks also build.


This is usually get 'twisted' and claim that if Apple is doing a GPU then they 'have to' come out with some 4090 killer ( whatever the largest , most expensive GPU currently out there... have to 'kill' that or it is a complete bust. ). That is nonsense. They don't have to do everything for everybody to be competive in a targeted area. AMD often got into trouble in the first half of this century trying to 'monkey see , monkey' do everything that Intel was doing.
AMD narrowed their focus and they got better ( reusable chiplets for both mainstream dekstop and severs. for a while did lower half of GPU range and then did upper half ). Apple also when Jobs came back and tossed have the model variations out the window.

Doing everything for everybody isn't necessary. Even in Mac Pro space.

But I think it's a huge, huge shame that there are companies like nVidia who make great GPUs, they innovate in the AI revolution, there are amazing games that make use of raytracing and all kinds of cutting edge technology, there's VR... and Mac users don't even know what the hell any of that is. Apple is like the North Korea of tech. "Our tech is perfectly good, you're not allowed to try anything else."

Just a few years ago we had nVidia GPUs in MacBook Pros. Then for some odd reason it was limited to AMD GPUs.

It is far, far, far, from "some odd reason". Nvidia did an 'embrace, extend , extinguish' attempt on OpenCL. They put CUDA in front of Metal once Apple moved onto Metal as the alternative. when there was a huge problem with Nvidia iGPUs they ran anyway and left Apple holding the financial bag. Finally toward the end Nvidia had some code that would "Halt and catch fire' for every new macOS kernel update. Throwing drivers out in a completely uncoordinated approach to Apple's plans. And Nvidia relatively publicly blamed all the quirkiness all on Apple. (we have drivers but we have zero idea why apple won't sign them. Hoping to get fanbase to put more heat on Apple to get a reprieve. ). That just put the icing on the cake.

It isn't 'odd'. Nvidia and Apple got into a 'war' as to who was more powerful and Nvidia basically got dropped as a bad partner to Apple. They were bad. Apple contirubted to the decline also. They are just as dogmatic about Metal first as Nvidia is about CUDA first. But it is Apple's systems and their operating system. And Apple's Metal first strategy was deeply intertwined with the iPhone business. Apple had zero incentiveto let Nvidia do an "embrace, extend , extinguish" move on Metal. There is no 'new' Mac Pro business they were going to drive that was going to offset the iPhone business.

Technically , Nvidia is just a 'subcontractor'. If you are a subcontractor who is always causing drama for the prime contractor ... eventually you get dropped. After a while Apple just stopped signing their drivers. That's it.
Some folks in the general PC parts world start inside out. First pick an Nvidia card and then wrap the rest of PC parts around that. In that alternative 'world' Nvidia is the prime contractor calling the shots. Apple doesn't work that way at all.


Both sides got much bigger over time on business areas that had almost zero overlap. Apple didn't 'need' Nvidia financially or technically and vice-versa. It was very easy for the 'war' to escalate until both sides 'blew up' their side of the bridge between them.


Nvidia is a pain to work with. Almost no potential 'business partners' wanted them to buy Arm. Similar dust ups with other vendors. Nobody trusts Nvidia except their end users who buy into the moat that Nvidia digs around their products.

But you could still get external GPUs, like the BlackMagic eGPUs via Thunderbolt. It seemed like things were expanding, that more options were becoming available...

There was a trend toward the end of the Intel Mac era where Apple's boot firmware got closer and closer to UEFI. Early on there were special "Mac boot ROM" requirements for video cards because Apple was mainly interested in EFI (not backward looking BIOS). That UEFI support crept in mainly because the Intel CPU packages needed it.

Given a free hand as to what the boot firmware was going to be Apple tossed UEFI out the window. Macs officially boot macOS and that's it. Can do some hackery with Linux but it has no official technical support coverage.

From the first WWDC 2020 when the Apple Silicon was announced Apple said the GPU driver coverage was only Apple GPU. That didn't change at all over the next two WWDC session ( 21 and 22 ). Maybe it would be 'low priority' so it would slide to '22, but nothing. The major driver coverage expansion announced at WWDC 22 was that DriverKit drivers should work on iPadOS on M1/M2 iPads. Crickets chirping on any 3rd party GPU drivers.

There are over 50+ cards that work in a Thunderbolt PCI-e card expansion box. Just not anything that deals with the early boot set up that isn't a generic driver like USB or NVMe or SATA. THere is no DriverKit coverage for display GPU in the object hierarchy for the framework ( unlike the preceeding IOKit framework). Thunderbolt works with Apple Silicon. That isn't the issue.

The big push for the Blackmagic eGPUs was for the laptops which had limited GPUs. The new Mn Pro and Max are no where near as limited. The Mini ... again now no where near as limited (especially if loop in a Studio Ultra).
The iMac Pro was less limited. And the MP 2019 not particularly limited at all in an "augment through eGPU" wise. Those last two were not where eGPUs were primarily targeting.

and now all Mac users are stuck with integrated graphics made by Apple regardless of how much they spend on their computer,

I suspect that is part of Apple's point of excluding 3rd party GPUs. They want developers extremely focused on optimizing for Apple GPUs. If there are not other options there are no distractions. Apple is trying to get rid of the notion that 'iGPU == slow' . That doesn't have to be if don't cripple the memory bus to kowtow to DIMM options.

a company that has no experience making GPUs and has not brought gaming, AI or VR to the Mac in any form.

That far enough in the delusional zone that appears to come from an alternative universe. Apple does billions in gaming on Apple GPUs. (the high end PC game market isn't the whole gaming market). No AI. chuckle you app exposure is limited. Apple is far more focused on AR than VR and have laid lots of foundation.

Apple's approach is closer to a console gaming than perhaps more hard core Windows PC gamers are comfortable with. It is somewhat of a hybird approach. Apple is looking to match highly optimized graphics code to a finite set of good hardware. A small enough set so that can optimize very well for all of them, but not so large where have constantly mutating drivers trying to plug every quirky hole that corner case chasing apps dive into.

When Apple does VR it will deep links to be mobile (on battery) VR. Ai on battery . Gaming on battery . 75+% of Mac sold are mobile capable. That is where the inertia is. The Mac Pro is gigantic chasm away from what is driving the vast bulk of Mac sales ( it is likely down in the 1% (or less) range). Highly focused , well optimized code when run on a plugged in and larger SoC will often run faster. (brute force code will proably require more expensive brute force to run faster.) They are not likely to completely miss out on all of the mid-high end range of performance coverage. Especially if have a high quality of optimized code on their GPUs than the competition.

They're just using their new CPU as an excuse to lock down their already massively crippled ecosystem even more.

Again no. Apple is more about creating a system than in a subcomponent. If looking to Apple just to make incomplete subcomponents you are kind of shopping for a pork sandwich at a conservative kosher deli.


If the unified memory wasn't buying them largely differentiated performance you'd have a point. It is a trade-off that apple is making. Don't like the trade-off then don't have to buy from them. Apple isn't trying to make everything for everybody.


Apple does need to add the ability to provide more general PCI-e bandwidth to the Ultra (and up) SoC that will never land in a laptop. Some acceleration that adds more performance in 'scale out' (in the box) workloads is likely necessary because there is only so much 'horsepower' you can put into a single package for a fixed fab process technology level. These days a supercomputer is a bunch of little computers cluster in a very expensive, custom network.

 

[goborgh]

I have a feeling they forgot about the Mac Pro when planning to move to Apple Silicon. Or maybe the Mac Pro haven't been released at that time and since they're so secretive even among themselves, the right hand didn't know what the left was doing and now here we are.

 

[deconstruct60]

Honestly though, if Gurman is going to argue that lack of support for 3rd party options is due to everything being on package or on the SoC (and it is currently not supported), then there's not much expansion options left to utilize those 8 PCIe slots.

If there is nothing then why are people buying these?

xMac Studio - SONNETTECH

3U enclosure to install and secure one Mac Studio with optional PCIe card expansion in a standard 19-inch rack.

www.sonnettech.com

Probably because if you go look at the Tech Specs there you'll find a PCI-e compatible card matrix that has over 50+ cards that work with M1/M2 generation Mac systems.

There are lots of cards that exist that are not GPUs. In fact, more non GPU PCI-e cards options than there are GPU ones. "Almost nothing to go into those slots".... not hardly.

And this xMac really doesn't fill out the role because a couple of cards the matrix list carry a footnote that say that the maximu capabilities of the cards are diminished due to Thunderbolt bandwidth limits. (e..g, put a x8 PCI-e v3 (or v4) card into the 'expansion' slot and only get x4 PCI-e v3 delivered. So 8K video capture card that doesn't quite do 8K HDR, but OK for 4K work. )

That list of 50 cards would be longer for some cards that bandwidth limitation is going to diminish their market value.
Faster networking.... what is the point if can't get the data to the computer in a timely fashion. four x4 PCI-e v4 M.2 RAID cards ... again what is the point if highly throttled.

The puzzling part is why Apple would not do a two x16 PCI-e v4 addition to their 'desktop' line of of SoCs and feed that into a similar two input PLEX PCI-e switch that is in the current Mac Pro 2019 that 'flex' provisions the data to six of the slots currently. There would be no backwards backslide on total aggregate bandwidth out to six slots.

Users that needed > 6TB of data could easily have better and faster data storage options inside the box. Older audio capture cards that have many interfaces and space more than one slot would have room. Could do 8K (and up) video capture. Who wants to do that on a machine that has 4 (or more ) top end video en/decoders built into the SoC that can do 8+ streams of 8K video ???????? If have reams and reams of 8K video , even compress , then likely have a storage capacity issue also.

Except for AV1 , Apple has some market whipping A/V en/decode inside their boxes.

Apple doesn't sell most of the 50+ cards in the main Mac Pro BTO configuration page. But they also don't sell any 3.5 HDDs and folks toss those inside of MP 2019 containers too after buying a bracket to put them into and 3rd party cords to hook them up.

Without display GPUs the entire PCI-e universe implodes into nothing-ness is myopic. GPU are not pragmatically the whole space.

A top end Mac Studio Ultra + xMac Studio Echo III (three instead of siots is $4,999 + $1,600 = $6,599. If Apple can bring in a Ultra powered Mac Pro for $6,999 that had three more slots and far more bandwidth ... why not?

The market is already there... product is being sold into it now.

If there is no M2 Extreme that would shrink the market , but even with a minimally augmented Ulra to decently provision out 6 PCI-e slots with 'same as MP 2019' backhaul bandwidth there are users that will find utility in that.
( Apple just could not squat on a M2 Ultra for 2-4 years ... they'd need something in 12-16 months. )

Longer term if Apple adopted PCI-e v5 and CXL 2 or 3 there are several new classes of accelerators that would come in handy. RAM SSD cards. Data format specific accelerators (where push specialized processing out closer to the RAM so don't need to 'round trip' to the CPU core and back to get work done) . etc. But Apple isn't going to get there if can't even provide a single x16 bundle out of the SoC. They need to start somewhere. Apple hsa overtly done next to nothing to adopt/promote CXL so I doubt it is coming soon. But if it continues to evolve rapidly over next 2-3 years or so , perhaps some "fear of missing out' will kick in at Apple.


Future likely is also going to have 'mini supercomputer node on a card" options where it isn't just GPU cores , but a CPU+GPU combo on the card running a lightweigth Linux image ( like a supercomputer node) that could plug into the Mac Pro. ... would need a a couple of double wide (or more) slots to stuff into the box. 'add on' TFLOPS don't have to be a display GPU.

 

[2Stepfan]

I believe the original Ultras “couldn’t scale” given the A series starting point.

As pointed out by people like CMaier there is a difference in the next (now released) architecture that doesn’t have that limitation.

I know members here are adamant that Apple is failing here, but it really does look like M2 was always supposed to be N3 and then the world blew up.

Would you be kind enough to summarise how the architecture has changed? I have seen allusions to this, but I don’t really understand it.

I think @cmaier may no longer be around, based on a cursory search, which is a shame because he was always a favourite poster.

 

[deconstruct60]

I have a feeling they forgot about the Mac Pro when planning to move to Apple Silicon.

The Mac Pro isn't any more 'forgotten' than when Apple drifted from 2010-2013 (3) with not much new. Or 2013-2019 (6) with not much new. MP 2019 drifing to the end of 2022 or into 2023 ( 3-4 years) is somewhat average for them.


Therefore, probably not. The Mac Pro had a T2 chip in it. Apple didn't say it but the 'T' is likely indicative of 'transition'. All the stuff with a T2 in it probably had a path to Apple Silicon. T2 chip usurped lots of the same stuff. It took over the boot process. It did a UEFI firmware handoff to the Intel CPU , but it was in control and largely put an end to shenigans with hacking firmware to bend the rules.

The T2 handled speakers , mic , camera , audio ... just the Apple Silicon is now. Mac Pro didn't skip that.

The T2 handed the SSD controller duties... Mac Pro didn't skip that and still highly likely true in the Apple Silicon version of Mac Pro ( SSD modules ... being not SSDs included ).

The Mac Pro has significant complicated infrastructure built in to provision video to four standard built in Thunderbolt ports. 4 Thunderbolt controllers built into the Mx Pro and Max SoCs. So that part could get tossed. Don't need a 'work-around' for something the new SoC fixes.


Six of the slots are all provisioned through a PLEX PCI-e switch to allow user to allocate a fixed amount of bandwidth with user discretion. If Apple cuts back to just two x16 PCI-e v4 lane clusters can still do six slots with an updated switch just fine.

The Intel W-3200 and projected (in 2018 or so ) W-3300 CPU were in the 300W range. 4 * 80W ( Max like TDP) is 320W. So the SoC package cooling is aligned with the chassis.

Using less power over the PCI-e slots ... really not that hard to do. But do retain flexibity for what might be an unexpected surprise.


Would be handy to have a rack version of the case with the same proportions. ... that is still true.

The outer case itself has no major breaks. The major fans attached to the case have not no major breaks. The power supply ? Nope. Thunderbolt I/O ports on replacable smaller scale daughter cards? Check... still useful even when provisioning the TB signals out from the SoC.

DIMM slots gone. ... move the SSD modules closer to the SoC. Done. If Apple was nice they could throw a M.2 slot where the SSD modules were on the backside. And leave some USB / SATA ports on the board (hooked to a discrete SATA controller.). J2i frame for HDDs.... reuse it also.

The MPX connector disappears and the whole outer case and space frame implodes as a design? Not hardly.

Re-using the other case and frame and the major fan mounts isn't hard.

Even if apple did another Intel version the logic board was going to change. Different socket. Different RAM. Different PLEX switch ... etc.

There being no 3rd party GPU drivers in macOS on M-series is a software and strategy policy issue. There is not direct coupling to the physical design of the current Mac Pro case at all. Apple shoved the M1 into the 2018 Mini case and left it half empty. It was expedient. Same thing could be in play here.


UlltraFusion means that Apple could stuff two x16 PCI-e cluster managers on an add-in shim between two "Max with UltraFusion" dies if they wanted. Or just don't have identical twins (trade excess TB and SSD and secure element controllers for PCI-e controllers) I suspect though that this may be a different chiplet disaggregation though. The Ultra and possibly the Max in the Studio would share a subset of chiplets with the Mac Pro ( where Mac Pro got extra enough to do the slot provisioning but not trying to exactly twin the Intel solution.)

The Mac Pro would be one that drifted from the plain Mn chip the most so do it last. So it likely always was at the end of "about two years" timeline. The pandemic just slid that into after two years. The pandemic was not in the 2017-2018 era planning process.


If Intel has done a timely and good job on the Xeon W-3300 (Ice Lake) processor then might have been a late 2020 or early-mid 2021 Mac Pro. That processor was late and hotter than planned. It was skipped by Dell, HP, and Lenevo. Apple also skipping it is no great surprise. If Apple had known much earlier that their Mn Extreme wasn't going to arrive on time they might have consider W-3300 as a viable "Plan B" to extend the Mac Pro a couple of years. But it wasn't.

Or maybe the Mac Pro haven't been released at that time and since they're so secretive even among themselves, the right hand didn't know what the left was doing and now here we are.

When Apple got on stage to demo the MP 2019 they had a slide deck that showed storage cards , video capature cards, audio DAW cards ... that could all be stuffed into a Mac Pro box. "We can put 5 Avid HDX cards in here. " ... that will still be true when the Apple Silicon version comes with 6 slots spaced the same way they are now.

All the 3rd party PCI-e card stuff was NOT retail off the shelf 3rd Party GPU cards. There were lots of folks who saw the case and viewed it only in that way. But that is not what Apple got up on stage and presented at all.

 

[DeepIn2U]

No. This simply can never happen. The entire reason the M1/M2 is so fast is because of unified memory. The old Intel computers used different memory in the CPU and GPU and then connected it with a PCIe bus. They moved data. the software had to push the data over the PCIe bus. So the number of bus lanes and the speed of each lane was a big deal because it was a bottleneck.

But with unified memory, the data NEVER MOVES. It is not that the bus is faster, there is no bus.

As soon as you have expandable RAM then you need some kind of bus to place it on. this ruins the entire advantage of unified RAM. Now all the data ging to and from you add-on memory has to go over a bus that connects the sockets to the main memory.

With the "M" series Apple removes a huge bottleneck which is the need to move data. They simply can not give his up without going back to having different kinds of memory

And why, really, do you need to add memory later? Just buy it up front and be done with it. The Mac Pro is for running some specific app that you need, and the cost of the Mac Pro is likely trivial compared to the cost of the work being done. No one is caring about saving $200 by adding RAM later.

The problem with just buying the ram one thinks they need una. Professional top end computer, at the start, is that software contributes to evolve.

Improvements
New features and functionality in updated and entirely new software/vendors respectively.

The whole premise of a computer is for it to grow to your needs. A tablet and smartphone can and always has been limited to the constraints of its OS being mobile. You see how far expandable memory in smartphones went right?! See my point.

If the software you have at time of buying the desktop top end computer continues to evolve so too will the need for ram, possibly Ana traditionally. When the G5 was launched, Logic Pro (before Apple purchased it and employed the staff/founders) it could handle 210 intruments or tracks simultaneously. When the Mac Pro on Intel debuted it was over double! Each was capable of more than 512GB of RAM. That was insane back in those days.

 

[SouthPalito]

It’s pretty obvious this product line is essentially DOA since the business case for it has been wiped out. Almost everyone who needs serious power moved long ago to windows/Linux workstations which support the latest in CPUs and GPUs and scientific libraries or runs all their intensive workflows on cloud instances.

 

[gank41]

It’s pretty obvious this product line is essentially DOA since the business case for it has been wiped out. Almost everyone who needs serious power moved long ago to windows/Linux workstations which support the latest in CPUs and GPUs and scientific libraries or runs all their intensive workflows on cloud instances.

You clearly haven't been to any music recording studios.

My dream for some day when I retire is to buy one of these machines, new, and hook up a 16 Channel DAW and setup a bunch of mics and equipment so all of my friends can come by and I can just press record in Logic and just let it go for hours and hours, recording all of the weird stuff we make up, never having to worry about it crashing because of some disk write buffer error, losing all that we had just played. We'd probably just put the music online for free, along with all of the other free music we make. My point is that that's just little ole me. Nearly every studio out there uses Macs. And these AS Macs just blow away the Intel ones. If I could have a faster interface with my DAW? And still be able to open it up and add more internal storage or maybe something else I haven't yet imagined? Sweet!

 

[Boil]

You clearly haven't been to any music recording studios.

My dream for some day when I retire is to buy one of these machines, new, and hook up a 16 Channel DAW and setup a bunch of mics and equipment so all of my friends can come by and I can just press record in Logic and just let it go for hours and hours, recording all of the weird stuff we make up, never having to worry about it crashing because of some disk write buffer error, losing all that we had just played. We'd probably just put the music online for free, along with all of the other free music we make. My point is that that's just little ole me. Nearly every studio out there uses Macs. And these AS Macs just blow away the Intel ones. If I could have a faster interface with my DAW? And still be able to open it up and add more internal storage or maybe something else I haven't yet imagined? Sweet!

OWC has RAID cards with eight M.2 slots, $12K for 64TB, (near) infinite tape roll...!

 

[Mr Rib]

I guess the only reason for existence of a Mac Pro in such a form is good PR. "See? We are still a power user company as well, not just mainstream!". If what Gurman says is correct, tt's pure PR because this product is, well, for noone. A good M2 Studio Max will fulfill the needs of most power users, for guys working in 3D, Macs are not that appealing anyway as they want to swap high end graphic cards for new ones every now and then - they'll keep using PCs. And for a tiny niche that needs SERIOUS power, they'll rely on cloud computing, rack computing server units etc.

 

[Yebubbleman]

If you are going to pay a hefty price for a Mac Apple should let you consider customizing it the way you want. It’s simple as that.

Why else do we pay premium prices for Apple products?

You pay a premium because you don't want plastic as an option. Also, during the Intel era, you didn't want a Core i3 option (other than the two or three times it was on offer). Nor did you want a non-retina display once Retina displays were a thing. With Windows and Android, you have the option to spend less money and get a lesser product. With Apple, you do not have this option.

I hope Apple has a plan here. If it is “just” an M2 Ultra, the motherboard will be in a tiny corner of the case (Given it fits within a Mac Studio. Without upgradable RAM and GPU you basically removed the reason you would buy the Pro over a studio. So you are left with a giant, heavy, case and an extra few $K in price for…what?

You do realize that there are other uses for PCIe expansion other than graphics, right?

The Mac Studio is the Trash Can Mac Pro, but implemented in 2021 when Thunderbolt 4 can actually handle the external expansion Apple envisioned.

Breakout boxes are still way expensive. I'm not saying the Mac Pro is inexpensive by any means. But Thunderbolt 4 as an expansion solution is far from ideal.

Best case scenario, Apple makes M3 pro-extreme daughter cards that can be added or removed for RAM and CPU upgrades. One motherboard for those that want PCIe slots with fewer daughter cards and another motherboard for those that want more daughter cards with no PCIe.

I'd hope that they'd make the Mac Pro chassis for years letting an aftermarket grow for used modules every time they come out with a new SoC.

Socket-able SoCs is probably our best case scenario.

What’s the point of the enormous chassis if you can’t add anything to the pCie slots?

You're acting like GPUs are the only reason to have PCIe slots...

"Popular with Creative Professionals"

I'm wondering who the market is for $20-50,000 Mac Pros, outside of Pixar and Marvel Studios.

Graphic Designers? Print Publishing? Web design? software development? Photographers? Audio rendering? 3D modellers?

There's nothing in those workflows that a Mac Studio or M2 Pro Mini can't handle easily, with spare capacity left over. And they've been shown handling video editing effortlessly.

An M2 Pro Geekbench-marks above a 16-core 2019 Mac Pro. The M2 Ultra Studio is going to be insane. For which creatives is that not enough?

Broadcasters. Extremely high-end audio folks. PEOPLE THAT NEED PCIe SLOTS FOR THINGS OTHER THAN GRAPHICS CARDS!

As offered in other threads, Mac Pro is a computer for the few (with plenty of cash) who want/need a very flexible, easily expandable & upgradable Mac. Those who do not need that have plenty of Mac options from which to choose. If this Mac Pro doesn't deliver on that, it faces the same issues that make the Trashcan and Mac Studio NOT be the new Mac Pro.

So I'm 'thinking different' about Mac Pro potential...

Just because Silicon locks down RAM & GRAPHICS & SSD doesn't mean a Mac Pro must too.

No, actually it totally means that a Mac Pro must too. See this video at around the 1-minute mark: https://developer.apple.com/videos/play/wwdc2020/10686/

It's inherent to the design of Apple Silicon as a Mac hardware platform. You can't get away from it.

For RAM:

• Slot the Ultras, so that someone needing more than the MAX associated with one Ultra can simply add another Ultra: 192GB RAM becomes 384GB. Create a "grand central"-like way to make them work together, like that technology on the old PowerPC Macs (what was that called???) that allowed tasks to be spread over several Macs linked together. Someone needs more than 384GB RAM? Slot in another ULTRA for 576GB. Make all 3 work together. Yes, this wouldn't be as fast as the EXTREME concept or jamming all of that extra RAM on one piece of silicon but it seems it could be as FAST as current Mac Pro slotted RAM modules.

Socketed SoCs can get you there. But that's honestly your best bet here.

• AND/OR support third party RAM anyway. It will simply be a little slower than the Silicon RAM. Again, in software, use a "Grand Central"-like approach to put most RAM demanding processing in FASTEST silicon RAM and put slower RAM processing needs in expansion RAM. Think of this like various caches in CPUs. This would introduce a concept of having FASTEST (Silicon) + FAST (regular RAM) + SLOW (SSD Swap "ram") in this setup. Those interested in buying Mac Pros now could still get their 1.5TB of "high performance" RAM made up of 192GB of fastest Silicon RAM and then traditional RAM that should be as fast as what it can be in the existing (aging) Mac Pro (if not faster).

Third party RAM CAN'T happen. RAM is intrinsically tied to the SoC. Apple is not going to re-architect how this works for it to function this way.

• BOTH: slotted ULTRAS for those who need more FASTEST RAM AND traditional RAM to the moon for those who need more than however many slotted ULTRAS could be put in the case.

This idea of slower RAM gets attacked but anyone who hops on an Intel Mac is unlikely to notice a RAM-based difference in speed of computing. Slotted RAM is no big speed bottleneck in typical uses of Macs. Yes, Silicon RAM is benchmark FASTER than slotted but real work experience probably can't notice except in very specific tasks. Every current Mac Pro user in the world only knows the speed of slotted RAM. So slotted RAM in this new one can't possibly feel any slower than what they know.

Slower or faster RAM performance isn't why Apple isn't going to do this. It's because they literally designed the hardware platform to use unified memory and then guided developers accordingly for the last two and a half years. They won't deviate from this here.

In many other threads, "we" are making justification arguments for half speed SSD chips in new M2 Macs because they are "good enough" and "no one can notice in real world use." That same logic(?) can apply here. Slotted RAM as an augment to Silicon RAM should not be slower or noticeably slower.

While the "good enough" argument used by pro-Apple zealots to justify the slower base-model SSDs on base-model M2 and M2 Pro Macs makes my teeth itch, that's not even applicable here. Yes, if it were, that'd be one thing. But that's not why it can't happen.

For GRAPHICS:

Anyone with an objective mind can find hard evidence that third party graphics are faster-to-much-faster than Silicon graphics (albeit at the expense of power vs. power per watt). I own a M1 Ultra myself and freely accept that to be true. Nevertheless, some people would like to use more power to get processing done faster than save a couple of dollars on a monthly electric bill so they can get things done slower but spin PPW as a huge benefit.

For those who need FASTEST graphics, Mac Pro should deliver the option of third-party graphics cards. Else that business that cares about power more than tiny PPW dollar savings goes to PC options. Mac Pro has no concern for battery life so Power should trump PPW anyway.

I see no way around this. If the existing market for existing Mac Pro is X and a chunk of X pays up for Mac Pro for the easy ability to keep up with graphics processing advancements, I doubt that chunk can be sold on "good enough" for life of device for ANY generation of Silicon graphics processing. Deliver the flexibility to use third party graphics card or just about force those people to buy their next Pro from PC makers. Those who feel it is Mac or bust will simply buy an Ultra in Studio, accept the full lockdown, and pay much less a Mac Pro... then "throw baby out with the bathwater" over and over to "keep up with" the latest Silicon graphics power.

Exception: revive the external graphics card option via thunderbolt connection? But do that and it seems it would have to be spread to all other Macs. So make this an exception INSIDE of Mac Pro and all other Macs are logically excluded from the option, while delivering a very clear differentiator in support of paying wayyyyyyy up for a Mac Pro.

Graphics expansion won't happen for the same reason why memory expansion won't happen: it's not how Apple designed Apple Silicon as a Mac hardware platform. It's also not how Apple designed the Apple Silicon builds of macOS. You have much greater flexibility when it comes to drivers on the Intel release of macOS. Not so with the Apple Silicon one. There is only one supported brand of GPUs on the Apple Silicon version. And, you guessed it, it's Apple's!

For SSD:

Up to 8TB is great (and insanely expensive when there's only one source of that storage). And Silicon SSD is VERY FAST. But even 8TB is not enough for everyone. Mac Pro owners enjoy the capability of using some of that internal space to fatten up their storage options, very cleanly, INSIDE the case.

This one seems easiest to address. The weakest way to support this need would be having some Thunderbolt ports INSIDE the case and basically shelves on which drives can be mounted. Plug into internal thunderbolt jacks just like one can plug an external enclosure into Thunderbolt jacks on the outside.

The best way to support this is to simply use the PCI-E lanes that will be necessary to support any cards and allow some of them to connect with storage on cards. Conceptually, this could lead to cards with many m.2 sticks mounted on them for gigantic and very fast RAID storage in a relatively small amount of space. Yes, this storage would probably not be as fast as Silicon storage but see the 3 tiers of RAM concept: Fastest SSD (Silicon) + Not Quite as Fast SSD (expansion cards/storage/internal Thunderbolt option) + External storage options like all Macs have now.

Just my "think different" 2 cents. Mac Pro has historically been positioned as "our more powerful Mac" not our most PPW-efficient Mac. I think new Mac Pro must be the most powerful Mac... not the same powerful (as Studio)... and not just as locked down as Studio.

The SSD is probably the one element that won't change over the 2019 model. It'll still most likely be in paired modules that you can replace aftermarket, but you'll still need to do a full DFU restore of the Mac in order to use them (with the data on the old ones being rendered useless as soon as that happens due to cryptographic pairing with the SoC - doubling as the storage controller as the T2 did before it). You'll probably still only be able to buy these from Apple at absurd prices.

SSD is already socketed in the Studio, so I don’t see expandable storage as an issue and you can always go external for that.

It's the same in the Mac Pro, albeit a little bit more flexible. My guess it that won't change with the next Mac Pro.

Tiered memory is a thing so they can implement that so users can expand RAM. It won’t be as fast as the integrated system RAM but it shiuld work well if that integrated amount is exceeded.

It is a thing, but it's much less likely for Apple to implement that than it is for them to just have the SoC be socketed and for them to sell the upgrade SoCs aftermarket. You may buy a base model M3 Ultra with 192GB of RAM and 60 GPU cores and later decide that the feature film you are working on demands a mid-range M3 Infinite (or whatever loony name that come up with) with 512GB of RAM and 100 GPU cores.

While Apple Silicon doesn’t support PCIe currently, I don’t know of any technical reasons it couldn’t. I can’t imagine they’re simply going to throw a massive fan inside that big case to boost performance and call it a day. They’ve go to to have a plan to utilize that space for expansion of some sort.

Not sure where you read that Apple Silicon doesn't support PCIe currently. It absolutely does and pretty much always has. Just because you don't see an interface doesn't mean that the bus isn't otherwise there and supported by the hardware.

Multi socket could solve these issues

See this video (at 26:26):

Apple is not a fan of multi-socket when it comes to Apple Silicon.

Do these M-series processors actually require everything be packaged on the same chip card or can they be configured in a more traditional manner?

Yes to the former; no to the latter. See the following video at around the 1-minute mark:

Explore the new system architecture of Apple silicon Macs - WWDC20 - Videos - Apple Developer

Discover how Macs with Apple silicon will deliver modern advantages using Apple's System-on-Chip (SoC) architecture. Leveraging a unified...

developer.apple.com

Apple could go any way they want. There are no requirements.

Umm. That's only true in the sense that it's Apple's machines and they can do whatever the hell they want. It's wrong in almost every other sense. Please see the above video at the 1-minute mark.

The benefits of "as is" is speed. It's generally much faster to build it all on a single piece of silicon.

However, the differences of that speed- while easily and very tangible in benchmarks- are not necessarily as noticeable in typical use. Defense arguments being offered in other threads about M2 half-speed SSD as "good enough" and "99% won't notice" apply here too.

IMO though, there is no all or nothing here. Apple could deliver BOTH. Silicon with fastest RAM and SSD embedded AND not quite as fast traditional RAM and not quite as fast traditional SSD options too.

They can't. That fundamentally goes against how they designed Apple Silicon from the get-go. It's not a matter of "Apple can do anything". They've built Apple Silicon around unified memory and around SoCs.

Such a Mac Pro could both address the purist fan view of "Silicon is best in every way" AND Mac Pro distinctions of flexibility, expandability & customizability INSIDE a case.

PCIe expansion isn't going away even if PCIe GPUs (for a Mac Pro) are.

Need more RAM? Load up on the speediest Silicon RAM and then add whatever additional you need in RAM slots. Yes, the latter will be a little slower but not noticeably slower unless you measure it with benchmarking tools.

Again, this won't work.

Need more storage? Load up with the speediest Silicon SSD and then add whatever additional SSD you need in m.2 slots. Yes, the latter may be a little slower but not noticeably slower unless you measure it with benchmarking tools.

There is nothing stopping you from buying a PCIe SSD and loading it into the current Mac Pro and that likely will remain the case for the next one as well. But, like the current Mac Pro, the primary internal storage will still be on a proprietary SSD whose controller will be on the SoC. Apple touts that as a security feature and you better believe that they're not changing that.

Any doubts? Dust off any Intel Mac you have and use it for your day-to-day functions. Do they seem noticeably slower because you are using an older processor, traditional RAM and traditional SSD? Almost anyone able to be objective would likely say no.

Your argument still neglects that Apple isn't changing what Apple is doing just because it's possible to do so. They feel the things done in the T2 and Apple Silicon eras of Mac computing are steps forward. They will not relent.

Yes, of course there are exceptions where Silicon as configured is noticeably faster but a key to this post is imagining a Mac Pro without dumping what makes a Mac Pro different than a Mac Studio.

Again, PCIe expansion and way higher-end SoC options will be what sets that apart. If they keep other elements of the current Mac Pro alive, then internal SATA will also be a differentiator too.

 

[DeLaSoul]

The PCI bus is from 1992. That's MacOS 7.0.1, floppy disks, serial bus, 68030 era.

Apple must be working on a M3 Extreme chip with a new high performance bus for memory, GPU and other cool media processing cards. One can dream

 

[Rigby]

Apple’s processors keep performing better and better, such that the lowest end shipping M series chip is performing even BETTER than the lowest end shipping Intel and AMD chips. Math says that, as they continue to market those low performance chips year after year and Apple improving year after year, MORE of what they make will be below the lowest of what Apple makes.

Except that Intel's and AMD's low end CPUs keep improving as well.

You’re right, they don’t serve those markets, BUT that doesn’t change the fact that those that do intentionally ship VERY underpowered solutions. It’s not said to be controversial, just a realization of the factors at play indicate that Apple’s lowest in processors will stick near the top of performance curves while most of everything else will be FAR below it.

You mean the performance/Watt curve. But that will change too as Intel catches up to TSMC with their manufacturing process, since that is the main factor that drives the power efficiency of Apple's CPUs. "Meteor Lake" later this year (first Intel CPU that is manufactured using EUV lithography) should significantly narrow the power efficiency gap.

Yes, and what Intel “aims” and what they “do” don’t have to align as long as they can physically make more chips than AMD.

And yet they are currently losing market share to AMD.

 

[LillDrutten]

Apple for sure need to prove that the concept can scale well. 24 core is just above the dream for laptops. For workstation 10 times is realistic. AMD have 96 core (192 threads) single socket CPU's that can be used in multi socket configurations. For sure those cores are far from a apples ARM cores. But the numbers compensate for that many times over. I think the Mac can be in trouble in to long run if the real pros does not get competitive performance on the mac pro.

 

[mdriftmeyer]

Apple’s processors keep performing better and better, such that the lowest end shipping M series chip is performing even BETTER than the lowest end shipping Intel and AMD chips. Math says that, as they continue to market those low performance chips year after year and Apple improving year after year, MORE of what they make will be below the lowest of what Apple makes. You’re right, they don’t serve those markets, BUT that doesn’t change the fact that those that do intentionally ship VERY underpowered solutions. It’s not said to be controversial, just a realization of the factors at play indicate that Apple’s lowest in processors will stick near the top of performance curves while most of everything else will be FAR below it.


Right, and that’s why ALL of them are near the top of the performance curves, not the bottom where most of what Intel and AMD ships to customers is.


If Intel ships a processor line that runs hotter than they said it would, the vendors who were in line to buy it are STILL going to buy it! They’ll just beef up their temperature control hardware to compensate. Intel missing their performance targets will still yield their expected sales because it’s not like those customers are going anywhere else. AMD STILL can’t meet Intel’s production volumes.


Yes, and what Intel “aims” and what they “do” don’t have to align as long as they can physically make more chips than AMD. What else are folks that need x86 solutions going to buy? I can predict now that they won’t be as performant and won’t run as cool as expected, but Intel will still ship a lot of them because ANY vendor that says, “No, this doesn’t suit me, I’m going to wait it out” is going to have marketshare taken by those that just alter their hardware to suit and ship it anyway.

The new 7040 series portable series with AMD XDNA Neural engine coming to market this quarter closes the gap and more so.

Lumping AMD with Intel is a fool’s errand.

 

[rm5]

This is super disappointing... I didn't read through this whole thread, but the lack of expansion and support for third-party graphics cards is going to make this product look very bad. Like I've said before—what is a "pro" machine that doesn't have multiple PCI-e slots, upgradable storage, upgradable RAM and GPU? Look! Even consumer Windows machines have this. You're already paying at least $6,000 for the Mac Pro and you're getting NOTHING? This makes no sense to me. It's gonna be the Trashcan all over again, but worse.

This makes me question the value of the Mac. Really, only the consumer Macs have good value, especially MacBooks. Even the Mac Studio is a pretty good deal. But you could build a PC that's at least 10x more powerful than the Mac Pro for $6000. Who cares about it being a "bad plastic case" and drawing a lot of power? That's irrelevant to me.

 

[Unregistered 4U]

The whole premise of a computer is for it to grow to your needs.

Maybe at one point in the distant past. The majority of light mobile computers that folks are actually purchasing has no allowances for ‘growing to needs’ except for if the OS does more or a new app that does a thing someone needs comes along.

 

[Unregistered 4U]

It’s pretty obvious this product line is essentially DOA since the business case for it has been wiped out. Almost everyone who needs serious power moved long ago to windows/Linux workstations which support the latest in CPUs and GPUs and scientific libraries or runs all their intensive workflows on cloud instances.

The folks using Final Cut Pro and Logic Pro are using windows/Linux workstations! That’s pretty clever of them!!

 

[Unregistered 4U]

Except that Intel's and AMD's low end CPUs keep improving as well.

Nowhere near as much as Apple’s. Again, it’s not that AMD/Intel absolutely couldn’t make it so that their lowest end chips match Apple’s lowest end’s performance. They’re in a different business and it just helps them to round out their bottom line when they sell large numbers of chips that, today, can’t come close to the M1 from the first MBA.

You mean the performance/Watt curve. But that will change too as Intel catches up to TSMC with their manufacturing process, since that is the main factor that drives the power efficiency of Apple's CPUs. "Meteor Lake" later this year (first Intel CPU that is manufactured using EUV lithography) should significantly narrow the power efficiency gap.

Catching up with a process node doesn’t magically rid Intel of the cruft of x86. No one should think that a decode heavy power hungry processor is ever going to be anything but decode heavy and power hungry. AND in reality Intel doesn’t HAVE to be very efficient. It’s not like folks that need an Intel processor are going to buy anything else but an Intel processor, regardless of how poorly it performs when it finally ships.

And yet they are currently losing market share to AMD.

If Intel were to stop making processors altogether, TODAY, AMD could not take up all the slack. Losing market share to AMD is a completely different thing from AMD being able to significantly take advantage of Intel selling a power hungry hot processor that outstrips its “roadmapped” TDP. Intel doesn’t have much to worry about at the mass market level.

 

[Realityck]

Maybe at one point in the distant past. The majority of light mobile computers that folks are actually purchasing has no allowances for ‘growing to needs’ except for if the OS does more or a new app that does a thing someone needs comes along.

I don't think our opinions change with how we purchase more capable Macs compared to light mobile crowd. It's still an investment to something you can use in the best possible way, be it consumer or enterprise.

This topic title is misleading also. "Mac Pro Enthusiasts Raise Concerns Over Upgrade Limitations of Apple Silicon" just comes across as an odd description of potential buyers. Who buys a computer or workstation from the standpoint of an enthusiast, it's not like you're going to show it off, admire it a lot, take it to shows?

Look at the definition of enthusiast

1. One who is filled with enthusiasm; one who is ardently absorbed in an interest or pursuit.
2. A zealot; a fanatic.
3. One who imagines he has special or supernatural converse with God, or that he is divinely instructed or commissioned.

Just the wrong terms to use, maybe potential or returning Mac Pro buyers have concerns. Yes I think the title of the thread was used for attracting views.

 

[Unregistered 4U]

The new 7040 series portable series with AMD XDNA Neural engine coming to market this quarter closes the gap and more so.

Lumping AMD with Intel is a fool’s errand.

They’re both x86, both trying to work around an outdated infrastructure that requires them to pump more power in in a way that doesn’t even provide a proportionate performance increase. PLUS, while they tout their performance as compared to Apple Silicon (who ever thought that would be a thing that AMD would even tout!!), they know Apple’s not their competition. Meeting the performance/power efficiency of Apple doesn’t matter to them because it’s not like Windows users have an option of using Apple Silicon.