Audio DSP cards, capture cards, PCIe NVMe SSD RAID stuff, high-speed networking stuff, weird scientific instrument interface stuff, etc.; they might still want slots, the edge connector allows for an optional expansion chassis...
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.
Concerns About an Apple Silicon Mac Pro
- Thread starter RedTheReader
- Start date
- Sort by reaction score
Audio DSP cards, capture cards, PCIe NVMe SSD RAID stuff, high-speed networking stuff, weird scientific instrument interface stuff, etc.; they might still want slots, the edge connector allows for an optional expansion chassis...
From reading various summaries as well as Nvidia’s letter to UK regulators, the US has clauses on the book that stops companies from using their market power to foreclose on their competitors ability to access essential tools needed to compete. More so in the UK and EU as my impression is that they also rely less on proving immediate quantitative harm to consumers. However an actual lawyer could probably chime in here. @cmaier?
The whole reason Apple went back to a tower design was having expansion slots. Aside from strange fixations on having another cube model or some unpractical technical marvel, there’s zero reason to expect Apple to have zero expansion to a new Mac Pro.
Limited? Maybe, but likely more due to expansion issues with Apple Silicon rather than desire to repeat the Mac Pro 6,1 mistake.
If I were gonna bet, I’d imagine that the INTERNAL expansion slots would be pci compatible a-la the mpx slots, and the machines would have an option of “afterburner” cards that have more gpu cores, or even cpu cores, to expand upon the base processor.
Also, this completely ignores the reality that graphical expansion isn’t the only use for pci slots. AVID cards for music production, and fiber optic network cards readily come to mind.
I think (hope) that Apple will be pragmatic about the upcoming Mac Pro, rather than repeat the issues with the 6,1 again.
Limited? Maybe, but likely more due to expansion issues with Apple Silicon rather than desire to repeat the Mac Pro 6,1 mistake.
If I were gonna bet, I’d imagine that the INTERNAL expansion slots would be pci compatible a-la the mpx slots, and the machines would have an option of “afterburner” cards that have more gpu cores, or even cpu cores, to expand upon the base processor.
Also, this completely ignores the reality that graphical expansion isn’t the only use for pci slots. AVID cards for music production, and fiber optic network cards readily come to mind.
I think (hope) that Apple will be pragmatic about the upcoming Mac Pro, rather than repeat the issues with the 6,1 again.
There is hardly any doubt that Nvidia is "the horse guarding the wheat" in this scenario, all they need to do is to angle Nvidia gpu`s to be superior in ARM. Which is in their best interest. And that effectively excludes all other gpus on the platform, and pushes all android to either use nvidia cpu/gpu or gpu. Is essentially a grab for the whole mobile market.
If this doesn`t get shut down in antitrust then nothing will.
If this doesn`t get shut down in antitrust then nothing will.
I doubt that Apple would make that mistake again so soon after fixing it with the latest Mac Pro. I'm also of the opinion that Apple will ship MPX GPU cards for the ASi Mac Pro for the same reason, the existing Xeon Mac Pro can have many multiple GPUs. I can't imagine that Apple won't have some ASi solution that is at least equivalent to the Xeon Mac Pro even if Apple only supports their own GPUs.
How much of a hit would we be looking at for off package RAM and external MPX GPU's? That and how many PCIe lanes does AS support now? Everything you guys talk about adding internally could be done via Thunderbolt, so really there is no reason to keep the existing MP form factor and they could go back to the "Trashcan" design.
I think the closest upcoming “Mac Pro Mini” might be a different beast. It might have internal expansion but won’t necessarily be as modular as the bigger variant.
I suspect a bigger more modular AS Mac Pro will eventually arrive BUUUTTT …. I’m not sure about discrete GPUs - at least not in the traditional sense. I think Apple is going to push hard on unified memory. So there may still be options here for internal modularity but it may not look like what we’re used to.
Even if Thunderbolt had enough bandwidth, for Apple to go back to that kind of design is just asking for ridicule. For Thunderbolt 4 at 32 GB/s vs x16 PCIe 5 at 63 GB/s also is going to make a pretty big difference.
I think that Apple is going to take a hybrid approach. Unified memory for the first 64-128 GPU cores and then x16 PCIe 5 for any add ons. PCIe isn't going to give 200-400 GB/s but more GPU cores at 63 GB/s is still better than nothing. Apple is much more clever than I am and I trust that they will solve this in a way that lets their customers with crazy GPU requirements get the most out of the Pro. Just for reference, the current Mac Pro has about 140 GB/s memory bandwidth and still uses x16 PCIe 3 @ 32 GB/s for GPUs.
Because PCI slots have much higher bandwidth than thunderbolt for starters?
Purely speculation on my part, but I believe in the same way the mpx modules are compatible with pci, there will be more connectors on the slots which allows the on-board M1 (pro or max) to directly access any cores on the module. Or maybe there will be distinct modules for that purpose.
In short: built in processor, goes through special interconnect (which we know exists on the m1 max) to extra processing cores or accelerators on expansion card.
Could be … though in the current MP the GPU has its own onboard ram so you can hide latency/bw issues better. I’ve seen others postulate full “PC Card” solutions where an individual MP can be turned into a mini-cluster. Which is interesting.
For me I’m keeping an open mind. So far Apple has surprised me, but this “SOC in a modular workstation” issue is a tough nut to crack - a lot of competing philosophies that don’t play well. For the smaller AS Pro, it’s more straightforward - a couple of (4x? 8x?) PCIe for cards and some thunderbolt enclosures and you can attach all things in @JMacHack ‘s list while still being a AS SOC. But the bigger machine? Huge RAM pools and dGPU is a more difficult problem and you don’t want to run into a “worst of both worlds” rather than “best of both worlds”.
Last edited:
I don’t think we’ll see a huge pool of RAM in the upcoming MP. Apple will likely handwave it saying that the memory is fast enough that you don’t need it.
Like I said, I’m confident that some internal expansion will be available, if limited.
And it’s not difficult to imagine ways that an SoC could be expanded upon that can match a dedicated gpu system. My example I believe is the most likely way, some dedicated direct access bus to essentially “dumb” cores. I don’t believe RAM will be expandible.
Re: large pools ram. I don’t think so either for the smaller MP. If a bigger one comes around … maybe. Dunno.
I also believe the replacement for 7,1 (not the 'cube', if it even is a thing) will continue to have expansion.
There were a couple of reports saying that it would half/a third the size of 7,1. People shouldn't be fooled by how large that machine actually is - it's a beast when you see it in person. Half of that is still a very large desktop computer, and if you consider that MPX, the large power supply and the three fans are all going to be reduced in size or gone, then you easily have space for regular PCIE cards.
There were a couple of reports saying that it would half/a third the size of 7,1. People shouldn't be fooled by how large that machine actually is - it's a beast when you see it in person. Half of that is still a very large desktop computer, and if you consider that MPX, the large power supply and the three fans are all going to be reduced in size or gone, then you easily have space for regular PCIE cards.
AT&T acquiring T-Mobile would be hard to do. That's because they offer the exact same service, which could actually decrease competition and come close to a monopoly.
How is Nvidia acquiring ARM, which is a strategic acquisition to increase Nvidia's CPU capabilities and combine IP remotely the same? Nvidia and ARM do not offer the same products.
I'm still waiting for your law that says Nvidia can't acquire ARM.
Concerns About an Apple Silicon Mac Pro
From reading various summaries as well as Nvidia’s letter to UK regulators, the US has clauses on the book that stops companies from using their market power to foreclose on their competitors ability to access essential tools needed to compete. More so in the UK and EU as my impression is that...
forums.macrumors.com
The laws being cited by the FTC relate to foreclosure on competition. The UK is considering the same objection as we all as others.
You can read the complaint here:
Nvidia/Arm, In the Matter of
The Federal Trade Commission filed a law enforcement action to block U.S. semiconductor chip supplier Nvidia Corp.’s $40 billion acquisition of UK-based semiconductor design firm Arm Ltd., the largest transaction in the history of the semiconductor industry. The FTC’s action seeks to preserve...
It’ll have all of the relevant information on the basis for the suit.
Last edited:
And yet T-Mobile was allowed to acquire Sprint. There were conditions placed upon T-Mobile however in order for the deal to be approved.
As others have pointed out, Nvidia does not need to acquire ARM in order to improve their CPU capabilities. Apple has improved theirs without acquiring ARM. Apple avoided even making a bid because they knew the regulatory hurdles would be difficult, if not impossible, to overcome. In addition to providing CPU designs, ARM provides GPU designs as well.
There is no law that specifically names Nvidia or ARM, but you know that. These types of deals are subject to regulatory oversight in almost every jurisdiction. The UK and EU have also expressed reservations about Nvidia acquiring ARM. If Nvidia can convince all of these regulatory bodies, and be willing to make some concessions to allay their concerns, then the deal will likely be approved. If they are unable to come to an agreement, the deal with fall through. I've given you a few examples where that has happened.
You seem to be confused.
There's a difference between actual laws that prevent the acquisition of this type and approval from regulatory bodies.
Decisions from regulatory bodies are opinionated, dependent on the current regime, and can be lobbied by other companies. Decisions from regulatory bodies can also be sued and challenged.
Laws are different. No law exists that says Nvidia can't acquire ARM.
- Nvidia needs its own world-class CPU team to compete with SoCs designed by Apple, Qualcomm, Mediatek, AMD, Intel, etc. Nvidia also thinks it can accelerate ARM's competitiveness in other devices besides low-powered ones.
- Yes, Nvidia doesn't need to buy ARM to design ARM-based CPUs. But Nvidia thinks it can help accelerate ARM's adoption in devices other than low-powered devices, improve its designs by combining IPs, inject resources into the company. Remember that ARM doesn't actually make that much revenue or profit despite dominating mobile devices.
- A standalone ARM does not actually have enough money and resources to continuously compete with Apple, AMD, Intel designs. And it shows. Apple designs are far ahead of ARM designs on mobile. Intel and AMD designs are significantly more powerful in non-mobile devices (albeit at a lower efficiency). For context, ARM made around $2 billion in revenue in 2021. Intel $78 billion. AMD $15 billion. Apple $365 billion.
- ARM has been too slow to get into laptop, desktop, workstation, server, and supercomputer market. This is where I think Nvidia can help accelerate a lot and add a 3rd option to Intel/AMD duopoly. Only Apple has managed to successfully bring ARM to high power computers.
- Nvidia + ARM adds one more competitor to the laptop, desktop, server, supercomputer landscape. We've seen a duopoly by AMD/Intel for far too long which set innovation back a decade in the 2010s.
- Nvidia has been losing mega government contracts to AMD/Intel because they can't supply CPUs.
- Nvidia would like to control their entire hardware stack for server solutions. Right now, they have to use Epyc/Xeon.
- Even their upcoming Grace CPU simply uses stock Neoverse cores
- The future of mobile devices (maybe even desktops) is clearly a SoC/APU approach. This leaves Nvidia vulnerable because they don't make SoCs/APUs. This is a strategic acquisition for Nvidia.
- It's not fair to ARM employees and shareholders to block this deal. Blocking this deal benefits Qualcomm, AMD, Intel shareholders. You're just choosing one side over the other by blocking this deal.
- ARM has contracts with existing customers. Even if Nvidia buys them, they can't just deny existing customers with licenses.
- Nvidia is too smart to ruin ARM's business model. They're not going to spend $40b just to destroy all of ARM's revenue overnight.
- Qualcomm has clearly spent a lot of lobbying money to convince the US government to block this deal. It's actually weird that the US government would block an American company from acquiring important IP from a British/Japanese company. I'm guessing Qualcomm has bought off the right politicians.
- I think this deal benefits consumers, adds more competition. But unfortunately, there are too many myths and politics in this deal.
Right, there are laws governing mergers and acquisitions. They don't prevent them at all, but they are subject to approval. While the regulatory agencies do have some latitude, the laws create the criteria to determine what criteria an agency can or cannot use to decide if an acquisition can go through. The decisions can be challenged in court, but even the courts must apply the laws as written. This isn't up to the whims of the FTC, they have guidelines they must follow.
Apple does not own ARM. Yet, they have a world class CPU team that has made their own custom designs that implement the ARM ISA. There is nothing preventing Nvidia from doing the same. Your first point is moot. Qualcomm recently made another acquisition to do what Apple is doing - and they didn't need to own ARM to do it.
The problem with Nvidia + ARM is that it gives Nvidia the potential to stamp out AMD + ARM, or Qualcomm + ARM. In fact, Qualcomm has objected to the acquisition, they don't believe it will benefit them. Google and Microsoft have expressed concerns as well. It isn't just US companies or the US government who have reservations about this deal.
You are correct in that so far ARM has been primarily used in mobile. However, Apple isn't the only vendor selling ARM-based laptops. They're the largest, and they've made the biggest commitment, but they are not the only ones. Likewise, others are bringing ARM to the server space. The ARM ecosystem is doing just fine without Nvidia owning ARM.
In order to make fair revenue comparisons, you would need to separate out the revenue generated by Apple's chip division rather than the whole company. Apple is a juggernaut, and is really several Fortune 100 companies operating under the same organization. Those billions in revenue also include iPhone and iPad sales (which far exceed Mac sales), accessories like AirPods, HomePods and the watch, and of course the fastest growth is in services revenue.
But ultimately it isn't up to me or you. Like it or not, business has gone global, so several governments need to sign off. As I have mentioned, the UK and EU have also expressed reservations, and those must be addressed before the deal can go through. Your own personal approval is neither sought nor required, but if you'd like to contact your representatives and the FTC on Nvidia's behalf, I'm sure they would be grateful for your endorsement.
Apple does not own ARM. Yet, they have a world class CPU team that has made their own custom designs that implement the ARM ISA. There is nothing preventing Nvidia from doing the same. Your first point is moot. Qualcomm recently made another acquisition to do what Apple is doing - and they didn't need to own ARM to do it.
The problem with Nvidia + ARM is that it gives Nvidia the potential to stamp out AMD + ARM, or Qualcomm + ARM. In fact, Qualcomm has objected to the acquisition, they don't believe it will benefit them. Google and Microsoft have expressed concerns as well. It isn't just US companies or the US government who have reservations about this deal.
You are correct in that so far ARM has been primarily used in mobile. However, Apple isn't the only vendor selling ARM-based laptops. They're the largest, and they've made the biggest commitment, but they are not the only ones. Likewise, others are bringing ARM to the server space. The ARM ecosystem is doing just fine without Nvidia owning ARM.
In order to make fair revenue comparisons, you would need to separate out the revenue generated by Apple's chip division rather than the whole company. Apple is a juggernaut, and is really several Fortune 100 companies operating under the same organization. Those billions in revenue also include iPhone and iPad sales (which far exceed Mac sales), accessories like AirPods, HomePods and the watch, and of course the fastest growth is in services revenue.
But ultimately it isn't up to me or you. Like it or not, business has gone global, so several governments need to sign off. As I have mentioned, the UK and EU have also expressed reservations, and those must be addressed before the deal can go through. Your own personal approval is neither sought nor required, but if you'd like to contact your representatives and the FTC on Nvidia's behalf, I'm sure they would be grateful for your endorsement.
I'm glad we're able to distinguish between laws and the approval/disapproval process of government agencies. At first, you claimed that laws exist everywhere that would prevent Nvidia from acquiring ARM. This is simply not true. Nvidia clearly didn't think they will break any law before planning to acquire ARM, because no law exists.
Of course Qualcomm objects to the acquisition.
Qualcomm has just invested heavily into Nuvia. They plan to build custom ARM cores for everything, starting with laptops. If stock ARM cores, boosted by Nvidia IP and resources, perform as well as Nuvia cores then Qualcomm can't differentiate. Qualcomm wants to be the Apple of ARM designs for Android phones and laptops.
Qualcomm has the most at stake here. They've clearly bought off the right politicians.
Let's not use Qualcomm, Microsoft, Google as stalwarts of competitive behaviors. Look up the dictionary for anticompetitive and you might find Qualcomm, Microsoft, and Google on it.
Nvidia acquiring ARM does not decrease competition. It increases it and accelerates ARM's adoption in things other than mobile phones.
Nvidia + ARM adds a real challenger to Apple, AMD, Intel, Qualcomm, Mediatek, Samsung SoCs, CPUs, GPUs.
Of course Qualcomm objects to the acquisition.
Qualcomm has just invested heavily into Nuvia. They plan to build custom ARM cores for everything, starting with laptops. If stock ARM cores, boosted by Nvidia IP and resources, perform as well as Nuvia cores then Qualcomm can't differentiate. Qualcomm wants to be the Apple of ARM designs for Android phones and laptops.
Qualcomm has the most at stake here. They've clearly bought off the right politicians.
Let's not use Qualcomm, Microsoft, Google as stalwarts of competitive behaviors. Look up the dictionary for anticompetitive and you might find Qualcomm, Microsoft, and Google on it.
Nvidia acquiring ARM does not decrease competition. It increases it and accelerates ARM's adoption in things other than mobile phones.
Nvidia + ARM adds a real challenger to Apple, AMD, Intel, Qualcomm, Mediatek, Samsung SoCs, CPUs, GPUs.
Last edited:
I fully acknowledge that my opinion about the Nvidia and ARM deal isn't a popular opinion on the internet. But to me, it makes perfect sense for both companies. Nvidia can help ARM and vice versa.
ARM is an architecture or a specification that anyone could license. NVIDIA has very low creditability to manage that legacy. They cannot avoid existing contracts but they have no obligation to honour hew ones. Combining that with NVIDIAs low creditability of having common standards means that there will be a risk that it will be NVIDIA only property. If NVIDIA had licensed CUDA to AMD and Intel and who ever else from start, at least I would see differently on NVIDIA.
Key architectures should be in IP pools for anyone to use under a licence. Look at what x86 brought us. Very expensive CPUs with very little progression between generation. That is the result of no real competition in the field.
Na, we don't need Nvidia to help ARM. Raspberry Pi, Amazon and Apple have enough momentum. And Nvidia is join gin the ARM bandwagon regardless of whether it owns ARM or not.
Back to the original topic, it will be pretty impressive if Apple can pull off having chip designs that span everything from the W and S-series for wearables, the A-series for phones and tablets, M-series for laptops/desktops, and manage to make powerful workstation chips. That would be quite the portfolio. I would imagine some of the competition is quietly pleased that Apple only uses their chips for their own products.
I would very much agree. There’s no one else offering the level of performance at the features and levels of efficiency Apple is. Apple was available in quantity on the open market, Intel’s i3, i5 and i7 lines would lose to Apple Silicon easily (even accounting for a Windows/Linux version of Rosetta).