References to Intel's Latest Ice Lake SP Processors in Xcode 13 Beta Hint at Forthcoming Mac Pro Update

[Yebubbleman]

Here's a thought that just occurred to me: What if the smaller size of Mac Pro is the Apple Silicon version and the larger one is the Intel version rumored here and what if they're getting launched simultaneously? Those whose software is already native and good to go on Apple Silicon would buy the Apple Silicon model while those that still need Intel for dependability/reliability with apps and plugins that aren't yet Apple Silicon native would buy the Intel one. There is now a Mac Pro for everyone.

 

[chucker23n1]

Here's a thought that just occurred to me: What if the smaller size of Mac Pro is the Apple Silicon version and the larger one is the Intel version rumored here and what if they're getting launched simultaneously?

Yup, that's my thinking as well. They do one last speedbump to the Intel Mac Pro and at the same time, introduce the all-new ARM Mac Pro.

(It still leaves questions for me, though. Why have so many different desktop Macs when desktops are now a niche? Why, then, not have more laptop Macs?)

 

[chucker23n1]

And I firmly believe the original 2013 Mac Pro intended on replaceable components, but Apple realized AMD's and NVIDIA's GPU roadmap didn't allow for lower powered GPUs, so they axed the idea halfway through. I don't think Apple designed the removable cover just to add new RAM. So you can't really rule out PCIe 5.0 slots. A 4x PCIe 5.0 slot equates to PCIe 3.0 16x speeds. Maybe 8x can fit.

The thing is, I never understood why the 2013 Mac Pro came with two GPUs.

Like, I'm sure there are areas where that's a useful approach, but is it really the overwhelming majority of potential Mac Pro buyers? What about software developers? They have little use for one discrete GPU, let alone two.

Was there a point in the design phase where they thought you'd also be able to configure it with 2 CPUs and one 1 GPU?

 

[Miha_v]

"Mini Pro" looks like a nice option for somebody, who doesn't need portability but wants power. I'm still skeptical about the graphics performance though - for consumers, integrated Apple graphics may be just fine, for professional use it's a different story (3D, animation & motion graphics related projects, etc).

 

[chucker23n1]

"Mini Pro" looks like a nice option for somebody, who doesn't need portability but wants power. I'm still skeptical about the graphics performance though - for consumers, integrated Apple graphics may be just fine, for professional use it's a different story (3D, animation & motion graphics related projects, etc).

The thing about GPUs is, their workload tends to be heavily parallelizable. Throw more cores at it and it'll be faster (in a way that doesn't work for the majority of CPU-bound code). So the question isn't so much 'can Apple make a GPU that's faster than AMD's and Nvidia's' — they absolutely can. Anyone could. Just add cores. The question is: can they do so while being more energy-efficient than them.

 

[MacFan23]

So 1 year into the "2-year Apple Silicon Transition" we've gotten one chip and a rumor of updating the top end Mac with Intel. Ok.

How many chips do you want them to produce each year? 🤔

 

[eVasilis]

• what will they name that half-size model? I'm guessing both are named Mac Pro, just like multiple different sizes and tiers are named MacBook Pro (sigh).

Mini Me Pro.

 

[chucker23n1]

So 1 year into the "2-year Apple Silicon Transition" we've gotten one chip and a rumor of updating the top end Mac with Intel. Ok.

No competing laptop chip comes close, so I think they're doing fine.

 

[farewelwilliams]

The thing is, I never understood why the 2013 Mac Pro came with two GPUs.

Like, I'm sure there are areas where that's a useful approach, but is it really the overwhelming majority of potential Mac Pro buyers? What about software developers? They have little use for one discrete GPU, let alone two.

Was there a point in the design phase where they thought you'd also be able to configure it with 2 CPUs and one 1 GPU?

Because of GPGPU programming/OpenCL. You can accelerate many non-graphics tasks using GPUs via Open CL so Apple bet big on that. Lots of Apple's frameworks were accelerated by OpenCL. But then many apps were optimized for NVIDIA'S CUDA so OpenCL adoption was low. Then Apple dropped OpenCL for Metal.

 

[chucker23n1]

Because of GPGPU programming/OpenCL. You can accelerate many non-graphics tasks using GPUs via Open CL so Apple bet big on that. Lots of Apple's frameworks were accelerated by OpenCL. But then many apps were optimized for NVIDIA'S CUDA so OpenCL adoption was low. Then Apple dropped OpenCL for Metal.

Sure, in areas like analytical science…

But again, there's got to be a lot of people who buy a Mac Pro who do neither GPU nor GPGPU stuff much at all.

 

[farewelwilliams]

Me personally? Nothing. I'm not a Mac Pro customer that needs PCIe expansion on a Mac. But Thunderbolt 3/4 to PCIe breakout boxes aren't cheap even if the bandwidth is sufficient. I know it wasn't sufficient with Thunderbolt 2. I'm not sure that Thunderbolt 3/4 is sufficient for EVERY PCIe card that would go into a Mac Pro even if it is for most. Either way, forcing users to buy breakout boxes, let alone expensive ones, is not going to make customers happy.



Too early to tell how that will be handled. For all we know, Apple may supply GPU booster cards a la Afterburner for those that need more than the SoC provides. They're not going to make the GPU a one or two size fits all the way they're doing graphics on all of the M1 Macs released to date. You're correct that the GPUs are the main things to command the full PCIe x16 slots. But even so, my point above about breakout boxes still stands.



I think you're 2/3 right about that. Certainly, if Apple wanted to, they could work with AMD on a GPU that was still newer than the AMD Fire Pro D300-D700 but not much faster for the sake of continuing to pump out Mac Pros. It's not like newer GPUs/cards couldn't provide the same performance at lower power draw at the very worst. But yeah, changing to a proprietary board design and interconnect was probably a bad move on Apple's part.


No, they also fully intended the SSD to be user replacable and upgradable. They still didn't give the 2013 Mac Pro enough room for other cards. The smaller of the two rumored Mac Pros could absolutely be what you're describing, but I'm totally skeptical that the 2013 Mac Pro had PCIe expansion (that wasn't rooted in Thunderbolt 2 expansion to a breakout box) potential.

My Thunderbolt EGPU box was only $150. Not quite expensive. And many Thunderbolt devices don't need a breakout box.

I don't think SSD+RAM warranted a full cover removal.

And I'm thinking Apple intended cards to be a special size/shape considering the proprietary connection to the base and the proprietary cooling that required cards to be positioned in a certain way. Also perhaps they didn't want dangling PCI-e powercables connected to the GPU so they decided a proprietary connection would delivery proper power and data.

 

[Yebubbleman]

(It still leaves questions for me, though. Why have so many different desktop Macs when desktops are now a niche? Why, then, not have more laptop Macs?)

The Mac Pro is not designed for your average desktop user. Non-workstation desktops are now a niche by comparison to how popular laptops are. But workstations are no less important now to the people they've always been important to than they were in 2007 when the Mac Pro evolved out of the Power Mac G5. And the need for an Intel AND an Apple Silicon version is so that users who are good to go on Apple Silicon can go all the way with Apple Silicon and those that can't or shouldn't move until their third party software or hardware is natively supported can still buy a beast of an Intel Mac. Those Ice Lake-SP Xeons sound like beasts.

The thing is, I never understood why the 2013 Mac Pro came with two GPUs.

Like, I'm sure there are areas where that's a useful approach, but is it really the overwhelming majority of potential Mac Pro buyers? What about software developers? They have little use for one discrete GPU, let alone two.

People who work with video. To my knowledge, they didn't have them working together in a CrossFire configuration. But even then, you have one GPU handling display output while the second GPU is focused squarely on rendering. At least, that was the point in having two non-CrossFire-linked GPUs in the 2010/2012 Mac Pros.

Was there a point in the design phase where they thought you'd also be able to configure it with 2 CPUs and one 1 GPU?

I strongly doubt it, but that would've been a great idea. Though having to choose between two GPUs and one CPU or one GPU and two CPUs without any other option would be a silly corner to put themselves in that only further draws attention to how silly it is. What if you need two of both?

Though, I still don't get why, there weren't two SSD slots, given that this predates the T2 and given that the SSD slot is on one of the GPUs. Just give it to me on both so I can put TWO PCIe-based SSDs instead of just one. The damn thing only maxed out at 1TB from Apple, which is paltry for a system that used to feature four SATA slots. Though, I get that the idea was external expansion was the primary mode of expansion.

 

[UBS28]

“Taking over” with 8.9% is adorable.

If you think a little bit, you will see that Intel is loosing market share and AMD is growing. And that is especially impressive since servers are not upgraded that frequently like the consumer market.

AMD increased it's market share by almost 2%, and however you want to twist it, Intel is loosing ground to AMD.

What makes this even more impressive, AMD is actually supply constraint. If AMD could deliver all their units, they would have taken more of Intel their sales.

 

[pmau]

The GPU support from Apple is so minimal that I cannot consider buying any MacPro.
It is really a shame. Lots of people will buy the M1 based Macs, but just as casual devices.
If you need anything you can reliably reconfigure to your needs, you are stuck with "PC" components.

Except iPhone and maybe a MacBook from time to time, there is very little insentive to invest in Apple Hardware.
I like iOS, I enjoy iPads and iPhones, but for my work I have very different requirements than what Apple haas top offer.

 

[Stephen.R]

If you think a little bit, you will see that Intel is loosing market share and AMD is growing. And that is especially impressive since servers are not upgraded that frequently like the consumer market.

AMD increased it's market share by almost 2%, and however you want to twist it, Intel is loosing ground to AMD.

What makes this even more impressive, AMD is actually supply constraint. If AMD could deliver all their units, they would have taken of Intel their sales.

Realistically the major threat to Intel CPUs in the data centre, is ARM.

That doesn't mean We're going to see ARM as a majority next year, but that's where things are likely to head, if for no other reason than better energy efficiency and less x86 cruft.

For the vast majority of data centre usage, the choice between AMD or Intel practically meaningless.

 

[farewelwilliams]

Sure, in areas like analytical science…

But again, there's got to be a lot of people who buy a Mac Pro who do neither GPU nor GPGPU stuff much at all.

Not really. OpenCL was adopted in some of Apple's common frameworks. Lots of apps used Grand Central Dispatch and that was accelerated by OpenCL. All users benefitted in some way. So Apple bet that most software would benefit from OpenCL, therefore dual GPUs seemed like a good idea at the time.

 

[Yebubbleman]

My Thunderbolt EGPU box was only $150. Not quite expensive.

"Only $150" compared to "included". Consider that multiplied by however many cards one needs attached to their Mac Pro and it becomes a lot of money that you're spending just to not have PCIe slots.

And many Thunderbolt devices don't need a breakout box.

Many Thunderbolt devices don't need a breakout box, but we're not talking about Thunderbolt devices. We're talking about $3000 AJA Kona cards and things of that nature. Thunderbolt breakout boxes are not the solution there. That was the entire idea behind the 2019 Mac Pro to begin with!

I don't think SSD+RAM warranted a full cover removal.

How else would you have wanted to access the SSD and RAM? A panel? Also, the Mac Pro is not the kind of computer to use SO-DIMMS either.

And I'm thinking Apple intended cards to be a special size/shape considering the proprietary connection to the base and the proprietary cooling that required cards to be positioned in a certain way. Also perhaps they didn't want dangling PCI-e powercables connected to the GPU so they decided a proprietary connection would delivery proper power and data.

I don't think they ever intended on standardizing the slot that the video cards used. If they did, we'd have seen aftermarket cards for it by now. I do believe that they customized the form factor of the cards to fit in that enclosure. But I don't believe they ever intended on third party boards to be produced for that Mac Pro the way that there are third party graphics cards for PCIe-based Mac Pros.

 

[farewelwilliams]

"Only $150" compared to "included". Consider that multiplied by however many cards one needs attached to their Mac Pro and it becomes a lot of money that you're spending just to not have PCIe slots.



Many Thunderbolt devices don't need a breakout box, but we're not talking about Thunderbolt devices. We're talking about $3000 AJA Kona cards and things of that nature. Thunderbolt breakout boxes are not the solution there. That was the entire idea behind the 2019 Mac Pro to begin with!



How else would you have wanted to access the SSD and RAM? A panel? Also, the Mac Pro is not the kind of computer to use SO-DIMMS either.


I don't think they ever intended on standardizing the slot that the video cards used. If they did, we'd have seen aftermarket cards for it by now. I do believe that they customized the form factor of the cards to fit in that enclosure. But I don't believe they ever intended on third party boards to be produced for that Mac Pro the way that there are third party graphics cards for PCIe-based Mac Pros.

You’re contradicting yourself a bit. If you need to connect a $3k card, a $150 breakout box is not a huge obstacle.

The point was that manufacturers would build a Thunderbolt variant of the card.

Yes, a panel or panels.

Not a standardized slot, but more of like Made for Mac Pro program that’s similar to what Apple is doing with Lightning. It wouldn’t make sense for nvidia to create a hot card that would harm the performance of other components. So Apple would have to set rules for cards.

 

[chucker23n1]

The Mac Pro is not designed for your average desktop user. Non-workstation desktops are now a niche by comparison to how popular laptops are.

Well, exactly. The Mac Pro is a niche within a niche. Why have two models of that?

(Maybe there's a lot of feedback that, actually, some people really did prefer the 2013 form factor, so now they want to please both crowds?)

But workstations are no less important now to the people they've always been important to than they were in 2007 when the Mac Pro evolved out of the Power Mac G5.

Sure they are. For a lot of those use cases, smaller machines now have plenty performance, or you rent stuff in the cloud. Workstations are a much rarer breed than they used to be.

People who work with video.

Yes, I get that. But what about others? If you're, say, a developer, why would you buy a computer that has one CPU and two GPUs? It's a bizarre setup for that.

If you think a little bit, you will see that Intel is loosing market share and AMD is growing. And that is especially impressive since servers are not upgraded that frequently like the consumer market.

AMD increased it's market share by almost 2%, and however you want to twist it, Intel is loosing ground to AMD.

What makes this even more impressive, AMD is actually supply constraint. If AMD could deliver all their units, they would have taken more of Intel their sales.

There's a huge leap from "AMD increased its market share" (you mean 2 percent points, not 2 percent) and your original assertion of "AMD is taking over the server market.

Yes, they're gaining, because Zen is a lot more interesting than Intel's 14nm+++++++++++++, but they're nowhere near taking anything over, and now that Ice Lake is coming, and stuff like ARM Neoverse is appearing, I don't think that's a foregone conclusion at all.

Not really. OpenCL was adopted in some of Apple's common frameworks. Lots of apps used Grand Central Dispatch and that was accelerated by OpenCL. All users benefitted in some way. So Apple bet that most software would benefit from OpenCL, therefore dual GPUs seemed like a good idea at the time.

Pretty bad bet.

Did they try and then abandon making clang use GPGPU? Did they try and abandon making JavaScriptCore use GPGPU?

 

[trash eighty]

Well its good news if Apple are still releasing a new Intel Mac in 2021 as it means macOS support for Intel Macs will continue for awhile yet...

 

[Icelus]

OS - macOS Tahoe 26

macOS Tahoe 26 with a new design, more ways to work seamlessly across devices, and new features to turbocharge productivity every day.

www.apple.com

This lists, in the footnotes, some dictation features being M1-only.

Also interesting is a feature requiring at least 16 GB memory (while there are 8 GB M1 systems sold).

 

[farewelwilliams]

Pretty bad bet.

Did they try and then abandon making clang use GPGPU? Did they try and abandon making JavaScriptCore use GPGPU?

probably doesn't make sense as GPGPU mainly benefits on tasks involving SIMD.

 

[chucker23n1]

Also interesting is a feature requiring at least 16 GB memory (while there are 8 GB M1 systems sold).

Yeah, that's not great.

probably doesn't make sense as GPGPU mainly benefits on tasks involving SIMD.

Well, that's what I'm saying. Lots of people who might have bought the 2013 Mac Pro if it had been more of a general-purpose high-end machine.

 

[Stephen.R]

a feature requiring at least 16 GB memory

Yeah, that's not great.

It does make for some pretty great popcorn-time what with all the people who parroted "buying more memory is useless, the 8gb baseline is enough, future versions aren't going to increase that any time soon, 'future-proofing' is a waste of money" and "M1 just doesn't need more memory, it's so efficient, 8GB is plenty for most people and 16GB is more than enough for <insert task they have zero actual knowledge about>".

 

[hunapu]

Why do they write "about half the size " and then show a fake image 1/4 the size?