Mac Pro Enthusiasts Raise Concerns Over Upgrade Limitations of Apple Silicon

[Detnator]

So you pretty much just admit that you had no point to make. You don't really have to work in a corporation to know the difference between planning the demise of your own product and planning for contingencies. In fact, these two different types of planning are handled by two different departments. Either you're just really confused or purposely trying to conflate the two to make a non-point.

Or you completely missed the point.

 

[Boil]

Folks talking about a $50K ASi Mac Pro...

• Base M1 Ultra Mac Studio - $4K (binned SoC/64GB/1TB)
• Loaded M1 Ultra Mac Studio - 8K (full-die SoC/128GB/ 8TB)
• Base M2 Ultra Mac Pro - $6K (binned SoC/96GB/1TB)
• Loaded M2 Ultra Mac Pro - $10K (full-die SoC/192GB/8TB)

With the 2019 Intel Mac Pro, it is the BTO options that drive the price into the stratosphere...

• Xeon CPU - up to $7K
• RAM - up to $25K
• SSD - up to $2.4K
• MPX GPUs - up to $11.4K
• Afterburner card - $2K

That is $47.8K in BTO options for a fully loaded 2019 Intel Mac Pro...

Even if one figures in two ASi GPGPUs (say $5K each), a fully loaded ASi Mac Pro would be $20K...

Add in a third-party 64TB RAID card (eight 8TB M.2 NVMe SSDs), a 8K video I/O card, and two audio DSP cards and one would still be under $35K...!

And a hypothetical M3 Extreme Mac Pro (full-die SoC/384GB/8TB/dual GPGPUs/64TB RAID/8K video/dual audio DSP cards) would be about $45K; still under $50K...!

 

[AdamNC]

Could they use just the M2 Ultra chip, and have separated ram, SSD, and video card slots?

 

[adamw]

Could they use just the M2 Ultra chip, and have separated ram, SSD, and video card slots?

I am sure Apple could do that, with a lot of design time and money spent, but since Apple decided to go with the System On A Chip (SOC) route with the M-series of processors, it is unlikely they would give up the Media Encoders, Neural Engine, built in GPU cores, and Unified Memory speed gains, by separating out all of the components, and having slower buses built between each of the components to move data around.

 

[PauloSera]

This.
Some people just want to ignore some facts and would rather prefer Apple to go back. I don't know but I suspect people that want to add or remove RAM and GPUs are Intel employees.

I'd rather they stay focused on their current path, and wait for the top end of Apple Silicon to improve to become realistic options for other top end configs.

With the current lineup including the Mac Studio, and a theoretical 2x Mac Studio (called Mac Pro), they will be hitting more than 95% of the Mac market. Maybe 98%.

There is no reason for them to completely change focus/direction to appease the smallest Mac market. Those people will whine and complain loudly (because the thing they want is macOS, they don't care about the hardware), but they are not a big market.

 

[Joe The Dragon]

they have ssd sockets on systems.
Now apple can lower the price of the upgrades or they can do an you must get 4 sticks for max speed and really jack up the base price.

Or just link up 4 pci-e lanes to an m.2 slot so that you can get your own SSD.

 

[avkills]

I think that there’s a lot of people that just fundamentally don’t understand the SoC. I’d guess they don’t even understand that “graphics over a PCIe bus” would be going backwards because “that’s what everyone else is doing”.

Perhaps, but my PCIe bus graphics are still spanking the hell out of Apple Silicon graphics overall.

 

[NT1440]

Perhaps, but my PCIe bus graphics are still spanking the hell out of Apple Silicon graphics overall.

How long do you think they can keep cranking the power and heat requirements before they hit a wall?

 

[avkills]

How long do you think they can keep cranking the power and heat requirements before they hit a wall?

How do we know that Apple can match performance without cranking the power and heat up to the same level?

 

[Joe Dohn]

How do we know that Apple can match performance without cranking the power and heat up to the same level?

I would like to hear that too.

 

[CMMChris]

In addition to that, they could still do something with normal memory slots. They could treat the unified memory as a cache, or a faster portion of RAM, but then add standard RAM that may be a bit slower to cover the need for more RAM.

Uh? What kind of benefit would that have? Would be a huge bottleneck with near to no benefit compared to plain old swaping on an ultra-fast SSD. Plus, it would totally destroy any optimization attempts with regard to the UMA architecture.

 

[NT1440]

How do we know that Apple can match performance without cranking the power and heat up to the same level?

Apple has headroom to double their power and heat and still be at a moderate draw….

I’m not sure you’ve though your response to me.

 

[BenRacicot]

Hmmm... Gurman's statements are a little weird...

Keep the current Intel Mac Pro case, but don't support anything that necessitates that type of case!? No memory slots, no PCIe slots, no 3rd party GPU support. Huh? So that entire case is for "expandable storage" and a "better cooling system".

1. The cooling system in the Studio is more than enough to keep it cool even under a continuously heavy load. so there's no reason you'd need the Mac Pro case just to keep the M2 Ultra cool enough to get full performance.

2. While I can believe the M2 Extreme may have been scrapped, it would be a huge misstep to release a freaken Mac Pro with just an M2 Ultra. They can afford to wait until the end of the year for the M3 Ultra/Extreme.

3. Apple was well into development of their M-series SoCs when the Mac Pro debuted. They would've known that eventually those SoCs would end up in the Mac Pro and planned accordingly; expandable memory, PCI slots, 3rd party GPU support, etc. They designed that case for all of those reasons.

Yes! It is strange isnt it? Wasn’t his report corroborated by other analysts too?

Anyways, yes. If all that is true then the only thing they could do is overclock the chips. That’s it.

 

[DeepIn2U]

ARM windows will be a Chromebook competitor. By the time Snapdragon gets to be anything close to useful as laptop replacement Intel will be using state of the art ASML machinery which will prevent Qulacomm to play in anything but bottom feeder laptops.

You’re mixing up EUL Lithography with the type of processor architecture - two completely different things I think. Btw all cpu manufacturers use the same lithography, how it’s applied and what their guiding differs.

 

[foo2]

Uh? What kind of benefit would that have? Would be a huge bottleneck with near to no benefit compared to plain old swaping on an ultra-fast SSD. Plus, it would totally destroy any optimization attempts with regard to the UMA architecture.

IDK what RAM he has in mind, but whatever it is, it's massively faster than even the fastest SSDs; Apple's stating 100GB/s on the low-end M2 up to 800GB/s on the M1 Ultra. Even half to 1/3 that (M2) speed (typical for DDR4 in DC config) is PDQ.

An SSD is 6GB/s, on the most expensive Mac configuration, and 1.5GB/s on the least. That's crawling - at least, compared to RAM speeds. Doesn't matter if it's 6000MB/s or 1500MB/s, it's still _crawling_ compared to RAM. Sure, you might not notice it much, but it's still a slowdown. OSs have been getting better at making swapfile / virtual memory usage better for decades, but it's still not as fast as RAM.

You want to avoid swapping, even to an 'ultra-fast ssd', all that you can. To do so, get RAM. Or better, figure out what RAM you need, then get said RAM.

 

[vladi]

You’re mixing up EUL Lithography with the type of processor architecture - two completely different things I think. Btw all cpu manufacturers use the same lithography, how it’s applied and what their guiding differs.

I am not mixing anything. I am saying that with Intel's new radical chip design which will use state of the art lithography, if it turns out to be successful and delivered on time, consumer ARM soc designers such as Qualcomm will not be able to fight for anything but bottom feeder PC laptops. To put it this way, can Qualcomm design a decent performance chip to enter Windows PC market? Sure it can but question is when? Sources are estimating 2024. They will need all the Microsoft's help they can get in order to get traction in both tablet and laptop market. I wish them luck and I want them to succeed because in the end we as consumers all win. After all if Windows on ARM becomes a relevant thing it means bootcamp will be back in no time.

To make it less confusing let's put it this way. If Intel's x86 chip redesign in 2025 pays off and if AMD grows with linear improvement and scalability we will finally witness the x86 efficiency breakthrough in PC mobility market on par with current cellphones and iPads. Even Apple will have a hard time keeping up with it let a lone Qualcomm but at this point it's all one big speculation. Much like it is Intel's upcoming server Xeon with 512 cores or workstation Xeon with 144 cores which will sip only 425W. Let's wait and see.

 

[Imhotep397]

...Threadripper based Mac Pros would have been nirvana - and would have sold a lot more than the 7,1. They would have sold even more, if Apple had signed off on the Nvidia drivers for both Turing and Ampere cards.
....

Video production using Apple only software.

Everybody else was abandoned - I am one of those folks - I do 3d art, and I finally threw in the towel when the 7,1 was launched. My Ryzen system outperforms the base 7,1 for less than 1/3 of the price.

Can you imagine Sony with a reborn Vaio brand selling EPYC and Threadripper based workstations, Sony Vaio EPYC Server racks with Nvidia drivers running MacOS? I feel confident Sony would be willing to buy back the Vaio company/brand if Apple suggested this kind of plan that ended with Apple Silicon working with AMD/Nvidia GPUs under MacOS Viao computers.

 

[Nugget]

Can you imagine Sony with a reborn Vaio brand selling EPYC and Threadripper based workstations, Sony Vaio EPYC Server racks with Nvidia drivers running MacOS? I feel confident Sony would be willing to buy back the Vaio company/brand if Apple suggested this kind of plan that ended with Apple Silicon working with AMD/Nvidia GPUs under MacOS Viao computers.

I'm familiar with the term "fever dream" but I don't know if there's a corresponding "fever nightmare." Not sure how else to describe this.

 

[Imhotep397]

I really don't understand the negative reaction these Mac Pro rumours are getting...

Back when Apple introduced the last Intel based MacPro they said that the those would stay on Intel for Pro users for quite a while. It's not unreasonable for users to believe that meant more than one iteration. It's also not unreasonable for Pro users expect an affordable option that didn't cost over 10x+ the price of the average Mac Pros prior to the last one.

 

[Imhotep397]

I'm familiar with the term "fever dream" but I don't know if there's a corresponding "fever nightmare." Not sure how else to describe this.

In what way is this unfavorable for Pro 3D users preferring to use MacOS? Do you work in any of those areas for a living?

 

[foo2]

Can you imagine Sony with a reborn Vaio brand selling EPYC and Threadripper based workstations, Sony Vaio EPYC Server racks with Nvidia drivers running MacOS? I feel confident Sony would be willing to buy back the Vaio company/brand if Apple suggested this kind of plan that ended with Apple Silicon working with AMD/Nvidia GPUs under MacOS Viao computers.

People certainly have made MacOS machines (with AMD graphics, largely) and AMD Threadripper hardware, running $currentMacOS. It's very doable. Sure, Intel VT-D's absence means virtualization is out the window, but it's so fast that emulation isn't that bad. I do hear of some issues with some Adobe product. But generally the other stuff runs fine.

Reddit's Hackintosh reddit would have more info.

 

[foo2]

IDK what RAM he has in mind, but whatever it is, it's massively faster than even the fastest SSDs; Apple's stating 100GB/s on the low-end M2 up to 800GB/s on the M1 Ultra. Even half to 1/3 that (M2) speed (typical for DDR4 in DC config) is PDQ.

An SSD is 6GB/s, on the most expensive Mac configuration, and 1.5GB/s on the least. That's crawling - at least, compared to RAM speeds. Doesn't matter if it's 6000MB/s or 1500MB/s, it's still _crawling_ compared to RAM. Sure, you might not notice it much, but it's still a slowdown. OSs have been getting better at making swapfile / virtual memory usage better for decades, but it's still not as fast as RAM.

You want to avoid swapping, even to an 'ultra-fast ssd', all that you can. To do so, get RAM. Or better, figure out what RAM you need, then get said RAM.

I should also note that when paging, the SSD will get more like Random IO QD1 speeds, so perhaps 50-80MB/s tops. Contrast that with 100GB/s RAM speeds.

 

[DavidSchaub]

Back when Apple introduced the last Intel based MacPro they said that the those would stay on Intel for Pro users for quite a while. It's not unreasonable for users to believe that meant more than one iteration. It's also not unreasonable for Pro users expect an affordable option that didn't cost over 10x+ the price of the average Mac Pros prior to the last one.

Perhaps, but Apple hasn't done an "iteration" of a Pro desktop since 2012, so it seem strange to expect that now. That statement was pre "public" Apple Silicon, so now there is a new plan. What Apple says never matters, just what they do.

 

[DeepIn2U]

I am not mixing anything. I am saying that with Intel's new radical chip design which will use state of the art lithography, if it turns out to be successful and delivered on time, consumer ARM soc designers such as Qualcomm will not be able to fight for anything but bottom feeder PC laptops. To put it this way, can Qualcomm design a decent performance chip to enter Windows PC market? Sure it can but question is when? Sources are estimating 2024. They will need all the Microsoft's help they can get in order to get traction in both tablet and laptop market. I wish them luck and I want them to succeed because in the end we as consumers all win. After all if Windows on ARM becomes a relevant thing it means bootcamp will be back in no time.

To make it less confusing let's put it this way. If Intel's x86 chip redesign in 2025 pays off and if AMD grows with linear improvement and scalability we will finally witness the x86 efficiency breakthrough in PC mobility market on par with current cellphones and iPads. Even Apple will have a hard time keeping up with it let a lone Qualcomm but at this point it's all one big speculation. Much like it is Intel's upcoming server Xeon with 512 cores or workstation Xeon with 144 cores which will sip only 425W. Let's wait and see.

So ... with Microsoft Surface Pro X was not a Performance ARM chip created by Qualcomm? Not tracking success but it did perfect decently.

Now onto the Surface Pro 10, with Microsoft's one-up behind the scenes tweak it's slowly getting to a decent Cor-i3 performance level. Another year we'll see but I'm thinking it'll take another 4yrs.

Does intel have ANY decent road map to compete with TSMC's N5 or N3 design process?

 

[foo2]

So ... with Microsoft Surface Pro X was not a Performance ARM chip created by Qualcomm? Not tracking success but it did perfect decently.

Now onto the Surface Pro 10, with Microsoft's one-up behind the scenes tweak it's slowly getting to a decent Cor-i3 performance level. Another year we'll see but I'm thinking it'll take another 4yrs.

Does intel have ANY decent road map to compete with TSMC's N5 or N3 design process?

Apple's Mac M1 chip just humiliated the Surface Pro X

How fast is Apple's Mac M1 chip? It can run Windows faster than Microsoft's own Surface Pro X

www.tomsguide.com

So far results for the Microsoft chip aren't particularly encouraging even compared to the last-generation and slower old M1 chips, but Microsoft does nothing if not iterate.