MP All Models Everyone's complaining about the Mac Pro's 192 GB RAM limit and no video card support. Do older MPs with 1.5 TB RAM outperform it?

[Boil]

People want more memory, and more powerful GPUs in Mac Pro. Apple cannot provide those under Apple Silicon architecture. Whether as-is is enough for some professional workflows matter less in the argument. The new Mac Pro architecture is much less flexible than the old one. Not to mention limited PCIe bandwidth, non upgradable RAM, and lack of dGPUs as lots of people have been talking about but to me is of a lesser concern in my judgement being a failure.

But if Apple gives the end user upgrade paths for CPU(ASi)/RAM/GPU(ASi)/SSD, and allows lower cost third-party options on the RAM & SSD; then the argument would be that they don't allow third-party GPUs (where even if they allowed AMD, folks would whinge about Nvidia) or cannot dual boot into Windows...

Apple could move to LPDDR5X RAM for the M3 Ultra / M3 Extreme ASi Macs (Studio/Pro/Cube) which should allow up to 1TB of capacity...? Oh dookie, forgot Apple will have to spec ECC RAM or their hardware is useless...!

Apple is done making PCs, it's all Apple Compute Appliances from here on out...! ;^p

 

[kvic]

Apple is done making PCs, it's all Apple Compute Appliances from here on out...! ;^p

"Apple Compute Appliances" as in lack of flexibility of resource allocation of your money.

I should have elaborated a bit more the lack of felxibilty in Apple Silicon Architecture. Memory, CPU and GPU are essenatially scaling together in Apple Silicons, which is far from ideal for workstations. Workstations are expensive investment. You want to allocate money at hardware that benefit your ROI effectively. So some people may want less CPU cores, more GPU cores. Some may want more CPU core and more memory, but less GPU cores. Now I hope make myself clear. I pointed this out years earlier in this sub-forum. The new Apple Silicon Mac Pro doesn't give users such choices.

By that, if you call them "Apple Compute Appliances", I'm in full agreement with you.

 

[h9826790]

What Is the Difference Between Intel® Optane™ Memory and an...

Explains the difference between Intel® Optane™ memory and a solid-state drive (SSD).

www.intel.com

“While a solid-state drive (SSD) is a device used for data storage, Intel® Optane™ memory is a system acceleration solution installed between the processor and slower storage devices (SATA HDD, SSHD, SSD), which enables the computer to store commonly used data and programs closer to the processor. This allows the system to access this information more quickly, which can improve overall system responsiveness.”

AFAIK, Optane memory's sequential speed is lower than most NVMe indeed. However, it has very high 4K random read performance (about 5x higher than NVMe). This is why it can accelerate the system. It has nothing related to that 8GB/s or 800GB/s. Even the Optane memory can only do ~220MB/s, most NVMe can only do ~50MB/s or lower.

Intel Optane Memory Review

Optane Memory, formally 3D XPoint, is utilizing the first new class of NAND in nearly…

www.storagereview.com

 

[mcnallym]

We are here because Apple built the 7,1 and gave it the ability to use other operating systems natively.

Do you have a 7,1 or predecessors?

I think reading more into apple and operating system support then there actually is and will explain why.

Mac on 68000 series, whilst third parties developed ways of booting Linux on Macs with 68020 or later then did Apple provide anything.

PowerPC Macs, again what support for other OS did Apple provide, certainly no boot camp for windows equiv.

intel era - Apple provided boot camp to allow easy windows installation with native unpatched windows. apple moved to Intel as powerpc not providing the processors they wanted, not because it allowed them to provide support for windows OS or other x86 based OS. If apple really wanted multi os support then would have used standard efi rather then own modified version allowing standard hardware and OS as if a standard PC. basically other OS support came as a consequence of move to Intel not by a conscious choice to move to Intel to provide multiple os support.

asi era - dropping of boot camp though people have made it work with windows in parallels. Not seen anyone with a native raw windows arm installation. (However willing to be shown to be wrong). No reason why if apple wamted they could not have worked with Microsoft to get windows arm booting natively if they wanted to provide such support. Am aware of providing support to ashashi (think spelt right) Linux people but hardly a mainstream os amongst non-technical users.

Apple didn’t deliberately give other os support on the x86 pc’s it just happened to inherit the ability though using x86 system and was not difficult to do but not a driving factor when designing a Mac in how easy to boot another os.

you even have to go in and enable the ability to boot off another drive on the t2 equipped Mac’s so putting a small road bump in doing so.

and yes I used to own a 5,1 and had a t2 mini 2018 model previously.
replaced the 5,1 with a hack when outgrew the 5,1 - 6core model, and now to studio with sonnet enclosure for extra nvme ssd.

other then the Intel era and even then apple made some differences to make people work at it, then has not been forthcoming support from apple, so ASi era simply more a return to when Apple running non generic wintel hardware such as 68000/PowerPC.

I don’t see how apple deliberately gave the multiple os support, is just more, a oh as we moved to Intel then we can do boot camp for windows if want to run windows as well as a happy coincidence of moving to Intel rather then we want to provide this feature. With non standard efi and t2 lockdown then wasn’t that important too them.

 

[Longplays]

All of which can use external PCIe enclosures with the Mac Studio. Personally I run 3 x16 PCIe TB3 enclosures, 1 pcie x4 pcie enclosure and 2 thunderbay 4s. Fiber and internal NVMe storage are the only PCIe express cards that can take advantage of the performance.

That is why there is a Mac Studio at $3k less between those M2 Ultra Macs.

Want less than a M2 Ultra? Then a M2 Max saves you $5k from a Mac Pro base.

For a production studio that wants something in one box, the Mac Pro is a convenient one stop box solution and

Thunderbolt 4 caps at 40Gbps. PCIe 4.0 slots offers more than that.

tax writeoff.

This is what the most vocal MR user that complains about the $1k hike do not understand.

It is a business expense. Charge it to the client that in turn becomes a tax shield.

Unless of course you do not pay your fair share of taxes.

 

[Longplays]

intel era - Apple provided boot camp to allow easy windows installation with native unpatched windows. apple moved to Intel as powerpc not providing the processors they wanted, not because it allowed them to provide support for windows OS or other x86 based OS. If apple really wanted multi os support then would have used standard efi rather then own modified version allowing standard hardware and OS as if a standard PC. basically other OS support came as a consequence of move to Intel not by a conscious choice to move to Intel to provide multiple os support.

Either in 2006 or 2007 there was an online contest for the development of "bootcamp". It ended when Apple announced their 1st party solution.

It wasn't part of Apple's original plan. They just did not want a 3rd party making one that would eventually lead to malicious code and bad press to the company.

asi era - dropping of boot camp though people have made it work with windows in parallels. Not seen anyone with a native raw windows arm installation. (However willing to be shown to be wrong). No reason why if apple wamted they could not have worked with Microsoft to get windows arm booting natively if they wanted to provide such support. Am aware of providing support to ashashi (think spelt right) Linux people but hardly a mainstream os amongst non-technical users.

Windows in Parallels is different from cold booting as it isn't direct to hardware but through a layer.

I think Microsoft wasn't interested. It means R&D cost with little or no licensing fees. How many of the 28.6 million Macs shipped in 2022 would buy Windows 11 on Apple Silicon? Would it be sufficient to cover cost?

Windows on Intel Mac had near zero cost to Microsoft as it would run on Intel/AMD/Nvidia hardware.

 

[Longplays]

But if Apple gives the end user upgrade paths for CPU(ASi)/RAM/GPU(ASi)/SSD, and allows lower cost third-party options on the RAM & SSD; then the argument would be that they don't allow third-party GPUs (where even if they allowed AMD, folks would whinge about Nvidia) or cannot dual boot into Windows...

There is a profitable niche and the unprofitable niche.

Apple kept a Mac Studio with PCIe 4.0 slots because the number of users it is profitable.

Apple removed the swappable parts because it was not profitable and the loss of business from them is a net gain.

Apple could move to LPDDR5X RAM for the M3 Ultra / M3 Extreme ASi Macs (Studio/Pro/Cube) which should allow up to 1TB of capacity...? Oh dookie, forgot Apple will have to spec ECC RAM or their hardware is useless...!

Apple is done making PCs, it's all Apple Compute Appliances from here on out...! ;^p

In 2022 Apple shipped 28.6 million Macs worldwide.

If the Mac Pro were ~75,000, ~15,000 or even ~7,500 units shipped annually would it be worth their time to not be "Apple Compute Appliances"?

Everyone's use case is valid... even the Nvidia Quadro ones but you may be too few for Apple to bother with.

The rumored M3 Extreme failed probably due to yields so they couldn't deliver 384GB unified memory. This would have reduced the number of niche user complaints.

Try again 21 months from now in Q1 2025.

 

[Longplays]

For those who may not be aware...

AMD announces Instinct MI300X GPU with 192GB of HBM3 memory

AMD announces Instinct MI300X GPU with 192GB of HBM3 memory - VideoCardz.com

AMD Expands Leadership Data Center Portfolio with New EPYC CPUs and Shares Details on Next-Generation AMD Instinct Accelerator and Software Enablement for Generative AI — AMD unleashes the power of specialized compute for the data center with new AMD EPYC processors for cloud native and...

videocardz.com

 

[anselpela]

For very specific use cases swap or paging in are unacceptable due to audio errors.

That's why this fella is keeping his 2019 at 768GB memory until a future Ultra or Extreme chip will allow for that much RAM.

For your convenience I fast forwarded to his concern with the 192GB RAM limit

Sounds like his wildly inefficient workflow needs some improvement.

 

[Longplays]

Sounds like his wildly inefficient workflow needs some improvement.

Or maybe he knows how it works and others do not.

There are very valid use cases for more than a 2019 or 2023. It just costs Apple more than it is worth now.

Maybe by Q1 2025 they'll align RAM with a 2019. That's a reasonable 6 year gap.

 

[Boil]

If the Mac Pro were ~75,000, ~15,000 or even ~7,500 units shipped annually would it be worth their time to not be "Apple Compute Appliances"?

Would just be a ultra-high-end compute appliance...! Not a bad thing, just no upgrade path(s)...

When you get a new compute appliance you can trade in the old one or start your own ad hoc compute/render farm...

The rumored M3 Extreme failed probably due to yields so they couldn't deliver 384GB unified memory. This would have reduced the number of niche user complaints.

Yields have nothing to do with it, the Ultras (and theoretical Extremes) would all be built using Mn Max SoCs...

Probably more down to issues with the interconnects for a four-way system...

And entry level for a Mn Extreme Mac Pro would probably be US$12K...

 

[Longplays]

And entry level for a Mn Extreme Mac Pro would probably be US$12K...

It's a tax deductible business expense. 🤷‍♂️

$12k is likely a unbinned M3 Extreme with more than 384GB max RAM

 

[Boil]

It's a tax deductible business expense. 🤷‍♂️

Which is fine for business purchases, "freelancers" might be a little put off by the six figure price tag...

$12k is likely a unbinned M3 Extreme with more than 384GB max RAM

$7K gets one an ASi Mac Pro with a binned (GPU cores) Mn Ultra SoC and 64GB of RAM...

Add another four grand for the upgrade to a binned (GPU cores) Mn Extreme SoC and 128GB of RAM, and another thousand for the interconnect; US$12K starting cost...

Full-die Mn Extreme SoC, 384GB RAM, and 8TB SSD; about $18K or so...?

And if Apple makes the smart move by bumping the ASi Mac Pro up to LPDDR5X RAM, then the cost may nearly double for a variant with 1TB of RAM...

 

[Longplays]

Which is fine for business purchases, "freelancers" might be a little put off by the six figure price tag...

Freelancers should pay their fair share of taxes.

No ifs or buts about it.

Also, they have the option of Ultra, Max, Pro and plain vanilla M.

Mac Pros are predominately bought for business and not a hobbyist's basement activity.

$7K gets one an ASi Mac Pro with a binned (GPU cores) Mn Ultra SoC and 64GB of RAM...

Add another four grand for the upgrade to a binned (GPU cores) Mn Extreme SoC and 128GB of RAM, and another thousand for the interconnect; US$12K starting cost...

Full-die Mn Extreme SoC, 384GB RAM, and 8TB SSD; about $18K or so...?

And if Apple makes the smart move by bumping the ASi Mac Pro up to LPDDR5X RAM, then the cost may nearly double for a variant with 1TB of RAM...

tax
deductible
business
expense

 

[gnomeisland]

There are basically two kinds of high-memory workloads. Sometimes the workload would not technically need that much memory, but it's more cost-effective to tell the users to buy more RAM than to spend the development effort to make everything work well with less RAM. And sometimes the memory is actually necessary, because the workload is based on random accesses to large amounts of data. In the latter, memory latency is often the most important performance measure.

With large amounts of data, RAM latency is measured in hundreds of nanoseconds. SSD latency, no matter whether we are talking about NVMe drives or Optane, is in hundreds of microseconds. Once the system starts swapping, it becomes unusably slow for the latter kind of high-memory workloads.

If you have a PCIe card full of DIMMs and use it for swapping, it might be possible to get the latency down to microseconds. That would probably be good enough for the former kind of workloads, but not for the latter.

Yeah, I can't say. It appears that PCIe 4.0 has a native latency of 250n/s where as quality RAM is under 10n/s—add in swap over head and you might stay in the nanosecond range, but barely.

AFAIK there is no modern PCIe "RAM drive" so I certainly can't say how well it work. Still I think it is an interesting an idea that I think might become "new" again if it would work for many/most Mac Pro users who need more than 192GB.

 

[mcnallym]

It is likely the failed to launch M2 Extreme would have a memory bandwidth of 1,600GB/s.

Making it a tad more than 200x faster.

That figure though would come from 4 x Max each at 400Gbs. I don't think it is a coincidence that

Max = 400Gbps
Ultra (2 x Max) = 800Gbps

I would say that each Max within the Ultra can access its RAM at 400Gbps and have 2 so 800Gbps

 

[Boil]

tax
deductible
business
expense

Cool, cool, cool; business expense...

The point I was trying to make is that $12K would get a binned Mn Extreme with 128GB of RAM, NOT a full-die Mn Extreme with over 384GB of RAM...

 

[anselpela]

Or maybe he knows how it works and others do not.

Or maybe he knows one very specific tunnel visioned old way of doing this and refuses to adapt.

 

[AndyMacAndMic]

Or maybe he knows one very specific tunnel visioned old way of doing this and refuses to adapt.

Do you have any experience in the field where Neil Parfitt is working? Because, as far as I know, he certainly knows what he is talking about and he is a highly regarded professional in his field. Unless you know much more than he does of course.....

 

[Longplays]

Or maybe he knows one very specific tunnel visioned old way of doing this and refuses to adapt.

He makes a living out of this. He doesn't run a home lab to nerd out.

Many, not all, who have a cow with the Mac Pro's price and limits tend to be hobbyists who are on a $3k budget and just want a i9 & 4090.

 

[impulse462]

He makes a living out of this. He doesn't run a home lab to nerd out.

Many, not all, who have a cow with the Mac Pro's price and limits tend to be hobbyists who are on a $3k budget and just want a i9 & 4090.

I don't think judging hobbyist or enthusiasts vs a "professional" is really warranted and it comes off as pretty condescending; especially when that Neil Parfitt guy works as a producer in the entertainment industry, an industry that is by definition recreational and not needed.

Why should anyone put more stock into what he wants out of a workstation vs anybody else? Other "professionals" such as researchers/scientists and engineers should take precedence and their needs should be catered to first since they are *more important* than a silly audio producer.

See? I can make the same argument against an entertainment industry "professional" as they make against hobbyists.

Apple fanboys/apologists love to **** on hackers or enthusiasts, forgetting that the original most famous successful product this company produced, the product that gave apple its name and reputation, the Apple II was entirely developed by a hacker/hobbyist/enthusiast.

 

[Longplays]

I don't think judging hobbyist or enthusiasts vs a "professional" is really warranted and it comes off as pretty condescending; especially when that Neil Parfitt guy works as a producer in the entertainment industry, an industry that is by definition recreational and not needed.

It isn't a judgement but highlighting product mismatch with end user.

Mac Pros are designed for persons who make a living out of their work. He gets paid for his audio work for TV/films/music. He isn't running a charity. He specifically says that he has contracts that paid for 2019. Once those contracts were completed he has been running on gravy with his 3.5yo Mac.

Enthusiasts & hobbyists do not depend on that machine for their livelihood.

Anyone can get a i9 & 4090 and a whole gaming rig for less than $7k. Want additional 16 GPU cores? That's another $1k! That's 2/3rds of a 4090!

Why force a fish to climb a tree and call it dumb with slightly lower gaming benchmarks? Get a chimp for less that tops those fps.

Why should anyone put more stock into what he wants out of a workstation vs anybody else? Other "professionals" such as researchers/scientists and engineers should take precedence and their needs should be catered to first since they are *more important* than a silly audio producer.

On the Mac Pro product page his industry is targeted for that Mac

• Music production
• Video transcoding
• 3D rendering
• Animation

If it was meant for non-revenue use like Mac gaming then it would be prominently featured on the Mac Pro product page.

Apple fanboys/apologists love to **** on hackers or enthusiasts, forgetting that the original most famous successful product this company produced, the product that gave apple its name and reputation, the Apple II was entirely developed by a hacker/hobbyist/enthusiast.

Check the white paper for the 2019 Mac Pro. It expounds for use cases for that specific year model by 2021.

Use case of the 1977 Apple II has changed nearly half a century ago that led to 2023 Mac Pro.

There is nothing wrong with the use case before but it does not align with this specific Mac product line.

The hacker today would likely use Linux and an Ryzen laptop.

Apologist is another word for someone who understands hows and whys something occurs and points out the hows and whys to those who do not.

If there is dwindling demand then shift to where the demand is strongest. Hence the 4 listed major use cases.

If the customer's too few you're better off doing something else.

 

[Yebubbleman]

This is typical party line that "unified memory" architecture fanatics have been parroting Apple marketing for more than two years by now. To the extreme, 8GB in Apple Silicon is equivalent to what..16GB in Intel Mac. LoL

First off, I have never and will never make the argument that 8GB of RAM in an Apple Silicon Mac is equivalent to 16GB in an Intel Mac. That's not even remotely what I'm saying.

Don't want to be arguing like you..'cos I can repeat that to refute whatever you said too. LoL.

I mean, if you wanna pick fights on the Internet, be my guest. That's not my intention here at all and I won't partake.

The new Mac Pro is a failure because Apple Silicon architecture doesn't scale up well, easily and flexibly, even from Apple's standpoint not just users.

Apple's standpoint is that this machine is an upgrade over its predecessor. Also that it's no less flexible than its predecessor. That's not my standpoint. I do not necessarily agree with them. But, everything they have said about this machine publicly indicates that they believe it's a viable replacement product for the things that customers of its predecessor used it for.

Furthermore, unless you have the machine in front of you, you're arguing tech specs over real-world performance. I'm not saying that this machine can or can't outperform a fully configured 2019 model. Just that, from Apple's point of view, more than 192GB of RAM isn't necessary and that expandable graphics are also not necessary.

Again, I do not agree or disagree with this sentiment because I do not have either Mac Pro in front of me.

People want more memory, and more powerful GPUs in Mac Pro.

Apple doesn't agree with you. Or at least, they find that the advantages of the unified architecture vast outweigh those desires.

Personally, I don't disagree with you. But I'd imagine, especially after the trash can fiasco, they did enough homework to make sure that the 2023 iteration could properly serve the needs of the vast majority of those who own 2019 models.

Apple cannot provide those under Apple Silicon architecture. Whether as-is is enough for some professional workflows matter less in the argument.

No, that IS the argument.

The new Mac Pro architecture is much less flexible than the old one. Not to mention limited PCIe bandwidth, non upgradable RAM, and lack of dGPUs as lots of people have been talking about but to me is of a lesser concern in my judgement being a failure.

It's a failure if it doesn't meet the needs of customers. If customers don't need more than what Apple offers, how is it a failure?

Again, I'm also skeptical of the lack of expandability here. Similarly, not socketing the SoC, at the very least, seems like a massive misstep for them. However, if one buys a 192GB/76-GPU-Core model, fills out every PCIe slot and everything works just fine with every need being met, what does it matter?

 

[kvic]

First off, I have never and will never ... [snipped]

It's getting boring and tiresome. Have a good day. Byte~

 

[orionquest]

Apple's standpoint is that this machine is an upgrade over its predecessor. Also that it's no less flexible than its predecessor. That's not my standpoint. I do not necessarily agree with them. But, everything they have said about this machine publicly indicates that they believe it's a viable replacement product for the things that customers of its predecessor used it for.

Furthermore, unless you have the machine in front of you, you're arguing tech specs over real-world performance. I'm not saying that this machine can or can't outperform a fully configured 2019 model. Just that, from Apple's point of view, more than 192GB of RAM isn't necessary and that expandable graphics are also not necessary.

Again, I do not agree or disagree with this sentiment because I do not have either Mac Pro in front of me.

Of course Apple is going to say "it' the best Mac Pro we ever created" because they want sell you a computer. Why does this matter?

You do not need the Mac Pro in your hands in order to judge it. The fact that it does not allow GPU cards makes it a failure. A year or 2 of ownership and the hot new GPU comes out which is blowing the doors off everything and you cannot use it in your "Pro" desktop is a failure. No need to purchase or see it in person to figure this out.