Mac Pro Enthusiasts Raise Concerns Over Upgrade Limitations of Apple Silicon

[Alan Wynn]

My current workstation is a 12-Core Ryzen 9 series with 64GB RAM and a current total of 16TB SSD storage (spread over 5 internal drives including two M.2 drives).

Curious why you chose the SSDs rather than get a PCIe multi-M.2 card and do it all with M.2 drives? My BF has the OWC Accelsior 8m2 for 16TB of internal M.2 in his Mac Pro and is connected to a 100ish TB server at 25Gb/s for external storage.

For my purposes, I need lots of CPU cores, lots of RAM, and lots of storage. GPU requirements are minimal (generally, I run upwards of 12-20 virtual machines at a time while developing enterprise systems and databases).

How much did you pay for your current system? Are you an independent developer or was it purchased by your company?

Apple made it clear when they released the Trash Can that they were no longer serious about serving the corporate professional market. I had high hopes for a good workstation solution when they announced development of the new Mac Pro, but was quite disappointed when the new Mac Pro came out.

I think you misread the situation. Apple looked at their statistics and saw that fewer than 1% of their users had any expansion cards (they said this publicly many times after the release of the machine) and built a machine that would serve them. They failed to understand two things:

1. That 1% absolutely needed that expansion and while with Thunderbolt 3/4 it is a bit more reasonable, one would really need the ability to support multiple x16 cards in an enclosure to serve them.
2. Many of the other 99% think they will eventually want expansion slots even if they never use them.

It is why people buy extended range electric vehicles, paying thousands extra for range they never (or almost never) use (and would be better off renting a car for those few occasions).

To use a tired old car analogy: Apple used to build a nice reliable pick-up truck with the original Mac Pro (and the Power Mac G4 and G5 before it). They decided to replace it with a zippy convertible with an oversized engine when they released the trash can and tried very hard to convince people that they can get the same thing they had before by adding a trailer hitch. When that strategy failed, they decided to replace that with an 18-Wheeler.

I like your analogy. I think the Mac Studio is a great machine for 99% of Mac Pro class users. For most of the rest of them, a Mac Studio with a bunch of PCIe slots that is priced to start around $6,000 and go up to $12,000 - $18,000 would cover the needs of most of the remaining 1%. There are a few who need (or think they do) 1.5TB of RAM, but they are such a small niche, that I am not sure it makes sense for Apple to try to serve them.

 

[Alan Wynn]

Apple proving that a ARM laptop (& desktop) works and can be profitable is emboldening Qualcomm and other ARM SoC vendors to be a bit more aggressive with their on ARM laptop offerings.

A friend used to own/run a high end location sound hardware company (they make audio recorders for film sets). He said that FPGAs changed the way the industry worked. In the old days a company developed high end gear and then used that research to build cheaper versions for the larger market. FPGAs made it so it was better to do the R&D for the mass market and then just build one's high end gear by using a bigger/faster version of the same FPGA and adding ports/features.

Apple, IBM and Motorola never had the volume to make PowerPC successful for the high end. Motorola's largest customer for their PPC chip was Ford and they needed neither high performance nor really low power. Intel by contrast was selling millions of computers and speed mattered to them.

Now Apple has a market of hundreds of millions of iPhones and enough money to make the incremental cost of developing higher performance SoCs reasonable.

Unlike Intel (and for that matter Qualcomm), they do not need to build a general purpose chip, but one targeted at their specific needs. They can design everything together at once. Apple gets to say: "What do we want to build?" rather than "What will commodity chips let us build?"

R&D from SoC vendors that has a combined annual worldwide shipment of >1 billion smartphone/tablets will put a monkey wrench into x86 laptop sales.

Yes, and laptop sales are still where the money is.

Scaling up smartphone SoC to laptop and then desktop size will push down x86's PC share to ~20% in less than 20 years.

Servers are the final scale up. It makes sense for someone to build a version that was perfectly balanced for cloud servers used by AWS, GCP, Meta, etc.

x86's sole strength would be legacy x86 hardware/software.

Not a tiny thing, but becoming less important over time. It also has Intel's marketing dollars.

Pity the PC gamer or gear head who will experience price increases as volume of x86 parts decreases and economies of scale suffers.

I think one will just see ARM versions of these games, so those people will buy ARM s

x86 by then may one day be as relevant as mainframes during Y2

 

[jakey rolling]

Curious why you chose the SSDs rather than get a PCIe multi-M.2 card and do it all with M.2 drives? My BF has the OWC Accelsior 8m2 for 16TB of internal M.2 in his Mac Pro and is connected to a 100ish TB server at 25Gb/s for external storage.

For two reasons:
1) The one 4TB SATA SSD I purchased was significantly cheaper than M.2 equivalent drives, and still provides the performance for my VMs that I need.
2) There were a couple of SATA SSD drives that were repurposed from a previous workstation that I wanted to continue to use, so I did so.

How much did you pay for your current system? Are you an independent developer or was it purchased by your company?

I paid just under $2,700 CAD for this system. Part of the reason I was able to purchase it so inexpensively was because I was able to repurpose some parts from a prior workstation (some SSDs, the GPU, PSU, and case). If I included the cost of parts brought from the prior machine (a workstation built in 2014 and then iteratively added to), that cost would be closer to $4,000.

I work for a company, however my home workstation is something I've always purchased myself because it was part of me exercising my WFH privilege. My company generally provides us with laptops and I do the majority of my development work at home now. My company-issued laptop is only used on the very rare occasion when I come into the office for meetings or staff events, or when I'm working at a client's site (in which case I tend to work directly off of the client's servers instead of my computer).

I think you misread the situation. Apple looked at their statistics and saw that fewer than 1% of their users had any expansion cards (they said this publicly many times after the release of the machine) and built a machine that would serve them. They failed to understand two things:

1. That 1% absolutely needed that expansion and while with Thunderbolt 3/4 it is a bit more reasonable, one would really need the ability to support multiple x16 cards in an enclosure to serve them.
2. Many of the other 99% think they will eventually want expansion slots even if they never use them.

I know we have focused a lot on PCI expansion when talking about workstations, but storage expansion and the ability to iteratively upgrade is a huge thing for users like myself as well. Even the original G5 / Mac Pro was a bit of disappointment because it only held two SATA drives compared to 4 drives on the G4 (made adding a RAID card quite difficult at a time when SSDs were still a fever dream for most people).

Apple has been trying to move people to external devices for many, many years dating back to the original G3. A lot of Avid users were quite disappointed when they found they had to upgrade their >$1000 devices because Apple dropped the number of PCI slots in their machines when the G3 came out. The Thunderbolt argument of today is the same as the Firewire argument back when people who had already invested tens of thousands of dollars in their workflow were now being told to "just upgrade to the .002"

 

[sam_dean]

A friend used to own/run a high end location sound hardware company (they make audio recorders for film sets). He said that FPGAs changed the way the industry worked. In the old days a company developed high end gear and then used that research to build cheaper versions for the larger market. FPGAs made it so it was better to do the R&D for the mass market and then just build one's high end gear by using a bigger/faster version of the same FPGA and adding ports/features.

I became familair with FGPAs in these 3rd party hardware-accurate NES/SNES/Genesis/GameBoy clones that had HDMI output. I got so tempted to buy one then I snapped out of it that I'd probably never used it for play.

Now Apple has a market of hundreds of millions of iPhones and enough money to make the incremental cost of developing higher performance SoCs reasonable.

Indeed... Apple SoC found in all their devices outnumber AMD/Intel x86 units shipped annually. They have the R&D money of both companies combined with Apple tax!

AMD/Intel PC excluding Macs

- 2021: 322.2 million units
- 2022: 263.7 million units

Macs, iPads & iPhones

- 2021: 321.5 million units
- 2022: 316.8 million units

Now Apple is just optimizing supply chain processes to squeeze out as much economies of scale as possible.

R&D-ing their own 5G/WiFi/Bluetooth chips and even their own displays so they can detach themselves from rivals like Qualcomm and Samsung.

IIRC a reason why Macs never got a 4G/5G modem was because of the way Qualcomm charges for the chips. They allegedly base the fees on the MSRP of the device. So if the device is $3.5k then the 5G modem price would be based on the % of it.

Source: https://www.macrumors.com/2011/08/1...macbook-pro-with-integrated-3g-cellular-data/

Doing so would reduce the economies of scale of their rivals and raise parts prices for everyone else.

Further differentiating themselves from everyone else.

Unlike Intel (and for that matter Qualcomm), they do not need to build a general purpose chip, but one targeted at their specific needs. They can design everything together at once. Apple gets to say: "What do we want to build?" rather than "What will commodity chips let us build?"

Thus never being pressured to compete on parts costs. Their systems will be so further removed from anyone else that an orange to orange comparison will become more and more difficult to do.

They are creating value that no other company could hope to achieve because an Android phone is a commodity. Windows PC is a commodity... etc etc

Yes, and laptop sales are still where the money is.

~80% of all PCs are laptops. Since as early as year 2005 when Steve Jobs wanted to put a G5 in a laptop and push 3GHz on a PowerMac G5.

Unachievable because the G5 was stuck at 90nm while Intel was on time with 65nm.

Servers are the final scale up. It makes sense for someone to build a version that was perfectly balanced for cloud servers used by AWS, GCP, Meta, etc.

I saw the numbers of servers and they're not as numerous as one could hope. To the point that Apple did not want to pursue it for Apple Silicon. Sad as their 3 ex-Apple Silicon engineer formed their own company to create ARM server chips. These were bought out by Qualcomm...

I think one will just see ARM versions of these games, so those people will buy ARM s

I am not talking about solely games but any old Windows x86 software. Odds are ARM laptops will not be 100% full speed accurate on x86 especially those old systems.

 

[Alan Wynn]

Apple has been trying to move people to external devices for many, many years

By the way, just because they were early does not mean they were wrong.

 

[Alan Wynn]

For two reasons:
1) The one 4TB SATA SSD I purchased was significantly cheaper than M.2 equivalent drives, and still provides the performance for my VMs that I need.
2) There were a couple of SATA SSD drives that were repurposed from a previous workstation that I wanted to continue to use, so I did so.

Makes sense. He moved from a 4m2 Thunderbolt connected drive to an PCIe connected internal 8m2 and added 5 more M.2 2TB drives.

I paid just under $2,700 CAD for this system.

That is what $25 US? (I have family in Canada and I remember when the Canadian dollar was about $1.25 USD).

Part of the reason I was able to purchase it so inexpensively was because I was able to repurpose some parts from a prior workstation (some SSDs, the GPU, PSU, and case). If I included the cost of parts brought from the prior machine (a workstation built in 2014 and then iteratively added to), that cost would be closer to $4,000.

A machine one builds oneself will almost always be cheaper than one that one buys, especially if one does not include the value of ones time (something completely reasonable if one enjoys doing it). When they are not cheaper, they are configured more appropriately.

I work for a company, however my home workstation is something I've always purchased myself because it was part of me exercising my WFH privilege. My company generally provides us with laptops and I do the majority of my development work at home now. My company-issued laptop is only used on the very rare occasion when I come into the office for meetings or staff events, or when I'm working at a client's site (in which case I tend to work directly off of the client's servers instead of my computer).

Curious if you could really do your work with your company provided laptop? I am a mostly a consultant these days. I used to have the top of the line Mac laptop and a Mac Pro desktop. Now I use an iPad Pro with a keyboard case and a Mac mini (that I hope to upgrade to an M2 Studio if they release one).

I found that I usually either work at a desk when I need a real computer (often attached to cloud resources), or I am mostly just taking notes/editing documents and using FaceTime/Zoom.

I know we have focused a lot on PCI expansion when talking about workstations, but storage expansion and the ability to iteratively upgrade is a huge thing for users like myself as well.

Yup, but users like you (corporate users who purchase their own machines and so do incremental updates) make up a tiny niche of a tiny niche. Now that 25Gb networking has gotten so cheap, I am not sure that any machine I ever buy again will have more than 2TB of internal storage. Unlike RAM and even GPU, I tend to want to keep my data and the speeds have not really gotten faster than even 10Gb/s Ethernet can handle.

Even the original G5 / Mac Pro was a bit of disappointment because it only held two SATA drives compared to 4 drives on the G4 (made adding a RAID card quite difficult at a time when SSDs were still a fever dream for most people).

Fixed in the Mac Pro.

Apple has been trying to move people to external devices for many, many years dating back to the original G3. A lot of Avid users were quite disappointed when they found they had to upgrade their >$1000 devices because Apple dropped the number of PCI slots in their machines when the G3 came out. The Thunderbolt argument of today is the same as the Firewire argument back when people who had already invested tens of thousands of dollars in their workflow were now being told to "just upgrade to the .002"

Once 10Gb/s Ethernet became reasonable, moving everything other than working storage out of a machine just makes more sense. Storage speed has not increased at the same rate other components have and storage requirements tend to be additive (I need to store everything I had and this new stuff too).

I have everything in a FreeBSD server that supports me and my BF. I connect to it at 10Gb/s he connects at 25Gb/s. That is now fast enough that he can do a proxy workflow and do his conforms off the server (it has 16TB of M.2 NVMe Storage as a read/write cache).

In your case, I am not sure why you would not do that at your next upgrade. Move the storage out of your office to your garage (or since you are in Canada, into your igloo) :-D and upgrade it as you need to do so, without having any impact on your desktop.

 

[DavidSchaub]

I know we have focused a lot on PCI expansion when talking about workstations, but storage expansion and the ability to iteratively upgrade is a huge thing for users like myself as well. Even the original G5 / Mac Pro was a bit of disappointment because it only held two SATA drives compared to 4 drives on the G4 (made adding a RAID card quite difficult at a time when SSDs were still a fever dream for most people).

Apple has been trying to move people to external devices for many, many years dating back to the original G3. A lot of Avid users were quite disappointed when they found they had to upgrade their >$1000 devices because Apple dropped the number of PCI slots in their machines when the G3 came out. The Thunderbolt argument of today is the same as the Firewire argument back when people who had already invested tens of thousands of dollars in their workflow were now being told to "just upgrade to the .002"

Isn't PCIe expansion and storage the same thing? The only internal storage that really matters these days are PCIe storage devices: either cards that act as SSDs, or cards that you can put a pile of PCIe based SSDs on to. For slow <= 600 MBps storage, external devices such as DAS and NAS are the right answer anyway?

Apple will always screw over the niche users if it feels it is worth it. That's just business as usual for Apple. They're a business, I can't really fault them for that.

If Apple really thinks it is worth building PCIe expandable computers, I would hope they do so consistently so board partners feel it is worth keeping up with drivers and support.

I think an interesting question will be the actual PCIe bandwidth on the Mac Pro. Will it be Gen4 or Gen5? How under-provisioned will the bridge chip be, given that the Mac Pro (2019) was already under-provisioned for PCIe?

At least TB4, and next gen's 80+ Gbps are OK speeds for most tasks. Apple is a company that still ships devices with 0.48 Gbps data throughput.

 

[Alan Wynn]

Isn't PCIe expansion and storage the same thing? The only internal storage that really matters these days are PCIe storage devices: either cards that act as SSDs, or cards that you can put a pile of PCIe based SSDs on to. For slow <= 600 MBps storage, external devices such as DAS and NAS are the right answer anyway?

With the inclusion of 10Gb/s Ethernet and the price of 25Gb/s Ethernet getting cheap, it certainly seems to be the right answer for all storage. Having a hybrid NAS (some NVMe cache and 100+TB of spinning disk), means we no longer really need much local storage (even for editing 6K Blackmagic RAW, where proxy workflows make more sense anyway).

That was not true when the 2013 machine was released, but it was getting very close.

Apple will always screw over the niche users if it feels it is worth it.

I do think this is a true or fair statement. I think that niche users cost a lot to support and do not have economies of scale. Their costs reflect that. AppleCare on a Mac Pro with 1.5TB of RAM costs no more than it does for a machine with the base configuration, but support costs are much higher. When one looks at the TCO with AppleCare, the costs are much more in line. If one does not buy support, one is probably overpaying.

If Apple really thinks it is worth building PCIe expandable computers, I would hope they do so consistently so board partners feel it is worth keeping up with drivers and support.

I think with Apple Silicon, we will finally see a regular upgrade cadence. Given they control everything, it will be much easier for them. I do agree that it will be important for them to cycle these machines with reasonable frequency to convince people to support them with drivers and hardware.

I think an interesting question will be the actual PCIe bandwidth on the Mac Pro. Will it be Gen4 or Gen5?

It would be really nice to have a reasonable number for PCIe Gen5 lanes. I could use them for my fusion reactor control system card. I would hope they will be Gen 5, but I am not really sure what cards really need it.

How under-provisioned will the bridge chip be, given that the Mac Pro (2019) was already under-provisioned for PCIe?

Another interesting question.

At least TB4, and next gen's 80+ Gbps are OK speeds for most tasks. Apple is a company that still ships devices with 0.48 Gbps data throughput.

I would hope that they would support Thunderbolt 5 before anyone else, and that the machine would ship with 25Gb/s Ethernet (either by default or as an option).

Apple has said on many occasions (as far back as 2013), that fewer than 1% of users add a card to their machines (not counting GPUs). I do wonder what cards people need, and how many of them could reasonably be replaced with a Thunderbolt 4 or Thunderbolt 5 interface.

 

[avkills]

I did add a USB3 type A (I think it supports USB 3.2) card to my 2019 MacPro; since the Apple IO card only has 2 type A ports. I also have a hub, but the card was so cheap, why not.

Depending on what Apple releases will determine if I get a M.2 PCIe card for more storage.

I sure wish Adobe would let you choose which GPU it uses for compute, because then I could put the 580X back in to drive my main monitors. I have four 4k monitors that I only hook up when I need to test stuff for work.

 

[jakey rolling]

Makes sense. He moved from a 4m2 Thunderbolt connected drive to an PCIe connected internal 8m2 and added 5 more M.2 2TB drives.

I definitely have looked at options like that as well. For the time being, I have not really had any major issues with performance or space with my current setup. And some of your other comments below actually illustrate a bit of a shift I am looking at with my setup (mainly because of my job shifting).

That is what $25 US? (I have family in Canada and I remember when the Canadian dollar was about $1.25 USD).

LOL - yeah, those days are long gone, sadly. As of today, $1 CAD is worth about 75 cents US.

A machine one builds oneself will almost always be cheaper than one that one buys, especially if one does not include the value of ones time (something completely reasonable if one enjoys doing it). When they are not cheaper, they are configured more appropriately.

True enough - I'm one of only a few people in my org who uses a workstation instead of a laptop for my daily driver. There are two others, but they never come into the office or visit clients, so they were issued workstations instead of laptops instead. That is the main reason why my workstation is a home-built one - I still want my company-issued computer to continue to be a laptop for those few times that I need one. If I were to go 100% workstation-based, I would not build my own any longer, I would be using the same Dell Xeon workstation that my colleagues use.

Curious if you could really do your work with your company provided laptop? I am a mostly a consultant these days. I used to have the top of the line Mac laptop and a Mac Pro desktop. Now I use an iPad Pro with a keyboard case and a Mac mini (that I hope to upgrade to an M2 Studio if they release one).

For the most part, I can. Up until this latest refresh, the company provided laptops were generally portable workstations for those of us who spent most of their time at a desk with minimal on-site visits (Dell Precision models). I had used USB-C drives for storage of my VMs, which do work reasonably well, but they had to also be portable, limiting the space they can hold. We do have a large Hyper-V server at work that I could use to run my VMs on (and some others I work with continue to do so), but doing so becomes a bit less flexible than I've become accustomed to.

I found that I usually either work at a desk when I need a real computer (often attached to cloud resources), or I am mostly just taking notes/editing documents and using FaceTime/Zoom.

And that is where my latest laptop refresh, thanks to having a workstation, comes in. Because I now have my work structured such that the "heavy lifting" work is mainly done at home, and client visits are mainly meetings and work on their servers, my laptop no longer has to be the beefy workstation. So instead I have a very elegant Dell Latitude 2-in-1 that serves this purpose really nicely.

Yup, but users like you (corporate users who purchase their own machines and so do incremental updates) make up a tiny niche of a tiny niche. Now that 25Gb networking has gotten so cheap, I am not sure that any machine I ever buy again will have more than 2TB of internal storage. Unlike RAM and even GPU, I tend to want to keep my data and the speeds have not really gotten faster than even 10Gb/s Ethernet can handle.

I realize I'm definitely a niche market. The reason I'm running multiple local VMs on my laptop is because I'm essentially approximating the entire server infrastructure of business enterprise systems a single work computer (for example, to do work for a single client project, I may have a VM running each for a SQL Server, Remote Desktop Server, Application/Reporting Server, and a Domain Controller, communicating to each other through its own virtual network). This is definitely not a common use-case.

But here's the thing with the "use-case" argument: there are a lot of "niche" use-cases out there where, just like my own, the hardware and software requirements simply boil down to being able to have a system grow with you over time. Running a bunch of VMs, or running a bunch of audio plug ins, or streaming multiple 4k streams, or playing virtual chess, or dabbling with AI... these could all be "niche" use-cases. The requirements of all these niches, however, aren't really that rare at all.

Once 10Gb/s Ethernet became reasonable, moving everything other than working storage out of a machine just makes more sense. Storage speed has not increased at the same rate other components have and storage requirements tend to be additive (I need to store everything I had and this new stuff too).

I have everything in a FreeBSD server that supports me and my BF. I connect to it at 10Gb/s he connects at 25Gb/s. That is now fast enough that he can do a proxy workflow and do his conforms off the server (it has 16TB of M.2 NVMe Storage as a read/write cache).

In your case, I am not sure why you would not do that at your next upgrade. Move the storage out of your office to your garage (or since you are in Canada, into your igloo) :-D and upgrade it as you need to do so, without having any impact on your desktop.

Interestingly, I actually just moved away from using a NAS in my home setup, because it was one more server that I had to maintain and back up at home. For everything but my VMs, my company's internal databases (currently in the process of being moved to a hosting provider) work well for what they store, and for my document storage, OneDrive has been the biggest key to keeping my laptop as up-to-date as my desktop. My VMs are now the only thing that semi-permanently exist on local storage, and those get backed up to an off-site file share now periodically.

 

[Alan Wynn]

I definitely have looked at options like that as well. For the time being, I have not really had any major issues with performance or space with my current setup.

Given what you are using, I presume you would have been served just as well by an 8-bay Thunderbolt RAID, something that has been available since the before the iMac Pro. There are not many applications where having large amounts of internal storage is needed for reasons of speed.

Again, just because Apple was early, does not mean they were wrong.

True enough - I'm one of only a few people in my org who uses a workstation instead of a laptop for my daily driver.

So even in your organization, needing a beefy desktop is rare.

There are two others, but they never come into the office or visit clients, so they were issued workstations instead of laptops instead. That is the main reason why my workstation is a home-built one - I still want my company-issued computer to continue to be a laptop for those few times that I need one. If I were to go 100% workstation-based, I would not build my own any longer, I would be using the same Dell Xeon workstation that my colleagues use.

That was why I was saying that you were a very rare special case. It matters to you to be able to upgrade your machine in pieces because you are paying for it yourself. I am betting that the ones that are company issued are not upgraded mid-cycle.

For the most part, I can. Up until this latest refresh, the company provided laptops were generally portable workstations for those of us who spent most of their time at a desk with minimal on-site visits (Dell Precision models). I had used USB-C drives for storage of my VMs, which do work reasonably well, but they had to also be portable, limiting the space they can hold.

The 4M2 is very small but can easily hold 16TB in a RAID or JBOD. Given that USB-C was fast enough for you, it means that Thunderbolt 2 would have been sufficient, and OWC among others had nice external solutions that would have worked fine with the 2013 Mac Pro.

We do have a large Hyper-V server at work that I could use to run my VMs on (and some others I work with continue to do so), but doing so becomes a bit less flexible than I've become accustomed to.

I get that. It is nice to have everything be local, and therefore somewhat faster. I would be curious if your containers could run fast enough on QEMU on a Mac Studio.

And that is where my latest laptop refresh, thanks to having a workstation, comes in. Because I now have my work structured such that the "heavy lifting" work is mainly done at home, and client visits are mainly meetings and work on their servers, my laptop no longer has to be the beefy workstation. So instead I have a very elegant Dell Latitude 2-in-1 that serves this purpose really nicely.

How often do you really need it to be a Windows laptop, and how often would an iPad be sufficient? Not even a high-end iPad Pro, but just an inexpensive iPad with a keyboard.

If you know, how much does the Dell Workstation that your company purchased for your co-workers cost? If they purchased it with external storage that they did not need to upgrade every cycle, would a mid-tier Mac Studio or even M2 Mac mini serve your needs?

I realize I'm definitely a niche market. The reason I'm running multiple local VMs on my laptop is because I'm essentially approximating the entire server infrastructure of business enterprise systems a single work computer (for example, to do work for a single client project, I may have a VM running each for a SQL Server, Remote Desktop Server, Application/Reporting Server, and a Domain Controller, communicating to each other through its own virtual network). This is definitely not a common use-case.

The reason I was arguing that you are a niche market within a niche market is that you purchase your own machine and therefore want to reuse parts when you upgrade.

From what I can tell, your use case is handled fine by Apple’s offerings, and was even fine with the 2013 Mac Pro. What makes it more of an issue is that you want to be able to do incremental upgrades because you are spending your own money. From my experience, it is very unusual for companies to even add RAM, disk space or a new GPU to an existing machine mid-cycle, I cannot think of a single company for whom I have worked or consulted where they reused internal parts or kept the case when they upgraded.

But here's the thing with the "use-case" argument: there are a lot of "niche" use-cases out there where, just like my own, the hardware and software requirements simply boil down to being able to have a system grow with you over time.

That is where we disagree. Just to confirm my thoughts, I polled various friends who have served as the CTOs and/or directors of technology at the Visual Effects Society Awards dinner tonight. I asked if any routinely upgraded machines in place, rather than took a machine that was not fast enough, gave it to another artist and bought/leased them a brand new machine.

Everyone of them said it was almost always the latter and almost never the former. They said it was too hard to handle non-standard configurations and that it was political issue to upgrade in place (“you did it for Artist A, why will you not do it for me?”), by making it so that they just purchased the new standard configuration, they are not doing something special.

Running a bunch of VMs, or running a bunch of audio plug ins, or streaming multiple 4k streams, or playing virtual chess, or dabbling with AI... these could all be "niche" use-cases. The requirements of all these niches, however, aren't really that rare at all.

Yep, but they are also handled fine by the line up Apple has, if one acts as the normal purchaser of these machines. It is only that you want to do it in pieces since it is your own money and you do it all yourself, that makes your case an issue for their offerings.

Interestingly, I actually just moved away from using a NAS in my home setup, because it was one more server that I had to maintain and back up at home. For everything but my VMs, my company's internal databases (currently in the process of being moved to a hosting provider) work well for what they store, and for my document storage, OneDrive has been the biggest key to keeping my laptop as up-to-date as my desktop. My VMs are now the only thing that semi-permanently exist on local storage, and those get backed up to an off-site file share now periodically.

For my personal and my B/F’s use, the NAS is treated as a local cache. Everything is backed up to one or two different cloud providers, and the local storage is just to make working with it locally easier. Frontier just started offering 5Gb/s service in our area. They plan to be at 10Gb/s by the end of the year. With those speeds, pulling down a TB or two would be fast enough that it would mean the cache could be one tenth the size for his needs and could completely go away for my needs.

 

[HDFan]

fixed the Mac Studio's intermittent noise issues.

Will an updated Mac Studio come out that bumps its capabilities and fixes issues

What issues? Not that many reports of problems. Working flawlessly for me.

 

[Joe The Dragon]

From what I can tell, your use case is handled fine by Apple’s offerings, and was even fine with the 2013 Mac Pro. What makes it more of an issue is that you want to be able to do incremental upgrades because you are spending your own money. From my experience, it is very unusual for companies to even add RAM, disk space or a new GPU to an existing machine mid-cycle, I cannot think of a single company for whom I have worked or consulted where they reused internal parts or kept the case when they upgraded.

But most that do have raid setups do hot swap the failed disks (apple does not even have raid 1 in there workstations)

 

[DavidSchaub]

What issues? Not that many reports of problems. Working flawlessly for me.

Oh, there are no serious problems.

My impression is that the Mac Studio fans are a little nosier than some would like?

Going up to 192 GB will also help some niches in the market; as would 40% GPU improvement (and perhaps more with the M3, if the Mac Studio gets an M3 instead of M2).

TB5 (or whatever it is called) will also expand the external I/O nicely.

I'm really just saying: I hope the Mac Studio is not another boutique one-off pro desktop from Apple, as the next version would be awesome.

 

[quarkysg]

But most that do have raid setups do hot swap the failed disks (apple does not even have raid 1 in there workstations)

Probably relevant with spinning HDDs. SSDs now are fast and reliable enough to not need RAID. Most professionals would go external now for massive storage?

 

[Nugget]

Probably relevant with spinning HDDs. SSDs now are fast and reliable enough to not need RAID. Most professionals would go external now for massive storage?

Even more importantly, 10Gbit ethernet is more than fast enough for spinning HDDs that there's no longer any performance justification for HDDs to be physically inside the chassis. Put all that heat, noise, and security concern into the server closet/room where it is easier to mitigate.

 

[HDFan]

My impression is that the Mac Studio fans are a little nosier than some would like?

I've never heard them, but do have a lot of noisy air filters running.

'm really just saying: I hope the Mac Studio is not another boutique one-off pro desktop from Apple, as the next version would be awesome.

My hope as well. My guess is we will know if they do/do not release an M3 version.

 

[wideEyedPupil]

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.

why can't apple offer regular DRAM that is 3rd party upgradable as a kind of tier two level of RAM and keep the SoC universal memory for core computing. the tier two could be a mega-cache of RAM operating at slow read and write speeds but still very superior to memory swap operations with the internal SoC SSD. or it could be even more tightly integrated if Apple could give a ****.

 

[wideEyedPupil]

Apple Silicon was the best thing that ever happened to portable Macs, but as low TDP doesn't matter for the Mac Pro, I don't see why they are transitioning away from x86. A Mac Pro with AMD Epyc would be unbeatable.

maybe Apple want to drop the need to continue support for x86 binaries in Xcode and macOS completely?

 

[wideEyedPupil]

It's a straw man that doesn't even exist yet. Let's all panic!

"Based on nothing other than some past mistakes, and because I have no real information, let's consider the worst thing that could happen."
"Did you hear that? Apple might do the worst possible thing."
"They're almost certainly going to do the worst possible thing. They did it before!"
"I can't believe it. Apple did the worst possible thing AGAIN?"
"I know, right? They're doomed.

Rinse and repeat.

who's panicking? we're speculating based on Apples previous strategies and decisions, and trends in computing. don't like it? read any of the other 10^15 web pages available for your amusement.

 

[wideEyedPupil]

In my opinion, Apple may not introduce the "final" Mac Pro until the M3 generation SoC's with their 3 nm process is unveiled late this year. Two reasons: 1) less heat-related issues and 2) the M3 is supposed to get a new memory controller better-suited for multiple CPU and GPU cores.

also they be able to claim the efficiency improvements against Mac Pros benchmarks with previous Mac Pro and start of the art intel based Linux/Windows PCs

 

[Alan Wynn]

why can't apple offer regular DRAM that is 3rd party upgradable as a kind of tier two level of RAM and keep the SoC universal memory for core computing. the tier two could be a mega-cache of RAM operating at slow read and write speeds but still very superior to memory swap operations with the internal SoC SSD. or it could be even more tightly integrated if Apple could give a ****.

I would guess because it would add a lot of expense (it would mean that they had to pull lots of lines off the CPU to some kind of external socket that could only be used by RAM, rather than general PCIe IO), and make it harder to get companies to optimize for a Unified Memory Architecutre.

 

[Alan Wynn]

A Mac Pro with AMD Epyc would be unbeatable.

It would be just like every other Epyc system, but with less software support than Windows or Linux.

If it was the only system with this architecture, it would get no benefit from all the Apple Silicon enhancements (like support for compression and Unified Memory Architecture) and yet would be too small a niche to get ports from those software companies that are on not on macOS despite currently being able to address the whole market.

 

[iAFC]

It would be just like every other Epyc system, but with less software support than Windows or Linux.

If it was the only system with this architecture, it would get no benefit from all the Apple Silicon enhancements (like support for compression and Unified Memory Architecture) and yet would be too small a niche to get ports from those software companies that are on not on macOS despite currently being able to address the whole market.

Only if you consider that the current Mac Pro is just like every other Xeon system.

 

[Luis Ortega]

I don’t know about you, but I pay more for Apple products *specifically* so I don’t have to customize them.
If I wanted to customize, I would’ve went with windows and Android.

Lol…maybe that’s ok for phones and watches, but not for pro workstation.