We are now several months beyond the 2 years that Apple Claimed would be needed to transition the entire Mac line from Intel CPUs to Apple CPUs. What they have done is nothing short of a miracle. Through a global pandemic, multiple lockdowns and limited production environments in various countries, and matching or exceeding performance per watt of their previous systems and the consumer available hardware, Apple has almost completed their lineup transition. However, aside from a larger iMac and a more powerful MacMini (with more ports), the most obvious machine that has not come to fruition is the Apple Silicon Mac Pro.
Pundits like to speculate as to why, and we are no different. There are several possibilities as to why we haven’t seen an M2 Extreme chip or the associated computer(s) announced or released.
- The Mac Pro is a niche product, and not likely the money maker that people think.
- The Mac Studio (with M1 Ultra), greatly reduced the perceived demand for the Mac Pro.
- Apple is having difficulties merging the SoC philosophy of Apple Silicon with the modularity demanded by the Mac Pro customers. They don’t want another MacPro 2013.
- Chip yields for the new SoC, with all components working at 100% is too low.
- The Post-COVID chip shortage is preventing them from having the discrete components to make it work (see the SSD on the M2 MacBook Air/13″ Pro), and have to re-engineer the SoC to absorb some of those components, and it is a bit more difficult than expected.
Of these possibilities 3 and 4 are the most likely.
From what we have seen so far, Apple has been directly connecting SoC and RAM, which makes the RAM a static item and unable to be upgraded. However, Pro users, will likely be investing thousands of dollars into this machine, and want the ability to upgrade it over time, to extend its lifespan. As a result, they will demand additional RAM, expandable storage, expandable/upgradable GPUs, hell, they want the ability to expand and customize their hardware to meet their demands.
RAM – It is possible and likely that RAM won’t be a huge issue, the would just need a memory controller that isn’t baked with the maximum RAM set, and the appropriate data lines to RAM sockets. However, in Pro environments, there will be demand for ECC Memory, which may be a hurdle that Apple is trying to overcome, not with the commercially available DDR5 that the would likely be using, but with their own system RAM. I do see RAM in this configuration being like cache was during the 486-Pentium era. There is the built in memory, then optional external memory. Allowing customers to buy their Mac Pro with 128GB of RAM on the SoC, and the ability to upgrade it to whatever level is deemed appropriate and possible with DDR5, at the time of release.
Storage – Apple can be stubborn, as seen in their integration of a storage controller into there SoC, making the removable storage cards in the Mac Studio non-upgradable at present. I expect a similar configuration for the Mac Pro. However, I’m hoping that Apple would also provide NVMe sockets, PCIe Gen 5, with the caveat provided to customers that it may not function as fast as the stock drives. And while it appears to be going away slowly, SATA/U.2/SAS ports available would be a nice addition.
Expansion – The Macintosh line, over its lifespan has had a dubious relationship with expansion. Sometimes, you get no options for expansions. Other times, you do, but they use proprietary connections, or similar connectors with different pin layouts that will fry the card and/or the computer if used incorrectly, or you have super powerful upgrade connectors that can not only replace your CPU, but totally change the configuration and operation of your machine. Thankfully, modern Pro Machines (aside from the previously mentioned 2013), have used industrial standards, PCI, AGP, PCI-X, and PCIe for expansion. And while we can assume that all of the internal communication of the SoC and the related connections to storage, video, and Thunderbolt are all using PCIe signaling, opening that up to unknown uses may be proving a little more difficult that originally envisioned, or they can’t provide enough dedicated lanes without a chipset, which Apple is likely trying to avoid.
GPU Support – GPUs aren’t just for rendering graphics these days. They also including neural engines, and other distinct functions that most CPUs are unable to handle efficiently on their own. Currently, Apple doesn’t support additional GPUs on Apple Silicon, which may be a software or hardware limitation. If the former, hopefully as time moves forward, that can be fixed for everyone. If the latter, that will have to be fixed eventually, as while Apple’s GPU numbers are impressive, especially for the power draw, they won’t always be king of the mountain, and Pro users, will want the ability to upgrade, either with Apple discrete GPU upgrades, or with hardware from Nvidia, AMD, Intel, or an unknown future supplier. That will be a dealbreaker to many Pros, especially those in scientific and production environments.
Yields – Integrated circuits are amazingly complex. The more processes, transistors, diodes, and other components you add, the more complex it gets. Start adding more than basic logic to it, it gets exponentially more complex and difficult to produce. Make those components progressively smaller, and it gets orders of magnitude harder. One of the reasons that companies like Intel and AMD produce a series of chips in each generation with different capabilities and prices comes down to binning. Binning in the process in which a produced chip has its capabilities tested, verified, and assessed. Some chips will have flaws, and components like integrated graphics don’t work, so they are either physically disabled or microcode is loaded into a small storage components in the die to disable that option. Perhaps one of the cores doesn’t work right, gets too hot, isn’t fast enough; Turn it off. This helps companies maximize their abilities to pay for production by reducing the number of chips that are total rejected. We saw this with M1 and M2, some chips had fewer GPU cores than advertised, because they were binned. There was difficulties getting some models of Mac Studios because the production of the SoC went so well, there weren’t enough binned SoCs to keep up with demand, and eventually Apple had to just start binning near perfect SoCs to keep up with demand. However, it is possible, that the rumored M2 Extreme, which 4 M2 Max cores, is proving to be a much harder chip to produce. And, they don’t want to run two M2 Ultra cores with physical separation after their presentation touted the advantages that M1 Ultra had with its interposer. If this is true, expect to see M2 Max and Ultra long before M2 Extreme, but if it isn’t true, all 3 should come out at roughly the same time (within the same quarter).
Ultimately, we won’t know for sure until we actually see an announcement, but hopefully this helps you understand some of the complexities that Apple is facing. And while they missed their 2 year goal, it is entirely understandable, and they are continuing to push forward.