Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
M3 Ultra Mac Studio Rumored to Launch in Mid-2024, But No Sign of New Mac Pro
- Thread starter MacRumors
- Start date
- Sort by reaction score
If I'm honest - a few years ago I ran windows 10 in the virtual box on my Intel imac at the time. I've never seen windows run so stably in my life (and I've built a lot of PC's and done installations)... Unfortunately, I don't know anything about the dual-boot and how windows runs that way... Whether that makes sense to use it on mac or not; I could understand it.
My father was an engineer in mainframe computing. He always said; Only doubling the complete performance of the computer is really noticeable and should be decisive for a new purchase. I would even say today - if you don't really need all the power of the computer all the time - buying a new one is perfectly sufficient with a fourfold increase.
Well I hope that's the case when I'm going from my 2019 27" Intel iMac to a ...well...maybe a Studio Max...?...Ultra?
Next year. Intel still going strong enough, but I think I' m missing out on these m chips. I know, I know...bad reason to upgrade ;-)
That's an interesting idea, and makes quite a bit of sense, given that the M3 Pro became its own die, rather than a Max missing some GPU cores and memory controllers. Apple seems to be making enough Apple Silicon that it is worth having more distinct dies.
There are a couple of design wins with that... One is that unnecessary pieces can be eliminated. The Ultra doesn't need 8 e-cores, and the Extreme certainly doesn't need 16 of them - anything that wants a ton of cores almost certainly also wants P-cores instead of e-cores. It doesn't make sense to eliminate the e-cores entirely - they're very useful for writing e-mail or the like. while waiting for a bigger job to complete, but the Ultra can go down to 4, leaving the Extreme with 8 (which is too many, but better than 16). Another example is that the Ultra doesn't need a duplicated Neural Engine, and the Extreme doesn't need four of them. Very little makes use of the Neural Engine, and the few things that do (largely Apple's own apps) are written to run on exactly one of them - it is consistent all the way from the iPhone to the Max, and the number of Ultras out there with a second one isn't large enough to interest developers. If Apple removed the duplicate Neural Engine and some of the e-cores from a custom-die Ultra, they could either make the die smaller than two Maxes or throw in extra units that do scale - either extra P-cores, extra GPU cores or a mixture of the two.
The second win is that there is no need to design in an additional interposer placement. If the current interposer is on the North-South axis of the chip, a four-way Extreme would also require one on the East-West axis. It might be significant design work to do that? Especially if the interposer is placed next to specific chip features, and a new one at 90 degrees to the original would require significant rearchitecting...
The main reason to use PCIe slots these days is for upgrading the GPUs, but AS Macs don't support any GPUs except the built-in one. so yeah it's pointless
The M2 Mac Pro is the *worst* Apple product released in the last five years, and easily the worst AS product.
Upgradable RAM? Nope.
Upgradable graphics? Nope.
Afterburner card? Nope.
What’s the point? Is it because it’s the only Mac that can be mounted in a rack? Might as well just go back to Xserve.
Upgradable RAM? Nope.
Upgradable graphics? Nope.
Afterburner card? Nope.
What’s the point? Is it because it’s the only Mac that can be mounted in a rack? Might as well just go back to Xserve.
I'm looking forward to the M3 Mac Studio models because it's really time for me to replace my 2017 iMac Pro. I like to by the "top of the line" computer and use it for 6-7 years. I can't imagine that Apple is selling many M2-based Studio models since the M3 Max MBP outperforms the M2 Ultra in several respects.
M3 Max has about RTX 3070 equivalent graphics. Even faster for video encoding. 128 GB RAM is the maximum amount.
4 M3 Max in an ’extreme’ SoC would therefore be x4 3070, that’s more power than a RTX 4090. With 512 GB RAM. It’s not 1.5 TB, but it’s closer.
It’s hard to predict what they’ll do. I believe if the rumor of the Jade 4C (extreme) was true, it was scrapped because they didnt know if there would be demand for it. How to find out? Release the Mac Pro with the Ultra and see how it does. If there are some people interested, release the Extreme chip in the Mac Pro in future M generations. If not, discontinue Mac Pro and skip Extreme.
It doesn't scale that way. Beside, if M3 Max is only RTX 3070, that would be a joke since M3 series are 3nm based while RTX 30 series are inferior 8nm. M1,2 series supposed to compete with RTX 40 series but none of them ever close to RTX 40 series but M3 and even M3 can barely achieve mobile RTX 4060's performance.
Also, if you say M3 Extreme = RTX 4090, then it just proves that it still not enough as Mac Pro can use 4x high-end GPU like RTX 4090 instead of just one. It's a workstation, not a regular desktop.
512GB ram is still far from 1.5TB and that was from 2019. Also, the biggest issue with Apple silicon is that it does not support ECC and therefore, it's quite risky for any workflows with high RAM usages.
Clearly, Mac Pro 2023 is not Pro after all.
Last edited:
Seriously doubt this is the case, Apple would have to completely redesign the M3 Max. N3B node is not compatible with the rest of the 3nm processes.
I still think the current Mac Pro could be a placeholder product, until the the M3 series Mac Pro is released.
Yeah, I was going to post something like that. You beat me to it.
I think the current Mac Pro was just a major comprise after having abandoned the idea of a Quad Mx Max a.k.a. Mx Extreme. Just as they’ve done with the M3 Pro SoC, I think Apple decided to step back and just create a unique design specifically for the Mac Pro. But it’s not coming to the M3 generation - the NB3 process is too expensive, so we won’t see an upgrade to the Mac Pro until M4 generation. IMO, WWDC 2025 is when the M4 Extreme Mac Pro will be announced.
Last edited:
Use M4 Max to create M3 Ultra? That’s a definite no. They just released the M3 generation, there’s no way the fourth generation (M4) will be ready before 2025. Especially since every company other than Apple has been waiting for the N3E process - there’s probably going to be limited availability and Apple is going to reserve all they can for the A18 and A18 Pro.
M3 series supposed to come out in early 2024 and yet, they announced them in Oct, 2023.
Not really impossible especially since N3B that M3 series are using are low yield chips and difficult to manufacture while N3E is different but simpler to mass produce. Beside, Apple already hated N3B.
Absolutely concur. Now that it’s 5 year later, at least 2.5 TB would be a more realistic high RAM utilization limit. But it doesn’t matter anymore, it has been too long, that [niche] market has already moved on from Apple to Linux
Too bad then since it was Apple who ditched pro markets for a long time since Mac Pro 2013 and now, they made a same mistake again. They learned nothing from their own past. Yet, Apple is start expanding their Mac marketshare with 3D, game, AI, and research since they aren't really selling well with Mac computers for more than 1 year.
I wanted to buy an MBA 15" but stumbled upon an mint MBP 16" I9 from 2019. It was too tempting (and the right decision)... If I were you, I would wait for a good deal for the next big IMac (32"), if yours can still stand upright ;-).
I'll wait another 1 or 2 years, then the next upgrade will come. Even if the MBP is very good - 512GB is too little... Until then, the MBA will also be at the level of today's MBP
That would be a terrible choice especially for SoC.
Max series chips are already too big and they are difficult to mass produce and yet expensive. Bigger the chip size, higher the price. It's already close to RTX 4080's die size.
And making a separate chip? That would be way more difficult unless Apple is using chiplet design which is the only way to solve the problem with SoC. Design each part of chip and then combine them in one chip just like AMD did. But it won't be ready till 2025 so I doubt it.
Likely answer: assuming the software got ported to ARM64, they'd perform almost as well on GPU-heavy AI training and serious 3D as a Xeon or Threadripper workstation running the same software on the exact same GPUs. The real question is how could an Apple Silicon Mac Pro tower with PCIe GPUs distinguish itself? The Unique Selling Point of small-form-factor and laptop Macs is power/performance, not raw speed - bolting in a couple of big, sweaty NVIDIA space heaters, dumping the SoC approach throws away the advantages of unified RAM will throw a lot of that under the bus. ARM will likely always be a bit more power-efficient than x86 but who actually cares about that in a high end personal workstation? The huge-form-factor Mac Pro certainly ain't a battery-powered laptop, nor is it a high-density computing system (where the power consumption and HVAC costs start to become serious).
The main market for the Mac Pro since 2019 has been users locked into a MacOS-only workflow - which will a shrinking corner of a larger pool which is also shrinking as laptops/small-form-factor systems get more powerful at one end and on-demand cloud-based processing power for "heavy lifting" becomes more prominent at the other. Not really where Apple would want to throw its money.
Not to mention Apple making the Mac Pro market so uncertain, with the OG Cheesegrater, the trashcan and the 2019 MP each being left in limbo for years followed by a drastic workflow-breaking change in direction.
Well, yes, but the demand for the "Real Mac Pro" seems to be from people who want a lot more RAM than on-chip unified memory can offer (not just the doubling you'd get with an ultra) and/or want discrete GPUs (which break the unified memory model anyway) - you can't have it all. As I said, its probably not a good idea.