tenthousandthings

I encountered these some years ago when trying to troubleshoot something (I can’t remember what it was). The scanned image of the photograph kept reappearing in Image Capture despite my attempts to delete it. I did an image search to find out what it was, just out of curiosity and annoyance. I thought maybe it was something I had scanned long ago (I work with lots of images). It eventually dawned on me that it was some kind of built-in test image. I remember encountering the bitcoin paper as well, but I can’t say I thought anything about it. 

On Craig Wright, LOL. His posturing about this just confirms he’s not the founder. I could make a stronger case that this proves Satoshi Nakamoto was an Apple software engineer…

YouTube TV bought the rights to NFL Sunday Ticket. They outbid Apple and others. They’re going to get an influx of new subscribers with that (even though it is an add-on), so this price hike is likely part of that calculus.

I don’t know what happened with MLB Network and the MLB.TV add-on. I’ll guess that will be resolved, but maybe not. I’ll also guess YouTube TV wants a special deal/rate with MLB because they’re going to have all the NFL fans, so they expect to have a higher percentage of sports viewers than other live television streaming providers. 

Well, that’s certainly one solution for the low-volume problem when we’re talking about the economics of building variable Apple Silicon GPU modules/components for the Mac Pro. 

macxpress said:

The iMac has always been a family oriented Mac. IDK why anyone would think it was any different. 

April 2017, Craig Federighi: “… The original iMac, you never would’ve thought as remotely touching pro uses. And now you look at today’s 5k iMac, top configs, it’s incredibly powerful, and a huge fraction of what would’ve traditionally required the Mac Pros of old and are being well addressed by iMac — whether its audio editing, video editing, graphics, arts and so forth. But there’s still even further we can take iMac as a high performance, pro system, and we think that form factor can address even more of the pro market.”

mjtomlin said:

tenthousandthings said:

mjtomlin said:

programmer said:

Serqetry said:

That is what I've always thought.  If Apple doesn't design a new ARM chip series for the Mac Pro that exceeds the limitations of the SoC M-series, there's really no reason at all to even make a new Mac Pro.  The dumb rumors about an M2 Ultra Mac Pro make no sense.  I think that is more likely to be misinterpreted info related to an upcoming M2 Mac Studio refresh.

There might be a new Mac Pro coming too, but it has to do a lot better than M2 Ultra... and it probably isn't coming as soon as people believe.

Why?  The Mac Pro is about a tower chassis with expansion capabilities.  There’s no rule that says it has to have higher performance than every other model.  The Studio has good thermals and that’s the primary limitation when it comes to performance.  And while it’s nice to think Apple could crank out a SoC just for the Mac Pro, the (1) economics of doing that don’t make much sense:  they just don’t move enough volume to warrant it.  When using Intel they had the luxury of a buffet of workstation and server class chips, plus AMD and/or nVidia to provide a selection of GPUs.  This is no longer true with ASi, so they need to figure out how to scale high enough without over investing.  Tricky problem.  

The whole M1 series was an impressive exercise in scaling, and if they can improve on the M1 Ultra’s efficiency bottlenecks then the M2 Ultra will be even more impressive.  It isn’t going to satisfy the discrete GPU adherents though, so either Apple backtracks and uses an AMD GPU (which kind of undermines their strategy), or they use multiple M2s (some sort of grid compute software solution), or they just stick with the M2 Ultra (for now, until they can improve their scaling story further) and put it in a slotted chassis, or they put off the Mac Pro until M3 (which leaves a hole in their product line).  My guess is they’ll use an M2 Ultra, perhaps clock it a bit higher, and focus on expandability… but at least have a product.  M3 is spooling up this year though, so perhaps a bit more delay and launch the next architecture with the Mac Pro?

1. The Mac Pro is already expensive because of this reason. For Apple, the biggest chunk of cost for their custom silicon is in R&D. They can get away with producing SoCs for the Mac because those SoCs use the very same core designs as the A-series. So the cost is covered by A-series volumes. It just a matter of plugging the blocks together to make a different SoC, or even a different chip altogether. The proof is in the fact that Apple has a huge line of SoCs (A, M, T, W, H, U, S-SiP) that all interchangeably use the same core IP blocks. So, creating a custom "X-series" CPU and/or custom "G-series" GPU for a high-margin system that uses the same IP as all their other silicon designs is not that difficult to believe.

I do agree that the Mac Pro "SoC" does not necessarily need to be more performant than the "Ultra", but it does need to be more flexible. Otherwise, it is just a Mac Studio with slots, which is a system they could've already released. And if they truly didn't care about the Mac Pro and what it represents, they would've done just that. And it would be an abject failure as a vast majority of users interested in a system like the Mac Pro are mainly interested in expandability and flexibility.

You’re not accounting for an essential component of what you’re proposing. The Apple Silicon lines that have graphics (A, M, S) have it integrated on the silicon. The only partial exception (so far) is the M1 Ultra, which uses a silicon interconnect fabric developed by TSMC to “fuse” two M1 Max together.

Using silicon for an interposer is expensive. That’s why AMD developed Infinity Fabric (now “Infinity Architecture”), which doesn’t use silicon. To do what you suggest, Apple would need to develop something similar. PCIe isn’t the answer, not with Unified Memory Architecture to consider, along with things like Neural Engine(s).

One realistic solution close to what you’re talking about would be to create another entry in the M-series Pro-Max progression: the Pro has one CPU unit and one GPU unit, the Max has one CPU unit and two GPU units, so the hypothetical “Max Extended” would have one CPU and four GPUs. So then you’d have the Ultra (2x Max) and the Ultra Extended (2x Max Extended). That’s a lot of silicon, and they would be expensive, but they would also have the kind of compelling performance Apple has said they are looking for.

You're getting hung up on Apple's marketing for their SoCs, which are designed and optimized for highly integrated systems, i.e. systems that are almost completely self-contained and not expandable. There's no reason the Mac Pro needs an integrated GPU or have a UMA. They can get rid of the SoC (System on Chip) and continue to use System on Board design of the current Mac Pro. They can do this by creating an ASi drop in replacement for the Intel Xeon and an ASi GPU MPX module along side the AMD models.

Current Mac Pro w/ Intel Xeon supports 64 PCIe gen. 3 lanes is any SoC going to be able to support that kind of system-wide throughput? And it gets even higher should Apple decide to move to gen. 4 or even 5.

[NOTE: The following was written prior to seeing Programmer's response to Serqetry above.]

Well, more power to you, it would be interesting to watch Intel's reaction, but I think you're glossing over the real costs of developing and manufacturing an Apple Silicon drop-in replacement for the Intel CPU. Like Programmer said, the "economics" of that are questionable. These advanced fabrication process nodes that Apple is using are super expensive. When Apple started developing the A15/M2 family, the TSMC N5P node was cutting-edge fab. Sure, the iPhone and iPad absorb it, but we're talking hundreds of millions of dollars in development costs. Adding another big, complex project into that mix is not simple or cheap, even if low-volume. The (rumored) TSMC N3 transition for A17/M3 is even higher-end fab. 

An MPX 2.0 (e)GPU that uses PCIe 5.0 is more feasible, and arguably is just more about packaging (and software engineering) than anything else, since they can just use the existing M2 Pro/Max GPU design. No more difficult than the "Max Extended" approach I mentioned. 

I do listen to the marketing, especially these two recent Apple Silicon interviews. I'll admit there's a reading-tea-leaves aspect to it, but I think it's foolish to dismiss it as sales gimmickry. They're sending a message, not just to customers, but to the industry, when they talk about how coming from the extreme thermal constraints of iPhone/iPad (and also MacBook re: M-series) has given them a different sort of approach/perspective and allowed them to do things others have not. It seems pretty clear to me that they have no intention of throwing all that out the window for the Mac Pro, because long-term that is the way forward, even for HPC. Apple is way ahead of the curve.