Apple is skipping ECC primarily because they don't have enough bandwidth in their system design. So it is more so they have repeatedly shown they will do what is convenient for their immediate interests. Scaling at the top end ... that has not gone without hiccups.
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.
Apple's M3 Pro, M3 Max, and M3 Ultra Chips Could Offer Even More CPU and GPU Cores
- Thread starter MacRumors
- Start date
- Sort by reaction score
Apple is skipping ECC primarily because they don't have enough bandwidth in their system design. So it is more so they have repeatedly shown they will do what is convenient for their immediate interests. Scaling at the top end ... that has not gone without hiccups.
Yes, for laptops, that would make a ton of sense. Take the efficiency win, while still racing ahead on performance.If Apple can get 20-30% without clock gains, then I’d say it’s even more likely we won’t see any significant clock gains, likely just staying in the 3.4 to 3.6 range again. But, hey that’s just me. I may be completely surprised, but one thing I won’t be is disappointed
For desktops, you can make a good case for driving clocks up a bunch. And maaaaybe for an MBP when not on battery, to the extent the cooling can handle it- which is not that well, for sustained load, but for bursty workloads, it's plausible.
How do you know that, about ECC?
You're certainly right about "without hiccups" or we'd have a better Mac Pro right now. Not sure how well that argument extends.
The rest of your post got lost in quoting errors. Please fix it so I can read it.
Oh I think there’s a case that could be made, I just don’t feel that Apple would care to listen.
As long as mobile platforms remains their bread and butter, their money is going towards producing efficient, performant processors to go in those mobile systems. Then they’ll hold their moistened finger aloft to detect for the winds of desktop market and toss a bone in that general direction. The bone being the same cores, roughly the same clock speed, just more of them.Lather rinse repeat until they don’t detect a breeze LOL
I feel like everything you are saying is based on what actually exists now, rather than on what could be.
As for the Mac Pro. The 2019 Intel Mac Pro came about because Apple started losing corporate and creative customers in the years following the 2016 Butterfly Generation releases (e.g. I used to follow a famous professional photographer who put out regular blogs on his techniques and equipment. I was appalled, and yet not at all surprised, when he ditched Apple for Windows, due to how far behind Apple was falling in hardware quality and capability, and creative software updates).
Apple, to their credit, created a research working group, in which they invited professional users in, gave them space and equipment to work with, and observed them, and asked them questions about what the do and what they want and need.
This resulted in massively improved hardware such as the 2019 Intel Mac Pro, and the 2019 16" MBP. The Mac Pro in particular was a seriously impressive machine, with a huge array of self-upgrade parts and options, including thing such as up to 1.5TB of RAM.
The current AS Mac Pro, sheesh, it is a shadow of the 2019 machine. Massively hamstrung by the restrictive limitations on customising and upgrading parts and options.
The overriding problem I see is that Apple is so damn focused on profits above all else, that it hasn't the imagination to create products purely for the joy of creating great products for niche markets, even if the revenues aren't super massive.
For example, who cares if the iPhone Mini only sells to 1% of iPhone customers. 1% of uberbazillions of total iPhone production is still an impressive number of phones. The sales volume of the iPhone Mini would still be so large that many smaller phone companies could only dream of such numbers. Sure, it wouldn't have added much to Apple's bottom line, but I bet it didn't actually make a loss, and even if it did, is it such a bad thing for a giant like Apple to have a little side line loss leader that gains good will from the 1% of customers that vehemently wish for a smaller physical phone size?
Same goes for Apple's discontinued, but much loved, wifi router. And so on and so on.
Same goes for all the customers that wish for the 12" laptop footprint.
And what is with the infatuation with somehow assuming that customers who want pro features also want large form factors, and vise versa. Why was there no iPhone Mini Pro? Or until recently 15/16" MBA (thankfully, they did just put out the 15" MBA, about bloody time)? But then, why is the MBP in the slightly larger 14/16", and the MBA a smaller 13/15"? Why did they have to kill the 17" MBP so many years ago? Why do you have to have the M2 Pro chip to get more than 24 GB RAM? And the M2 Max chip to get more than 32 GB RAM? Chip power and RAM do NOT necessarily go hand in hand. RAM size holds more data, that's all, it doesn't mean you have a need to crunch swaths of that data with complex algorithms, it may just mean you simply want to view it fast. Why can't you get more than 2TB SSD on a MBA?
Apple seems to cull any product that doesn't clean up a ton of profits. It's a damn shame. Just let these little departments run. No harm done, and a handful of customers absolutely love these products.
The Cynic in me thinks Apple bought all the 3nm TMSC space to try and "buy time", to stagnate progress from Nvidia and AMD.. to allow Apple to close the gap a bit more with their igpu's. Apple is at least 3 yrs behind on the GPU front compared to Nvidia and needs to take big steps to actually close that gap. Right now you are paying $4-5K to get the mid-tier last gen Nvidia performance.. AI/ML its even worse.. Apple is probably closer to 5 yrs behind from what I have seen.. an Ultra comparing more to a GTX 1080 in ML training and Inference, than the normal 3070ti comparison at 3d.
I'd be surprised if there's a significant number of people who upgrade a Mac from generation-to-generation.
That's a f*cking shame that people are forced to use third party gadgets for this. At least implement the bloody display port daisy chaining in MacOS finally.
There are issues with parallel computing that no engineering can solve. Even without considering the overhead of multithreading, you'll never achieve linear gains with more cores.
Amdahl's law - Wikipedia
In well written programs they create threads based on how much the cpu can handle. The big jump was having multiple threads, and somewhat at going from 2 to more. Once the program is adapted to that, there is not much work with any more jump. The question is how much does it benefit from more threads, and it depends on the task (how much of the work is strictly serial) and the overhead of handling multiple threads.
Anything running on GPUs is easily parallelizable by default, I'd be surprised if there is any issue there. Games are little different, but the problem is on the CPU side and from the requirement of having fast feedback to user input.
DisplayPort daisy chaining wouldn’t have helped, the GPU still needs to have the 3rd display buffer to render the screen, which the M1 and M2 don’t have.
Which is a joke for the price we’re playing. $400 machines can do it.
I agree they really should have had that. In 2020, even the i3 Intel MBA could output to 2 (or more) displays, while months later the M1 Air couldn’t, which is more or less the only thing downgraded.
That said I don’t see Apple too eager to, since the same M1/M2/M3 chip also goes into iPads and iMacs, where 99% of the target audience aren’t expected to ever use more than one ext screen. Adding an otherwise unused display buffer is just wasted silicon budget in other words.
I completely agree guys, I'm just merely stating that you'd expect performance increases every generation.
I don't think the silicon costs would be particularly substantial. Intel CPUs have been able to do this for a long time with only a tiny fraction of the transistor counts, and the iGPUs in these Intel chips were even smaller. These were seriously tiny compared to the M1/M2 chips.
If Apple wanted to put these in there, they could have very easily done so without really making many sacrifices to do it. At first I thought it was just a first generation product limitation, but after the M2 came out with the same limitations, I have started to believe this is much more likely an upselling thing for Apple.
I have an M1 Mac Studio Max 32 GPU. I will only upgrade this computer to an m3 Max if I see a significant improvement in both CPU and GPU. At this moment, it looks like CPU will be a lot faster with the M3 chip but I still have doubts about the only 40 GPU cores at maximum. We'll see. I am looking forward to benchmarks.
I just had a flashback to the dark intel mac days - sometimes BARELY any speed increases between releases.
If Apple wants to get out of the Mac sales niche in comparison to Windows PCs, they must deliver at higher AMD/Nvidia 4080 performance level (M3 Ultra).
Very few people will buy a Mac for gaming with the current Apple tax rate. I can buy a MacBook Pro for work and a PC with top hardware for less than the price of the current M2 Ultra Studio.
For a lot of professionals, that level of performance would mean nothing unless it has Cuda capabilities. Well, anything Apple Silicon won't have it.
And finally, Apple Silicon itself is a problem. What works wonders in lower segments makes the higher end messy. I'd be happy to have a 24-core cpu, but I'd be fine with an 8-core gpu. For others, they'd be fine with 8 cpu cores, but would wan't the 60-core gpu. A lot of people starts paying for unnecessary hardware at a very high premium. But Tim Apple worships the God of supply chain management, and that is an unforgiving God.
I love to use Macs, and I like how they integrate with my other devices, but with their current pricing and product situation what makes sense is to buy the cheapest possible Mac and use it as thick client to a much more powerful pc/vm.
That's part of what I liked about the idea of having the unbinned M2 Pros for the laptops. You can still get virtually the same CPU performance without the extra GPU cores if you wanted it (which is how I configured mine).
If I were Apple, I'd take advantage of the fact that Intel is desperately trying to earn Apple's business back. I'd say "hey, you already know how to write GPU drivers for MacOS since we have millions of Intel ones that were shipped with your Iris Pro chipsets. Think you can take some of that work and port it to Apple Silicon for your Arc GPUs? (for the Mac Pro, where there are PCIe slots)"
Probably wouldn't be the most profitable venture for either party, since the majority of people would just stick with the ones integrated into the M2 Ultra series chipsets. That's part of the problem. There wouldn't really be anything stopping Apple from allowing it, but a company would have to feel like they could sell enough GPUs for a very limited and niche market for it to be worth it for them to create the drivers.
If they go the third party gpu way, it should be nvidia because of Cuda.
I agree, but unfortunately Apple's transition to Metal ruffled some feathers at Nvidia and we haven't had drivers since.