A mac has also never announced in September for the last 9 years.
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.
The secret sauce of Apple Sillicon
- Thread starter Kung gu
- Start date
- Sort by reaction score
Here:
Sounded like you were predicting next week either a new MBP or obliquely, a new chip that would be used in the MBP derived from the A15.
There was a report of a leaked benchmark on the A15 GPU with 13% improved performance over the A14. iCaveDave video on YouTube this morning.
You know, there was great hope and excitement that the new MBP laptops would come with WWDC with an “M1X”, an M1 with more CPU and GPU cores. Then there were rumors that tied these MBP with codenames of Jade C-Die and Jade C-Chop, assumed to be full fat 32 core and chopped 16 core GPU chips. But there was also talk of Jade 2C-Die and Jade 4C-Die. I assume those would be multi-chiplet systems with 2 and 4 times the CPU and GPU cores, respectively. Perfect for a MacPro… but that system hasn’t been rumored earlier than this late 2022.
So, if half of that is right, it doesn’t make sense the 2021 MBP SoC is an extended M1 design based on the A14 core because that logic train would end with next years Mac Pro also being based on a multi-chip M1 when that year’s phone SoC is 2 generations newer. Hopefully the summer 2021 MBP rumors were just hope and they were always planned for just after the A15/iPhone 13 (just like the original M1 was announced near the iPhone 12). So the MBP gets an ”M2X” with A15 cores and the 2022 MBA gets a base M2.
Does the ~30” iMac get the “M2X” or a two chiplet version? Don’t know. Now I’m starting to think the large iMac and new Mini gets the ”M2X” in 2022 as the rest of the consumer line up is updated with M2 SoC. What about the Mac Pro? Does it wait until the end of 2022 to get an A15 based 4 chiplet design? Probably not, I‘d hope. Perhaps that machine is the final product conversion and launches with an M3/A16 based SoC platform as the rest of the line up is refreshed with M3/A16 based SoCs and the rolling switch over from Intel is complete basically in 3 waves. Consumer grade first, then mobile and thin “Pro” hardware, then the ”less power constrained” Mac Pro. At that point Apple would make a phone SoC, a consumer Mac SoC, and a Pro SoC that can be used in multi-chip arrangements to flesh out the rest of the product line. 3 basic SoCs to design and update on a yearly basis.
I continue to disbelieve that an “M1X” with M1/A14-style cores is a thing that exists.
I agree. Looking forward to the next round of updates… I have a 2012 mini fully loaded that is buried in my home office table (along with a UPS and other equipment)… it’s high time for a serious upgrade and a new high DPI monitor
Completely agree. It would be really weird if 2 years into the ASi transition and all they have to show for it are an increasing number Firestorm, Icestorm, and G13 cores. However even this reality is a bit unfortunate for the Mac Pro since it'll be using A15 cores instead of whatever the A16 has in late 2022 (especially if those rumours of the M-series chips using 3nm before the A-series ends up being true).
the iphone will never be better performer than a mac even with the exact SoC....physics are the limit
So they need to address the iphone 12 performance issues in the iphone 13, there were some rumours about vapour chamber heat management for the next gen
Or perhaps the Mac Pro is a summer 2022 release and the transition to Apple Silicon is complete in 2 years flat. If it really is using multiple MBP SoC, does it really take a year to stitch them together into the Mac Pro?
In any case, seeing how they scale the memory bandwidth into the SoC and between chiplets will be interesting.
Gurman, who's been quite accurate lately, has been using the M1X name when referring to the next MacBook Pros.
I don't trust any of these rumors, not the self-publicists who spread them. However, if I was going to put money on this, I'd bet on cmaier being correct, rather than Gurman. We'll just have to wait a couple of months to see who's right.
I could be wrong, of course. But the whispers I’m hearing are different than the whispers Gurman is hearing. Maybe different whisperers.
Yes, this will be interesting. It made sense that A14 and M1 is fairly close from an architectural point of view, but I think that M2 and A15 will be a bit more separated on the CPU side of things so not just a moar cores thing. But we will see soon enough.
I know nothing. But what if the plan had been to release the M1x Macs in the summer, thus keeping them in the "M1" era of SoC's, but the mini LED shortage led to the delay until after the A15/M2 series of chips were ready? I'm not an expert, but I'm pretty sure that if they were spinning up production on computers over the summer, they wouldn't be swapping out M1x for M2 SoC's, even if they were pin compatible.
Also, the rumors from a few months ago were saying that the M1x would be more powerful than the M2, due to the performance/efficiency core setup.
The rumors from the summary were something like 8 performance/2 efficiency cores. That would not make any sense if these were M1 cores. Ratio would be very inefficient.
Look. I SAID I know nothing. And also that I'm not an expert.
But from my "the last microelectronics class that I took was when the 128k Mac was king" I would have assumed that the 4/4 ratio on the M1 was to maximize battery life while still allowing the SoC to burn rubber when needed. So (to me) the idea of 8/2 or 10/2 (which was in another rumor) just means that the chip will be faster, but will need more juice. Which is less of an issue in a Mini or iMac. And the 16" could just have a bigger battery. But I guess that's not at all how this works, is it?
Look. I SAID I know nothing. And also that I'm not an expert.
Not exactly. Each time you run a thread that doesn’t need high performance on a performance core, that means it is taking time slots from threads that really need it. Particularly where the high performance cores don’t have hardware multi-threading, that’s a bad idea - it increases power consumption *and* reduces overall system performance.
That’s why I surmised that, if true, 8/2 or 10/2 or whatever suggests that the low performance cores operate very differently than M1 (or the high performance cores are better able to recover from context switches, etc.). Something different than M1.
To summarize, it sounds like you're saying the new MBP's should feature a new generation of chip for two separate reasons: (1)They want the new MBP's to stay current with the latest iPhone; and (2) The switch from 4 to 2 efficiency cores requires a fundamental redesign anyways.
It's technically possible they could have used A14-generation technology to create a new chip with an 8 performance/2 efficiency architecture--but that wouldn't make sense.
Others, of course, are pushing the A-14-based M1X idea (e.g., https://www.notebookcheck.net/Next-...lorful-2022-MacBook-Air-release.549263.0.html). It will be interesting to see what happens.
As an aside, I think when discussing the changes to the CPU/GPU cores, folks lose sight of potential changes to all the fancy coprocessors that are an essential part of AS's performance. I expect those will be significantly upgraded in the new MBP's.
Last edited:
I always assumed that the performance cores were similar to the efficiency cores, just "beefier." Like the engine in a Mazda 3 vs the engine in a Bugatti. But you're saying there's some kind of qualitative difference as well? And, again, knowing nothing, if I had 8/2 SoC, even if it needed 4 efficiency cores and assigned two of the performance cores, that would still leave 6 performance cores, so wouldn't that still be better than a 4/4 M1, even if it isn't the best use of resources?
Yes, the differences are qualitative (and depend on what chip we’re talking about).
As for your follow-up questions, if you look at your process monitor you will see hundreds of threads running at a given time. Most of them are low-priority. It’s better that they run on efficiency cores rather than performance cores, at least if you actually have real work to do, because the performance cores can’t simultaneously do more than N things at once, and you don’t want any of those N things to be things that could have been done on efficiency cores.
So the performance cores can only do N things at once, but the efficiency cores can do N+X things at once because they are able to multithread? My Mac currently has 513 threads running. So you're saying that hypothetically 500 of them are low-priority and 13 are high priority (or 450/63, whatever). So the efficiency cores might be able to handle 100 low priority tasks each, while the performance cores can only handle 10? But they can run those 10 at 10x the speed that the efficiency cores can run them? And I'm talking about the M1 and a hypothetical M1x.
The efficiency cores are not multithreaded, but they (typically) have more pipelines, more flexibility in reordering instructions, sometimes they can do functions in hardware that otherwise need to be done using other hardware (for example, sometimes there’s a hardware multiplier in the performance cores, but the efficiency cores have to do multiplication by using the adders), etc.
In any case, the simple point is if I have a job that needs high priority (because the user is interacting with something, or there is an important calculation going on, or whatever), you want it on a performance core. But if something that takes a long time but the answer isn’t needed right now is running on that core instead, the overall system is slower.
True, but most low priority tasks can be batched on a P core with a low duty cycle. I look at Apple products from a system design perspective (FD: for which I am not qualified). There are many gains to be had— for instance using dormant specialized ML units to enhance the more mundane traditional CPU cores. My hope is that their (less burdened) silicon design is less monolithic and relies on tools to meld intellectual property in a comprehensible and provable way. Hence my (unaffiliated) plug for Clash in another thread. Thoughts?
Last edited: