You're right. It is not proven, in the same sense that it currently not proven that aliens will provide the next gen chip.
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.
User predictions for Apple Prosumer Silicon (M1X, M2 etc.)
- Thread starter leman
- Start date
- Sort by reaction score
So you dont have an official link. Please, stay on topic. If you have nothing to say/contribute for the topic, just dont flood it with nonsense
Thank you so much
New microarchitecture at TSMC 5nm+ or 4nm. 8 performance + 2 efficiency cores. Faster clocks (3.5 GHz?). 16 GPU cores (maybe up to 16 GPU cores so maybe some binning). Same number of Neural Engine cores as the M1. Name not guessed but not M1X. I'm guessing not M2 either. New memory controller for up to 64 GB. If LPDDR5 is available I'm guessing Apple goes with that. It looks like Micron has 128 Gb (16GB) LPDDR5. Both Intel Tiger Lake and Qualcomm Snapdragon support LPDDR5. Either 2 or 3 TB/USB4 controllers. PCIE4x4 SSD controller. Two or three 6K displays. That's a lot of cores and I/O so I'm guessing it is a pretty large SoC probably close to double the M1's 119 mm².Since most leakers seem to agree that new Pro Macs will make an appearance at the WWDC, and given that there were a lot of recent speculations on what kind of chips these Macs will use, I though it would be a fun thing to collect all the different predictions folks here might come up with. After (and if) the products are announced, we can evaluate all these predictions and see which guesses came closest.
You can post your predictions in this thread, and I will do my best to keep this post up to date with all the different ideas you ladies and gentlemen care to share.
Ground rules
Predictions
- Be realistic and reasonable
- Be as specific with your prediction as possible
- Include your reasoning
- One prediction per user (if you see multiple likely scenarios you have to pick one)
- Only sufficiently distinct predictions count as new (I will merge multiple predictions if they are very similar and if their authors agree that they can be merged)
- Once your prediction is registered, you cannot take it back or significantly change it! If new leaks come out that rule your guess out, you are out of luck. Minor adjustments are accepted
- Keep the bickering and ranting to the minimum (but do bicker a little bit so that this thread does not die)
Some background information/leaks
- A completely new chip with a new name (tentatively P1) based on a new prosumer-oriented microarchitecture with significantly higher per-core performance compared to M1 and possibly some new features such as ARM vector extensions (SVE/SVE2), hardware ray tracing and possibly others (@leman )
- The chips in the upcoming 2021 prosumer machines will be based on the next-generation Apple microarchitecture and will probably adopt a name variant based on M2, e.g. M2 pro, M2X etc. (@Fomalhaut , @senttoschool )
- No special chips, just scaled up M-series with more cores (@ader42)
- The performance cores are 4.35Ghz (420 KB per core (performance cores, 252 instructions + 256 data) 256 KB per core (efficient cores, 256 instructions + 128 data) with L2 (18 MB (performance cores) 8 MB (efficientcores) (copied from @Serban55's post)
- M1 variant with the name of M1X (@Lemon Olive)
https://www.bloomberg.com/news/arti...ook-pro-macbook-air-revamps-with-faster-chips
Apple's follow-up to M1 chip goes into mass production for Mac
TSMC-made chipsets to replace Intel offerings in laptops set to launch in 2Hasia.nikkei.com
https://www.chinatimes.com/newspapers/20201026000140-260202?chdtv
This has been almost spot on with what I have been personally thinking. I've been personally calling it "D1" for dedicated, but thinking about it, P would probably fit better. As someone else mentioned, M1 Pro could also work, but in any case, I believe we will get a different naming scheme for pro levels of desktop hardware.
One other thing I'm currently thinking. I think the rumor 8+2 chip we keep hearing about is meant only for the 14 inch macbook pro. I personally believe the 16 inch is coming later this year, but we will see.
In that case I can add your name to the first prediction if you want.
I think the math does not support this though. The rumored Jade C-die should be in the ballpark of 65W cumulative TDP (8*5 for CPU, 20W for GPU, 5W for RAM). This could fit a 14" chassis (30W TDP with throttling) or a larger 16" chassis (70-80W without throttling). Having a larger chip for the 16" would not bring any advantages as it could not reach it's peak performance...
The current 13" 4TB uses a 28watt cpu that goes up to 40 watts. Ming chi Kuo said the 14" will be getting a new thermal design that should allow for more TDP.
”Peak performance” is a slippery concept as it is so dependent on where on the power vs. performance curve makes sense for your product. Wide-and-slow brings better performance per Watt where the wide aspect is applicable, but only there which is why modern processors can change the clocks of for instance different cores dynamically to take advantage of the thermal headroom available when only parts of the whole processor need to run fast. Which makes sense and is pretty nifty. Granularity and speed of response then becomes areas of interest.
Since we are talking about laptops here it’s safe to assume that power draw will be a major consideration, so I doubt that Apple will push clocks higher on a similar process in general. How they manage thermals on a wider, but still strongly thermally constrained chip will be interesting to see if they let us. It’s one area where we might see changes from the smaller mobile parts. The wider your design is, the more you have to gain by managing those resources intelligently.
Which is exactly why I assume that the sustained peak per core will stay around 5W — this seems to be the target Apple has been using in the last 7-8 years for their chip design. And it makes sense for a 8-core processor in a compact laptop chassis.
There is a lot of overlap, true. The key difference is that #3 predicts that it's an M-series chip, but includes both a possibility of M1 or M2, while #5 explicitly says that it's an M1 family chip.
If it ends up to be M1X I suppose those who backed #5 just get extra points
Gotcha.There is a lot of overlap, true. The key difference is that #3 predicts that it's an M-series chip, but includes both a possibility of M1 or M2, while #5 explicitly says that it's an M1 family chip.
If it ends up to be M1X I suppose those who backed #5 just get extra points
For naming, Im holding firm that it won’t be called M2. (maybe P1, maybe M1X, or maybe just M1 and specified core counts)
Mainly because if this chip was M2, what would we call the 8-core M1 successor? (whatever goes into next years MacBook Air) would that be M3 then? I feel it would be confusing if M2 was a 10-core CPU / 32-core GPU, and then they release M3 that’s an 8-core / 8-core again. Wouldn’t make sense.
I think it would be smart to brand the performance chips with at least something (either an added X or use a new letter entirely) to help differentiate them.
8-core M2. Just like the MacBook Air's M1 (7-GPU cores) and the M1 (8-GPU cores) are both called M1.
I've always understand the suggested naming convention to be based on CPU "generation", in the same way as the Apple A-series and Intel Core CPUs. An M2 would be a SoC with a new core micro-architecture (see https://en.wikipedia.org/wiki/Microarchitecture so that we are referring to the same concepts).
Within each generation there will be design variations including number of CPU and GPU cores and other sub-systems. Similar to how we have an Apple A12, A12X and A12Z, or Intel 11th gen Core i7 CPUs with a whole load of variants (https://www.intel.com/content/www/us/en/products/details/processors/core/i7.html).
How Apple will name these remains to be seen. Is could be as simple as "M1 8-core" & "M1 12-core" (for an upper tier M1-based machine). There is nothing dictating that you can't release a new variant of the M1 after launching the M2, or that a smaller 8-core M2 has to be released before a 12 or 16 core variant.
There appear to be a lot of emotive posts with people staunchly defending the possibility a second generation "M2" SoC announcement at WWDC, as well as those decrying the idea.
Given the fact that we expect an A15 in iPhones/iPads in September, with some improvement over the A14, it would be a little underwhelming if the new MacBook Pro SoCs were still based on the current M1 (with added cores). I haven't read anything that provides a definitive answer in either direction - they are both possible.
It would be pretty awesome if the new chips were based on next-gen, arm v9 based A15 cores with a +15-20% bump, but keep in mind that with intel, xeon chips always lagged behind their standard desktop stuff by about a year for every generation.
Nothing is impossible, but it's not unreasonable to think that they will just be scaled up M1's at this point. I do agree that it's disappointing that they are coming out probably 8-9 months later though (July maybe), and the A15's will be out just 3-4 months following that.
It could be due to supply chain issues having nothing to do with the SoC per se. The mini-LED supplier has been reported to struggle for instance.
We'll hopefully know soon enough. From a practical standpoint, the "core generation" probably doesn't matter much if on the same process, there are other aspects of the SoC that matter a lot more, such as number of CPU, GPU and arguably NPU units, how they are utilised, if the on-chip cache subsystem is reworked, reworked or added auxiliary functionality, I/O and networking, what the main RAM solution will be and so on.
What I've been thinking about recently is: who says that there will only been one generation and one microarchitecture? ARM ships multiple "current" architectures that cater to different use cases. I think Apple might do the same. So far, we only saw their mobile architecture (Firestorm+Icestorm+G13 aka. A14 aka. M1), but I find it likely that they will have a "parallel" set of desktop prosumer architecture that is tweaked to deliver higher performance in a higher power envelope.
That's exactly the synopsis of my "prediction": that we will see two "sets" of microarchitectures going forward. A mobile one (for phones, iPads and entry-level Macs) and the prosumer one (for high-end Macs).
If you want to go on the list, you have to pick one
It isn’t likely that Apple will care too much about Aarch64 V9. It is mostly security features that Apple probably wouldn’t use because they already have their own Secure Enclave. The interesting bit is SVE2 which is optional on the V8.5 already. Maybe you were using V9 as a shortcut for SVE2?
Ha, if I had to put money on it, I'd guess that the M1X will be of the same generation core as the A14/M1 (same single threaded performance). Crossing my fingers though since you never know with Apple.What I've been thinking about recently is: who says that there will only been one generation and one microarchitecture? ARM ships multiple "current" architectures that cater to different use cases. I think Apple might do the same. So far, we only saw their mobile architecture (Firestorm+Icestorm+G13 aka. A14 aka. M1), but I find it likely that they will have a "parallel" set of desktop prosumer architecture that is tweaked to deliver higher performance in a higher power envelope.
That's exactly the synopsis of my "prediction": that we will see two "sets" of microarchitectures going forward. A mobile one (for phones, iPads and entry-level Macs) and the prosumer one (for high-end Macs).
If you want to go on the list, you have to pick one
Do we know if Apple is or isn't already using SVE2? Because yeah I tend to agree V9 might not be a big deal for Apple as people hope it is, as Apple may have already adopted most of the speed improvements.
What’s wrong with firestorm cores that would need to be altered to fit a more power-user computer?
I think that Apple would want more single core performance for their pro machines, and it is questionable whether Firestorm can be reliably clocked higher.
I also wonder when single core performance increases are going to slow down. For nearly a decade we were seeing 40-50% increases per year in mobile! (with a few bad years).
Most recently A13 -> A14 was roughly 20%. Still way better than the pathetic intel gains of 5-7% per year, but obviously a lot less interesting than 40% bumps (A12 to A13 for X to 11).
Apple’s CPU performance has gone up 20% per year for a very long time, not just recently.
Awhile back i posted a table I made comparing benchmark scores year-by-year, and it was more or less 20% every year on single core performance (there was a little outlier one year, but I can’t remember which).
I happen to have a relevant graph at hand...
Single-core performance has been improving at a very stable rate.
Single-core performance has been improving at a very stable rate.
Thanks! Looks like a7-a8-a9 is where it deviated a tiny bit from 20ish%, with a smaller amount 2013-2014 and a bitter amount 2014-2015, and since then it’s been like clockwork.