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.
Could we see M3 before A17?
- Thread starter senttoschool
- Start date
- Sort by reaction score
Hoping it will be used for M3 Ultra & M3 Extreme Mac Pro workstations that will debut at WWDC 2023; with higher clocks, LPDDR5X, hardware ray-tracing, & ASi GPUs...
WWDC 2023 will be the three year mark since the Apple silicon transition was announced and the Developer Transition Kit (A12Z SoC in a Mac mini chassis) was available; Apple has had plenty of time, they need to release an ASi Mac Pro ASAP...
It might not be M3 but how does this prove that it’s not M3?
15-Inch MacBook Air With M2-Like Chip Spotted in Developer Logs
An unreleased 15-inch MacBook Air with a processor "on par" with the M2 chip has been spotted in App Store developer logs, according to...www.macrumors.com
I highly doubt it.
Are we expecting more cores for M3?
It'd make more sense to me if M3 had the exact same number of cores as the M2.
M1: New node, new architecture
M2: More cores
M3: New node, new architecture
M4: More cores
M5: New node, new architecture
M6: More cores
Last edited:
I did think of that as well. His report says M2-like and refers to core count. For low end M3 I wouldn’t expect a change there anyway.
That combined with reported N3 volume production late 2022 + Apple only N3 customer currently + reported device ID is “Mac15,3” still has me hoping for an M3.
I would hope we see more cores for the M3 Max (and therefore also for the M3 Ultra & M3 Extreme), a 16-core CPU (12P/4E) in the M3 Max would give us 32 cores (24P/8E) in the M3 Ultra & 64 cores (48P/16E) in the M3 Extreme...?
15 inch MBA is running macOS 14 which wont gonna release until later this year. Since it's more likely M2 chip, this rumor points out that M3 devices wont be released in June and therefore, M3 will not come out sooner than A17.
So how are you so sure that they won't launch the MBA 15" on Ventura?
I think the tik tok approach would work well. Apple can drastically improve perf/watt with a new node and architecture. Then they can improve the MT/GPU performance in the year after. They did this with M1 and M2. AMD and Intel will sometimes do this too such as Alder Lake to Raptor Lake.
I personally would not want more P cores. I feel like improving the Neural Engine and GPU are more important going forward. AI applications are now more important than traditional software.
Asymmetrical SoCs then...
One M3 Max (12P/4E) & one GPU-specific (GPU/Neural Engine) SoC paired for an Ultra. two Ultras for an Extreme...
Same GPU-specific dies could be used for ASi GPUs...
The 15 MB Air rather needs lower power draw than more cores and performance (as does iPads!). Makes perfect sense with a power efficient 3nm M3 especially if the M2 sometimes throttles in the current Air/iPads.
I consider it very unlikely that a 3nm M3 will draw less power under load than the M2. And idle power of those chips is almost null anyway.
Fun fact: recent Apple patents describe a memory folding scheme, where physical RAM is moved around so that memory controllers can be powered down when bandwidth requirements are low. They also have patents describing how cache can be used as RAM, further reducing the need to communicate with external RAM. I have no idea whether any of those are implemented in the M2 family.
So a node shrink is not reducing the power efficiency? I thought that was a selling point for a node shrink. Please explain.
First, a node shrink will increase power efficiency.
Second, Apple develops chips to fit a certain power envelope. For example, the M1 CPU maxes out at around 10w. The M3 will likely max out at around 10w too. It'll just be a lot faster at 10w.
Nikkei has always been highly accurate in my opinion. They indicated N3E for A17 and M3, which means M3 is coming tail end of 2023 or early 2024.
asia.nikkei.com
Most consumers don't replace their Macs often enough to justify a 12-month refresh cycle.
Apple to use TSMC's next 3-nm chip tech in iPhones, Macs next year
Move will mark industry's first adoption of updated production technology
Most consumers don't replace their Macs often enough to justify a 12-month refresh cycle.
Does it make sense for Apple to use the same node for Macs and iPhones?
www.semianalysis.com
TSMC’s 3nm Conundrum, Does It Even Make Sense? – N3 & N3E Process Technology & Cost Detailed
Shrinking finally costs more, Moore’s Law is now dead in economic termsA couple of weeks ago, we were able to attend IEDM, where TSMC presented many details about their N3B and N3E, 3nm class…
www.semianalysis.com
The time frame with looking at October 2023 to offer first M3 devices is only 16 months from M2 introduction. That assuming the A17 does its intro in September. Given this is the very first 3nm SOC with its own production challenges to overcome and grow.Nikkei has always been highly accurate in my opinion. They indicated N3E for A17 and M3, which means M3 is coming tail end of 2023 or early 2024.
Apple to use TSMC's next 3-nm chip tech in iPhones, Macs next year
Move will mark industry's first adoption of updated production technology
Most consumers don't replace their Macs often enough to justify a 12-month refresh cycle.
We went though many node shrinks in the last decades, and yet computers do not consume less power. In fact, they often consume more, because performance demands have changed. A top GPU in 2003 was made on 150mm2 node and consumed 50 watts. A top GPU today is made on 5nm node and consumes 450 watts. Of course, the today's GPU has 600x as many transistors and is three times larger.
Apple has been using die shrinks to build more complex devices with better performance. As @senttoschool explained, they have a certain design target (5 watts per CPU core on Apple Silicon for example) and try to make the fastest hardware possible while staying within that target.
Now, one could of course use a new node to make the same chip, which would have the same performance but consume less power. But what would be the point of such a product? You might get 15-20 minutes more of battery under full load, if at all. That's a lot of money and resources spent on a product no customer is interested in. In fact, Apple should be interested in further increasing the performance potential of the chip, so that their desktop and laptop offerings are more competitive.
Last edited:
Yes, but in my opinion M2 performance is sufficient for most people so it is better to use the improved efficiency on longer battery life than a speed bump. "M2-like" might just be that. Minor speedup and longer battery life. The point of a portable device is to be free of an electrical outlet.
M3 Pro, Max, Ultra is for raw compute performance not the M3.
I'd question whether Apple would build up stocks of chips. For a few reasons:
- Chips (even Apple Silicon) are perishable. M1Pro/M1Max were/are heavily discounted since the M2Pro/Max release in February. Just 15 months.
- Tim Cook is a short-term financial management kind of guy. And he doesn't want working capital tied up in inventory - especially unfinished product.
- Tim is focused on Apple's value. He's not going to look good if mac sales take a big hit, like they have for what probably is the last two quarters (still waiting on CY23Q1). He doesn't want to upset the market, and the market is anticipating M3 hardware, and the market will probably love the M3.
Last edited:
I doubt that the 15-20 minutes of extra battery life under load will be worth it. I just happened to write a post in another thread on this very topic where I lay out my reasoning: https://forums.macrumors.com/threads/m2-macbook-air-really-nice-machine.2377091/post-32096078
If there is a foundry bottleneck you kind of have to stockpile if you want to satisfy the demand. I can imagine it’s a delicate balancing act, with many risks. I’m glad it’s not my job to make these decisions…