Did you really mean 25W? The most power I’ve seen in powermetrics for the performance cluster is ~15 W. But that is on a MacBook Air.
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.
New cores (A15)
- Thread starter cmaier
- Start date
- Sort by reaction score
If I remember correctly, most I've ever got from the P-cluster was 20 watts. Firestorm in M1 peaks out at 5 watts per core at max turbo.
Apple is not Intel, they don't need E-cores to achieve goo perf-watt. Most real-world parallel workloads are asymmetric, so E-cores have limited scalability. I agree with you that the rumored 8+2 config does sound weird, but at least those rumors were consistent. We'll see.
The lack of a CPU performance gain is a real puzzle.
Apple had a bigger battery, more transistors and a new process to play with, but chose (or were forced?) not to add any CPU grunt.
Maybe Apple judged that a battery life improvement is more likely to get someone to upgrade from an iPhone 12/Pro to a 13/Pro than a CPU improvement (and owners of older phones have enough reason to upgrade already), but I would still have thought even a circa 10% improvement would have made it onto the spec sheet if they could have managed it.
Apple had a bigger battery, more transistors and a new process to play with, but chose (or were forced?) not to add any CPU grunt.
Maybe Apple judged that a battery life improvement is more likely to get someone to upgrade from an iPhone 12/Pro to a 13/Pro than a CPU improvement (and owners of older phones have enough reason to upgrade already), but I would still have thought even a circa 10% improvement would have made it onto the spec sheet if they could have managed it.
I think there is a good explanation if one looks more closely. The new, brighter OLED display likely uses more power, and their CPUs are already more than fast enough. At this point I expect that A15 brings minor improvements, but is clocked lower than A14, for slightly lower power usage. Ultimately, this will be revealed once we get the benchmarks and the power consumption data. It is also possible that the CPU itself is not changed, just Neural Engine/GPU/Cache.
I did mean that, although now that I've rechecked since I was trying to remember numbers from months ago, I was wrong. It's actually ~21W. I also misremembered the figure for the E cluster - I can only make it use about 1300mW, not 2500.
I also am testing on an Air. You may have to run powermetrics with the option "-i 1000" (1000ms polling interval) or less to actually see this, because an 8-thread load will make the Air start backing off full frequency/power in only a few seconds.
The core eater I'm using is just "yes >/dev/null" in a terminal window. You can start eight copies all at the same time with "yes >/dev/null &; yes >/dev/null &;" etc., and later kill them all simultaneously by closing that terminal window.
It should be possible to make the cores use more power with a real power virus - some vector FP ought to do the trick.
It's still 5nm as the A14. It might be an updated node though.
You're probably right, I just thought that if Apple could bring even a small performance increase, it would.
I do think it's ultimately a marketing rather than technical decision that it decided not to do so.
"It appears Apple has not changed the CPU much this generation. SemiAnalysis believes that the next generation core was delayed out of 2021 into 2022 due to CPU engineer resource problems. In 2019, Nuvia was founded and later acquired by Qualcomm for $1.4B. Apple’s Chief CPU Architect, Gerard Williams, as well as over a 100 other Apple engineers left to join this firm. More recently, SemiAnalysis broke the news about Rivos Inc, a new high performance RISC V startup which includes many senior Apple engineers. The brain drain continues and impacts will be more apparent as time moves on. As Apple once drained resources out of Intel and others through the industry, the reverse seems to be happening now.
We believe Apple had to delay the next generation CPU core due to all the personnel turnover Apple has been experiencing."
semianalysis.substack.com
M1 Exploration - v 0.70
drive.google.com
We believe Apple had to delay the next generation CPU core due to all the personnel turnover Apple has been experiencing."
Apple CPU Gains Grind To A Halt And The Future Looks Dim As The Impact From The CPU Engineer Exodus To Nuvia And Rivos Starts To Bleed In
Apple's leading CPU team looks to be struggling as they have now delivered no CPU gains for the first generation ever. Apple likely had to delay their next core architecture due due to personnel overturn from the Nuvia exodus. Furthermore, with a new exodus surrounding Rivos, the future doesn't...
semianalysis.substack.com
M1 Exploration - v 0.70
M1 Explainer 070.pdf
drive.google.com
We will know more once we have detailed benchmarks and analysis.
Oh, it’s all marketing. Also, Apples performance comparisons are usually very conservative.
That’s pure speculation. Another theory is that Apple went for efficiency gains rather than performance. I don’t think anyone is clamoring for more CPU speed in an iPhone."It appears Apple has not changed the CPU much this generation. SemiAnalysis believes that the next generation core was delayed out of 2021 into 2022 due to CPU engineer resource problems. In 2019, Nuvia was founded and later acquired by Qualcomm for $1.4B. Apple’s Chief CPU Architect, Gerard Williams, as well as over a 100 other Apple engineers left to join this firm. More recently, SemiAnalysis broke the news about Rivos Inc, a new high performance RISC V startup which includes many senior Apple engineers. The brain drain continues and impacts will be more apparent as time moves on. As Apple once drained resources out of Intel and others through the industry, the reverse seems to be happening now.
We believe Apple had to delay the next generation CPU core due to all the personnel turnover Apple has been experiencing."
Apple CPU Gains Grind To A Halt And The Future Looks Dim As The Impact From The CPU Engineer Exodus To Nuvia And Rivos Starts To Bleed In
Apple's leading CPU team looks to be struggling as they have now delivered no CPU gains for the first generation ever. Apple likely had to delay their next core architecture due due to personnel overturn from the Nuvia exodus. Furthermore, with a new exodus surrounding Rivos, the future doesn't...
M1 Exploration - v 0.70
M1 Explainer 070.pdf
You also omitted GPU cores and the rest of party inside SOC. GPU core is about 1.5 W.
If I remember Firestorm should draw no more than 5 W per core, probably closer to 3.5, but I cannot give a source now. Probably Anandtech.
Max power draw for fully juiced M1 should be 20-25 W, so hypothetical 8P 4E 16GPU chip should fit into 45 W envelope at 5 nm node. Such 8+4 config was rumored long time ago.
Seems that they prioritised the GPU, with a whole extra core that should be significantly more powerful? That was also the rumour for the (probably A15 based?) M2, an extra 2 GPU cores (up from 8 to 10) is the most significant upgrade over the M1.
Agreed that Apple doesn't need E-cores for good perf/watt, but I was making the point that they offer truly great perf/W. That's why the rumors of moving to two in Macs haven't made sense to me; if anything Macs have more demand for low intensity background processing. (iOS is quite hostile to background activity.)
I don't think they should go way above four either. As you say, real world loads don't favor that. Most software which would do well on a sea of E cores is better off being offloaded to GPU compute or the Neural Engine.
8+2 makes sense for scalable chiplet, but this base 8+2 would be probably crappy.
A chips are 2+4
M1 is 4+4
So next logical step is 8+4
Why do you think it would be crappy?
No disagreement here. But the thing is, all chip-related leaks so far were very accurate. I was also skeptical when I heard about a 10-core M2 GPU (that's a weird number), but now we got a 5-core A15 GPU, so that rumor is pretty much confirmed.
It’s the plus node, for sure.
If the performance indicators are accurately telling the story, then it has to be clocked lower for battery life. The same cores are going to end up in Macs in the next year or so for sure, and I’m sure at the same clock there must be a single core performance improvement.
Here’s a random thought - could the A15 cores be, essentially, simply revs of Firestorm and Icestorm, with the same microarchitecture and some minor improvements? Maybe they increased the sizes of the caches and SLC - that could account for, say, 10% of the increased transistor count. Then you have the extra GPU core, for another 5% - maybe more if the GPUs are redesigned. Maybe the neural engines were completely redone. Tough to figure out how to add up to 3 billion more transistors. But, anyway, perhaps they’ve decided to decouple the i-device and Mac CPUs, so the M2 is not using the same cores as the A15? It would make a certain amount of sense - the A-series now go only into phones and lower-end ipads, where the CPU performance has more or less become “more than good enough” and where the focus is on other functional blocks (GPU, neural engine, image processing, etc.) to support things like the upcoming VR goggles, etc.
Anyway, just random thoughts. I’ve certainly worked on a lot of chips where we got 10-15% improvement without ripping up the core microarchitecture. And if there was an entirely new design, I feel like Apple might have spent a bunch of time talking about how great it is. Now I feel like we will get our next CPU brag from Apple in a few weeks at the mac announcement.
Anyway, just random thoughts. I’ve certainly worked on a lot of chips where we got 10-15% improvement without ripping up the core microarchitecture. And if there was an entirely new design, I feel like Apple might have spent a bunch of time talking about how great it is. Now I feel like we will get our next CPU brag from Apple in a few weeks at the mac announcement.
Last edited:
You stated Yourself (or cmaier stated) that 2 efficiency cores could be too small number for efficient handling of low priority threads.
And You both are a kind of gurus on this forum for me.
That's my thought as well. Notably, A14 introduction was done similarly, without much fanfare. The more technically interesting stuff was presented later, during the Mac event. I think we will have something similar here. Basically, Apple's message is "A15 is still years ahead of the competition, and now you also get a better battery", enough for an average iPhone buyer. Their presentation are not necessarily targeted at the tech geeks...
Another thought: if the CPU cores are largely unchanged and all the difference boils down to more cache and maybe a different GPU, why would they claim a "new" CPU with new codenames? Avalanche/Blizzard are official monikers at this point, straight from the Apple's headers.
Regarding M2: the rumours of a 10-core GPU in M2 is consistent with a 5-core GPU in the A15. I don't think that later fact was ever mentioned by any leakers, and it certainly gives credibility to earlier leaks about M2.
I don't know if two efficiency cores is not enough. Asymmetrical CPUs on desktop is still a very new thing. Regardless, 2 or 4 E-cores won't make a huge impact in any case.
Didn’t know about avalanche/blizzard.
I dunno, I’m stumped. Not enough information available.
Exactly, right now the speculation is pretty much meaningless. We need to wait until the devices get into the hands of the serious reviewers. And anyway, the Mac vent is just couple of weeks away.
8+2 had to be weird indeed for me at a first glance, then the idea of chiplet and x2 or x4 sets has materialized, which is reasonable: 16+4 and top notch desktop 32+8.
On the contrary, new Intel Alder Lake chips have relatively small number of performance cores and numerous efficiency ones.
I am not expert in hardware nor software engeneering, my occupation is medical, but many years ago, as a young man, I had Atari ST and I programmed some stuff in 68k assembler, just for fun. So all that Mxx story is a good fun for me and resembles my old days, since the architecture of 68k was so clean and logical compared to x86 and Mxx gives me similar impression.
Sorry for off-topic
On the contrary, new Intel Alder Lake chips have relatively small number of performance cores and numerous efficiency ones.
I am not expert in hardware nor software engeneering, my occupation is medical, but many years ago, as a young man, I had Atari ST and I programmed some stuff in 68k assembler, just for fun. So all that Mxx story is a good fun for me and resembles my old days, since the architecture of 68k was so clean and logical compared to x86 and Mxx gives me similar impression.
Sorry for off-topic
