I believe there is also the matter of the low-level IPC via interrupts, on which Apple diverges from the official spec. I remember reading that they have a custom fast interrupt protocol.
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.
Qualcomm Says New Snapdragon 8 Elite SoC Has 'World's Fastest Mobile CPU'
- Thread starter MacRumors
- Start date
- Sort by reaction score
I believe there is also the matter of the low-level IPC via interrupts, on which Apple diverges from the official spec. I remember reading that they have a custom fast interrupt protocol.
I too design CPUs faster than Apple...just gimme another decade to work out the efficiency issues
It doesn't.
The key question is: does it accomplish its higher multi-core score at similar thermals as A18 Pro? Its higher clock and higher core count suggest that it draws more power.
I mean, that's basically the three contenders, right? AMD isn't even playing in the efficiency game, Intel's Lunar Lake is about a generation and a half behind Apple, and Snapdragon may be slightly closer.
And Apple's design — M4 / A18 — is only half a year old, so they perhaps aren't that far ahead internally.
LMAOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
Samsung Dex is pretty much this idea and it's executed quite well. We all know Apple could to it even better with mobile chips now technically fully capable of running their desktop OS very well, but all their products are worse than they have to be because they want to sell you a gazillion other devices that do 99% of the same thing since that's so Mother Nature.
Yep.
I think Ubuntu Touch was first to the idea. I'm not sure if they had a name for it.
Windows 10 Mobile then offered Continuum, which was almost useful except 1) Windows on ARM (10 was NT, not CE, unlike Windows Phone 7) at the time was quite limited in what kinds of apps ran, and 2) by the time Windows 10 Mobile came out, it was already almost dead.
Samsung offers DeX. That's perhaps the best execution of this.
No, Apple usually tests their chips running in production, i.e. in the device. Usually Apple claims tend to be correct, even when nobody believed their claims on the M1, when it came out. Then it smashed all tests and people started wondering if the tests were wrong.
S8E beats the A18 pro in efficiency since it’s using a more refined N3E node from the experience TSMC gained from fabbing A18 pro.
So much for “Apple design” nonsense that fans were barking about. I told you all, chip design is intellectually child’s play and unimportant. The fabrication method is harder and important.
I’m not giving props to QC, since this is all due to TSMC. QC and Apple deserve none of the praise for the chips TSMC fabs.
So much for “Apple design” nonsense that fans were barking about. I told you all, chip design is intellectually child’s play and unimportant. The fabrication method is harder and important.
I’m not giving props to QC, since this is all due to TSMC. QC and Apple deserve none of the praise for the chips TSMC fabs.
The somewhat latent statistician and mathematician in me is dying on that one. Literally urgh.
Qualcomm has more cores, which means it can operate on lower frequency at the same wattage. You are ignoring the basics of computing and completely misconstruing reality.
What I find more interesting is that despite the core count advantage Elite 8 needs 40% more power to gain a 12% multicore performance lead on Apple. That does not look good for Qualcomm.
If your first and third paragraphs are true, your second one cannot be true, as no one would be able to design a better soc.
Ergo you can’t slam Apple as it would be impossible to do better.
There’s no way to screw up a pepperoni pizza “design” from dominos.
I’m not dinging apple, but you shouldn’t praise them either much like you wouldn’t praise an 8-year-old for designing a pepperoni pizza from dominos.
This is uneducated spiel. The PPW curve looks like a lognormal curve no matter what the node or design is. At a particular point, you tradeoff in power and performance flips where you need disproportionately more power to gain slightly more performance. Apple would have the same “problem” if they allowed higher power (Software limitation).
This inflection pt is set by the node, not the “design”.
As I have said recently, this is all lies.
Apple M3 (and all the other M chips) has both P and E cores. E cores have roughly 1/3 the performance of the P cores, and yet they are made on the same process. That alone is enough to prove that your statement is a giant steaming pile of BS - here we have two cores on the same process with drastically different performance and efficiency. But let's look a little more, comparing two different designs from two different companies.
Intel Lunar Lake compute tile is TSMC N3. So is Apple M3. HIQP would have us believe that the N3B process used for Lunar Lake is "more refined" because it's newer (it's not, actually - improvements made in a single process are continuous, and applied to all chips being made with that process). But if he were right, then that would just give an advantage to the Intel chip.
The Core 9 288V has a nominal maximum clock of 5.1GHz, which it may actually attain during a GB6 SC run. The Apple M3 has a nominal max clock of 4.05GHz, which it will also likely reach during an SC run.
Apple M3P cores score around 3100, according to the GB site. Dividing by 4.05GHz, you get about 765 points per GHz, which gives you a sort of rough idea about IPC. According to notebookcheck, the 288V scores about 2900. (The GB site shows many scores, mostly ranging from 2400-2800, but let's give it the benefit of the doubt.) Dividing by the nominal 5.1GHz boost clock, you get a bit under 570 points per GHz.
So that's pretty much all you need to know - ON THE SAME EXACT PROCESS running at lower clocks, the M3 scores better than the Lunar Lake. CPU design matters.
But of course, that's not really the whole story - there's more. POWER EFFICIENCY is a very big deal, for a lot of reasons, but maybe most obviously because if you're trying to run multiple cores, then you can run more of them faster without melting them. The M3 P core is significantly more efficient than the Lunar lake core. I'm too lazy right now to fish out those numbers, but feel free to do the homework yourself, or perhaps someone here can supply them.
The shape of the curve is one thing. The parameters of the curve is something else entirely. Different number of cores, different voltage, different chip design — all these things factor in, and this is what you are blatantly ignoring.
Also, you seem to confuse logarithmic and log-normal. So much to "uneducated spiel". Please keep in mind that you are talking to practicing statisticians and industry experts here. I don't know what your degree is in, but it is certainly not stats or IC design.
Last edited:
This doesn’t address my point at all I’m afraid.
Think of it like designing a pizza. TSMC is the oven. Anyone can buy an oven and put the pizza in, but it’s the quality of the ingredients and the dough maker that determines the quality of the pizza. TSMC can’t really screw an oven up, hence the difference in cpus on the same node is due to Apple being better engineers.
Hope this clears up your confusion.
Last edited:
“Different number of cores, different voltage, different chip design”
This is on par intellectually with ordering pizza from dominos.
“Also, you seem to confuse logarithmic and log-normal. So much to "uneducated spiel". Please keep in mind that you are talking to practicing statisticians and industry experts here. I don't know what your degree is in, but it is certainly not stats or IC design.”
I mistyped it. I meant to say logarithmic. I typed this quickly on my iPhone. Regardless, statistics isn’t related to IC designs. Your degree is irrelevant.
No I entirely agree with you. There's a ton of weirdness under the hood of AS. For better and for worse.
Yes, logarithmic. I made a small mental fart while typing this quickly on my phone. Way to ignore the overall point that Apple has low quality engineers and sucks at engineering.
Hmmmm
That does sound like a ding.
It might be easier if you think of it like designing a pizza. Intel covers their dough in botulism and cheese they found in the trash. Apple uses only farm fresh ingredients.
My first degree is microelectronics. My second one is mathematics (with statistics). Mine are not irrelevant.
You couldn't be more wrong. It is everything to do with statistics from statistical static timing analysis to manufacturing process monitoring. Literally the entire stack is built on statistics!
Hell we even have statistical models for memory errors and key extraction via timing analysis!
So if you want to talk statistics, please do. You do not speak like a statistician or anyone with statistical knowledge as you would be carefully avoiding making certain statements and using certain language in your statements and analysis.
By any chance do you work at Dominos?