To you perhaps
No one said you did. You “randomly” picked a massively overclocked chip. My point was the average is a realistic picture of the chips performance. Not a massively overclocked one.
no
I said most, not all tasks.
All the best.
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.
Why Intel and AMD don't make chips like the M2 Max and M2 Ultra
- Thread starter sack_peak
- Start date
- Sort by reaction score
To you perhaps
No one said you did. You “randomly” picked a massively overclocked chip. My point was the average is a realistic picture of the chips performance. Not a massively overclocked one.
no
Translation: single core scores reveal the giant hole in my reasoning, therefore I will pretend they don’t matter.
That's a totally different argument, and a definite win for the Apple CPU, but you do lose performance because of that. I have no problem saying that!
It's a tradeoff. If you want a portable device with massive battery life, you get an apple processor machine, but if you want real world performance, I'd take an intel (or AMD) machine every time.
That's why I carry an iPhone in my pocket, an apple Watch Ultra on my wrist, and I'm sitting at an intel based machine.
Cool!! I'd love to have a machine that could do that.
No it is not. There are use cases where it's important. Some algorithms are difficult to parallelize but many typical tasks get parallelized implementation. Typical tasks that need processing power like image processing, machine learning, games etc. all use multiple threads/cores. It is true that many apps do not use multi-threading but in most cases the reason for that is that they do not need performance in a first place.
Nonsense. It’s important for pretty much everyone. If single core performance improves, all tasks see the benefit. Multi core performance improvements need adaptation to software. Sometimes it can be done, sometimes it’s much tougher.
Yeah, all tasks benefit. For example, calculator app: instead of 4 ms will only take 2 ms to do the calculation. Who cares (as long as it's below 500ms)? Same is true for things like e-mail, notes, messaging and on and on.
What?
All Tasks benefit. Video editing, encoding, games, spreadsheets etc. Everything, literally everything benefits from faster single core performance.
All of the tasks you listed use multithreading thus it's the MC performance that matters to them not single core.
Lol, those tasks use both single core and multi core performance.
If Single core performance had been improved at the rate it was from the 60s to the mid-2000s, we’d probably never have been talking about multi core consumer machines. It was a work around for their (Intel/Amd) inability to continue scaling single core performance. If I had the option of one 10ghz cpu, or two 5ghz cpus. I would always take the single, faster cpu. It would in almost all cases, give significantly better performance.
Can you tell me the next goalpost spot so I can prepare please? You surprised me by shifting this one.
I never shifted. You just don't know what your are saying.
You do know that TSMC 3nm gives more advantage than TSMC 5nm? If you wanna compare the performance, it's important compare on same lithography or otherwise, it will be totally unfair to compare. It's like you are comparing between Intel 14nm vs 10nm and yet, 10nm has better performance. Also, Apple is the one who always take advantage with nm unlike other companies. What would you say when Intel gets 3nm CPU, huh? Or you are saying, Apple needs to use 3nm to compete and come close to Intel 7 CPU which is TSMC 7nm. How funny.
So you’re saying I’m not being fair to Intel because Intel’s failure to deliver a node on par with TSMC gives Apple an unfair advantage to the tune of 500% less power consumption?
‘Kay
What’s next, transistor count?
lol, A17 Pro may close to 12900K but it's not Apple Silicon's advantage. It still does not change the fact that Apple needs 3nm to compete with old Intel 7 or TSMC 7nm which is what you are trying to say.
Last edited:
This take of Intel applies prior to 2020.
Intel's failure to keep pace with TSMC is likely a OpEx related to deliver stockholders' value.
Seeming they're the dominant PC chip maker with all PC OEMs buying from them since 2006 it has less incentive to be leading edge unlike AMD and now Apple.
Intel likely sets engineering compensation standards so they do not lack for talent.
Apple's unlikely to slack off as their market is the top 20% of any given market. So they must be at next die shrink before anyone else. Their product refreshes reflects that starting with every iPhone.
Their performance per watt mantra is forced upon them because majority of all their profits is derived from iPhones.
So any devices derived from iPhone R&D will be centered around that.
Persons who need outright raw performance for a desktop, workstation or server will not be served in places where power is cheap or have no hard limit.
But the market for raw performance users at all cost s shrinking YoY so Apple's ahead of that curve.
Last edited:
That is all very lovely, and nobody is refuting Intel's absolute performance advantage (at an extreme cost in power consumption), but may I remind you that the point under the discussion was this one
The simple fact is that neither Intel nor AMD are currently able to make to make a CPU core that can get to ~ 3000GB6 points with 5-6 watts. And believe me, they would really love to as achieving this would easily improve their server performance by 50-80% for the same die area. Currently, they have to work around by increasing either the die area, power consumption, or both.
Single-core performance is the gateway to total performance. If it means nothing to you you are either ignoring the big picture or simply not arguing in good faith.
You can! Just buy a dozen or two of 13900KS and a few buckets of liquid nitrogen. Would make a fun little workstation.
Would it be safe to say it isn't possible to get that kind of performance at that wattage with x86?
I don't see why it wouldn't be possible. It is likely that x86 is handicapped with the more expensive decode, but micro-op cache goes a long way to addressing this issue.
Intel an dAMD have pretty smart folk, and if they haven't been able to do it yet, why would we think they can at all?
They have been pursuing other design strategies, and it has been working well in their market segment. But the race for more IPC is on. AMD seems to be very serious about it, and Intel is rumoured to substantially overhaul their basic approach over the next few years, so who knows.
AMD devices are such a joy to use. While the ROG Ally doesn't have great battery life, it's so silent I would even forget it's there if I weren't in front of the screen. And at 10W, it's dead cold in your hands.
Since when did a thread around here only stay on the original topic.
And I wasn't arguing against that point. It's just not an important point to me. Performance means more to me than efficiency.
The gateway, hmm. Anyway, I'm arguing in just as good a faith as you are, we're just looking at different things. A single core is no gateway to performance, a computer is NEVER just doing a single thing. The best gateway to performance is having multiple cores, even slower cores are okay if there's enough of them, and a good OS. I never look at single core performance when I'm buying a machine, or optimizing software to run faster, or basically anything else I do. I'm not a researcher or whatever you do, I'm an IT generalist (manager), I care about my users getting their work done on time, and that's pretty much it.
I'll pass, the 14th gen CPU is what I'd go for now and reliability is probably not that good on such an extreme overclocking. It would be interesting to play with though.
