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.
Intel 12900HK is faster than M1 Max for both single and multi cores.
- Thread starter sunny5
- Start date
- Sort by reaction score
Very interesting read-up. A couple of things I liked:
(And the M1Max was more than twice as fast.)
Of course, that's specifically for ray tracing. The writer explains carefully that the algorithms need to be finetuned to keep the processors going. So it's a very specific workload. But still! All that, and on battery? Amazing.
This is the way it has always been with Apple fans. I've been involved with Mac versus PC discussions during the comp.sys.mac.advocacy newsgroup heyday and there's nothing new today that I didn't see back then. I'm amazed at the blind loyalty so many people have to Apple.
That said I like Apple products but use whatever tool is right for the job. I use both Macs and PCs and have no loyalty to one over the other.
I am one of the people who said that I'll wait until there's a shipping product before I'll believe the results. Not because I am an M1 advocate but rather I've seen so many claims which never panned out. Likewise I dislike Geekbench because it is only tells us how fast a system can run Geekbench. IMO it has little real world value.
I experienced it, so sorry, you have no say in the matter. I've never had an intel laptop that needed to throttle as much as it could -- because it had a fan.
Go out and find it then and leave the rest of us the **** alone.
I'm surprised at Qualcomm. I would have expected them to come up with the good ARM desk/laptop processor. They got complacent or something, IMO.
Qualcomm is years behind Apple. They relied on reference ARM cores IIRC until around 2016, which is also when they started shipping 64-bit ARM chips. Meanwhile Apple moved to 64-bit ARM in 2013, and has been designing their own uArch for longer than that.
But you are right that they don’t have a lot of incentive. Apple doesn’t sell it’s SoCs, Samsung shut down their chip design team and are only now looking to spin it back up again (based on rumors / job listings). If you want to buy SoCs for mobile devices right now as a phone or tablet maker, I’m not even sure where you get them except from Qualcomm. Qualcomm has seemingly been more interested in rent extraction from their cellular patents than advancing state of the art in SoCs.
and if I did what do you think your comment would make me feel like?
I lied, I’m not sorry.
this is probably the post that is the most disconnected from reality in here.
Moreover if you even read half the posts from people around this forum you’d know that most people here aren’t coffee shop dwellers or youtubers. Many people here are programmers, engineers, videographers, musicians and many other vocations that you dismiss.
Which, I find ironic that you are so dismissive of Macs capability of doing “real work,” yet you praise gaming as if it’s somehow less infantile.
But that is misplaced value based only on purchase price, not total cost of ownership. IBM estimates that moving to MacBook’s saved them approx $400 per user and many other enterprise customers show the same:
Total cost of ownership: Mac versus PC in the enterprise. Apple TCO
If the upfront cost is more, how can Mac be less expensive than PC? We examine the numbers and put the Mac versus PC in the enterprise myths to bed.
Can You imagine that Intel sell ONLY laptop/computers not CPUs/Chipset ?
Yes this is one of the main reason why apple-intel comparisons are pointless.
So Intel shares its technology with other manufactures. Doesnt force them to use/buy
special motherboard, drives or memory etc.
Apple products are "closed". You must use only apples components/computers.
We cant even say what is cost of M1 chips. Without knowing this parameter it is
difficult to compare. Apple hidden real prices in the cosmic prices of the memory or
drives upgrade. Probably with high service cost (M1 are not repaired because it is SoC).
All Intels CPUs comes with three year of warranty. Apple say "pay more for extended warranty".
Many people pay more for larger memory/drives for future use because M1 is not
upgradable. But years after years memory/drives are getting cheaper. You can "invest" today
1000$ for 2 TB SSD. This same SSD will cost about 500$ after a year. You dont need it now
but must buy for the future. Yes this is hidden cost of M1 technology.
Yes this is one of the main reason why apple-intel comparisons are pointless.
So Intel shares its technology with other manufactures. Doesnt force them to use/buy
special motherboard, drives or memory etc.
Apple products are "closed". You must use only apples components/computers.
We cant even say what is cost of M1 chips. Without knowing this parameter it is
difficult to compare. Apple hidden real prices in the cosmic prices of the memory or
drives upgrade. Probably with high service cost (M1 are not repaired because it is SoC).
All Intels CPUs comes with three year of warranty. Apple say "pay more for extended warranty".
Many people pay more for larger memory/drives for future use because M1 is not
upgradable. But years after years memory/drives are getting cheaper. You can "invest" today
1000$ for 2 TB SSD. This same SSD will cost about 500$ after a year. You dont need it now
but must buy for the future. Yes this is hidden cost of M1 technology.
Just saw this, the article at hand: https://www.anandtech.com/show/17024/apple-m1-max-performance-review/5
I just don’t get what’s the whole ordeal here, in none of the CPU benchmarks does the intel beats neither the M1s 8+2 CPU (neither the AMD‘s for that matter). Not only that, in some benchmarks it scores close to three times the score of an Intel 11980HK (65W), almost double of a 11900K, it is just shy below an AMD Ryzen 5950X… I already forget what was the TDP of the M1, 30W was it to boot?
Honestly I don’t know what people want or what are we discussing anymore… sure the AlienLake from 2030 will for sure beat this, fine, I’ll give it that.
For Alder Lake to beat this it would have to see a sort of doubling performance from current gen, something that I have never seen happen often to be honest.
Your recent replies pointed me back to this post and honestly? You trolls need to come up with better material. Anyone who's actually used a M1 Air to run heavy loads knows that you're fibbing, or basically haven't owned a laptop before.
But let's look at objective numbers, shall we? Because I have actually measured how much my M1 Air throttles using powermetrics, a power and performance monitoring tool Apple built into macOS, and Hot.app, a temperature monitor.
M1's performance cores run at a maximum speed of 3.204 GHz. If I start up an all-core CPU load while the CPU is cold (I'm using eight threads to make sure the efficiency cores are loaded up too), P cores stay at 3.204 for 20-30 seconds until their temps rise a bit, then gradually ramp down clocks to keep temps under 100C. After 10 minutes or so, the system finally reaches equilibrium with the P cores at 2.1 to 2.2 GHz, about 2/3 of maximum frequency. (E cores stay locked at their max freq, 2.064 GHz, the whole time.)
So: In the Air, M1 CPUs throttle back to about 2/3 of peak. That's perfectly acceptable, if you ask me, because it's a number Intel expects you to deal with even on 125W fan-cooled desktop CPUs:
Intel® Core™ i9-11900KF Processor (16M Cache, up to 5.30 GHz) Product Specifications
Intel® Core™ i9-11900KF Processor (16M Cache, up to 5.30 GHz) quick reference guide including specifications, features, pricing, compatibility, design documentation, ordering codes, spec codes and more.
Max single core turbo is 5.30 GHz. Sustained all-core speed while running at 125W TDP (the "base" frequency) is 3.50 GHz. That, too, is about 2/3.
If you run this exercise with mobile Tiger Lake CPUs, you will find that 2/3 is a bit optimistic for a lot of them. Even the ones expected to be used with fans.
I am not lying at all, and personal attacks are definitely not cool.Your recent replies pointed me back to this post and honestly? You trolls need to come up with better material. Anyone who's actually used a M1 Air to run heavy loads knows that you're fibbing, or basically haven't owned a laptop before.
But let's look at objective numbers, shall we? Because I have actually measured how much my M1 Air throttles using powermetrics, a power and performance monitoring tool Apple built into macOS, and Hot.app, a temperature monitor.
M1's performance cores run at a maximum speed of 3.204 GHz. If I start up an all-core CPU load while the CPU is cold (I'm using eight threads to make sure the efficiency cores are loaded up too), P cores stay at 3.204 for 20-30 seconds until their temps rise a bit, then gradually ramp down clocks to keep temps under 100C. After 10 minutes or so, the system finally reaches equilibrium with the P cores at 2.1 to 2.2 GHz, about 2/3 of maximum frequency. (E cores stay locked at their max freq, 2.064 GHz, the whole time.)
So: In the Air, M1 CPUs throttle back to about 2/3 of peak. That's perfectly acceptable, if you ask me, because it's a number Intel expects you to deal with even on 125W fan-cooled desktop CPUs:
Intel® Core™ i9-11900KF Processor (16M Cache, up to 5.30 GHz) Product Specifications
Intel® Core™ i9-11900KF Processor (16M Cache, up to 5.30 GHz) quick reference guide including specifications, features, pricing, compatibility, design documentation, ordering codes, spec codes and more.ark.intel.com
Max single core turbo is 5.30 GHz. Sustained all-core speed while running at 125W TDP (the "base" frequency) is 3.50 GHz. That, too, is about 2/3.
If you run this exercise with mobile Tiger Lake CPUs, you will find that 2/3 is a bit optimistic for a lot of them. Even the ones expected to be used with fans.
And given what you say about the Air and 2/3 of maximum frequency, I think you agree with me more than not. All my other laptop experience is on laptops that have active cooling, and I've never seen throttling that severe (a full 3rd) on any other laptop. I never owned an i9 laptop btw, not many people have, they sell far more lesser CPU's and I buy thin and light, which you can't even get an i9 in.
I do have an i9 desktop and it doesn't throttle. (it's got cooling that can handle it.)
Prior to Nuvia's acquisition by Qualcomm, they had some interesting blog posts up comparing performance and perf/W between a bunch of different CPUs. They used GB5 CPU for these tests. When questioned on it, they put up a really interesting additional blog post which laid out a bunch of evidence that GB5 CPU scores are a very accurate predictor of SPEC CPU scores.
They also answered the obvious question: why not just run SPEC? Because SPEC is a lot harder to just download and run than GB5, particularly on mobile platforms like iOS and Android. Nuvia found it much easier and faster to first establish correlation between SPEC and GB5, then run GB5 on all the things they were interested in comparing.
Like it or not, SPEC (and therefore GB5, if you trust Nuvia's data) is a highly relevant benchmark in that CPU architects use it an awful lot. Say you have an idea for a new branch predictor, and you want to figure out if you should actually use it. Architects evaluating questions like this will run trade studies which predict the changes in area, power, frequency, and performance. Performance predictions are often done by running code traces from SPEC (and other benchmarks, but SPEC is common) in a cycle accurate simulator modified to implement the new idea.
Perhaps, in your desire to see yourself as a wise man above the fray, you have led yourself into distrusting things you should trust.
Or perhaps you're a certain other kind of poster I remember from my own days reading CSMA.
And anandtech’s M1 Max benchmarking shows the M1 Max smoking the 5980HS and i9-11980HK in SPEC2017.
”On the SPECint work suite, the M1 Max lies +37% ahead of the best competition, it’s a very clear win here and given the power levels and TDPs, the performance per watt advantages is clear. The M1 Max is also able to outperform desktop chips such as the 11900K, or AMD’s 5800X.
In the SPECfp suite, the M1 Max is in its own category of silicon with no comparison in the market. It completely demolishes any laptop contender, showcasing 2.2x performance of the second-best laptop chip. The M1 Max even manages to outperform the 16-core 5950X – a chip whose package power is at 142W, with rest of system even quite above that. It’s an absolutely absurd comparison and a situation we haven’t seen the likes of.”
Last edited:
Then why do so many of your posts read like a troll trying to pass as the calm, reasonable guy (so as to avoid moderation) while missing few opportunities to drop lies they know will get people riled up?
It's simply absurd to claim that the M1 Air throttles worse than anything out there. Objectively untrue, and very easy to prove if you're technical enough. Most people aren't, but I am, so I'm gonna push back.
I just showed you that a modern i9 desktop does, in fact, "throttle", in the sense that peak performance is not available under all-core loads. Unless you choose to overclock the hell out of it, sustained per-core performance under all-core loads will be 2/3 of peak single-core performance. That's exactly the same as M1 Air.
But hey you want to talk mobile parts? Sure. It just tickled me that even a 125W desktop Intel chip under the official factory PL1/PL2 limits should reduce clocks under all-core load every bit as much as a fanless M1.
First, let's consider what TDP to look at. In my testing, once clocks stabilized, M1 package power stabilized at about 12W. So let's look at a Tiger Lake i3 which OEMs can configure to 12W TDP. (Intel defines TDP as the long-term average power under a "complex" load. Much like M1 in the Air, one should expect this CPU to start out well above 12W for short periods of time, then settle down to 12W in the long term.)
Intel® Core™ i3-1115G4 Processor (6M Cache, up to 4.10 GHz) Product Specifications
Intel® Core™ i3-1115G4 Processor (6M Cache, up to 4.10 GHz) quick reference guide including specifications, features, pricing, compatibility, design documentation, ordering codes, spec codes and more.
2 cores, 4 threads, 4.10 GHz max single core turbo, 1.70 GHz all core speed at 12W TDP.
Compare to M1:
8 cores (4 performance), 8 threads (4 perf), 3.204 GHz max multi-core turbo, 2.1 GHz all core speed at 12W TDP
I dunno about you, but 65.5% is much bigger percent than 41%. And Apple's doing it with twice as many performance cores, with four more efficiency cores thrown in for good measure. Whether you're evaluating by percentage of peak or absolute performance, Apple comes out way ahead.
Personal attack again.
I did not say that. I said that I never had a laptop that throttled as much as my M1 MBA.
Only your interpretation of what I said.
I know they *can* throttle, why do you think I never bought an i9 laptop? And as for peak performance, if you can't perceive a difference in normal day to day usage, it's definitely not like my M1 MBA! I'm not talking benchmarking at all, I'm talking my real world usage.
Maybe you have a defective unit then, as that doesn’t track with almost every other user I’ve seen.