Intel 12900HK is faster than M1 Max for both single and multi cores.

[sentential]

It’s good to see better from Intel, but for laptops they need to at least halve their TDP to compete on a like for like basis imho.

For desktops Apple needs to up their game, which will happen in 2022. If the new full size iMac only has a 10 core CPU (M1 Pro or Max) I will actually be disappointed. In a desktop I don’t care about TDP and want better multi-core CPU performance on Mac without spending $10,000 for a Mac Pro.

Apple have designed a family of chips for Mac with a means of leveraging so that their tech can be used across a variety of devices and form factors. This will mean some side-effects such as excessive memory bandwidth that may not make as much difference in the real world on lower end devices but can still be “sold” as a feature. It will also mean more profits for Apple of course and bring benefits such as the unexpected doubling of the video cores - which again will probably end up being a further quadrupling in a Mac Pro leading to excess and unused cores. So there’s pros and cons to Apple’s approach.

It does feel to me that the new MBPs too heavily focus on video performance - most people are not video editors - but I await the real world tests to enlighten me.

Here’s the deal, what people aren’t talking about is that the Gracemont E-cores from Intel are equal to that of Kabylake at 70% less power.

So the first batch of Alderlake is 8P/8E. So we’re taking an 11th gen Tigerlake-H and adding two 7700Ks on top of it. We’re talking about the same processor that was used in the MBP16 of not even 2 years ago but literally the same one used in the current generation +2 more cores.

Meteorlake/Raptorlake adds 8 more E cores with Meteor adding a Xe Apu bringing the count to 24 total cores (8P/16E) that is going to be really tough to beat and I don’t think Apple can without pushing their core count to 32+.

For time reference Raptor will launch beside the M2 Max next year and if the rumors about the M2 being just a clock bump doesn’t look good as Gracemont matches the M1 in power efficiency.

 

[PlainBelliedSneetch]

I couldn't find any link on the OP-article about the source of this geekbench score for 12900HK. Is it confirmed in any way?

I think Pat Gelsinger vouched for it.

 

[huge_apple_fangirl]

It probably wouldn't even exist, businesses like backwards compatibility. I bet there would be a heck of a lot less computers too. Windows made the personal computer market.

One thing is for sure, Apple wouldn't have any more business market share than they have now, since Apple doesn't do backwards compatibility.

On the contrary, Apple would likely have greater share because Apple has advantages despite (or maybe because of) their attitude towards backwards compatibility. Meanwhile Windows without compatibility is worthless.

 

[magbarn]

I am sure many people still would, I definitely would so there goes your "nobody". I don't use programs that were made in 2001, but I know Windows can still run those programs. Part of why Windows is so horrible is because it is so bloated and brings all those legacy items.

One of the big advantages of windows is that my brother’s core2quad q6600 from 2007 with 4gb of ram is still able to run windows 10 and can still surf modern websites while a Mac mini from that vintage is running an old OS X version with no new versions of safari or chrome available.

 

[machinesworking]

I love these threads.

Apple destroy completion : “omg! Look at these leaks. So much power! So good!”

Intel perform better than apple : “got to wait for real world tests”

In your wording you hint at why it's not as hypocritical at you make it out to be. The Geekbench score shows Intel performing only marginally better than the M1 Max, with a chip that isn't out yet, that mostly likely will consume almost twice the power. For the kind of thermal difference we're going to see here I would expect better than maybe at most a 10% increase in performance.

 

[UBS28]

Is anyone really worried about a desktop chip beating a laptop chip? Can you imagine the battery required to run a power hungry computer and display for 14 hours.

It is not a desktop cpu, it is a laptop cpu with 6 high performance cores.

 

[gatortpk]

The M2 should be out early 2022 and should come with improved CPU and GPU performance over the M1, followed by the M2 Pro and M2 Max.

So this Intel CPU should be going head to head with the M2 and it's variant instead of the M1.

Exactly. The M2 should be based on the A16. The A16/M2 performance cores will likely be 20-30% faster than the A14/M1 cores. Even at just 20% faster at the same power still gets a single core Geekbench of 2100. That likely a good deal faster than any other CPU in 2022.

 

[Spindel]

LOL this thread is funny.

In ST this BM indicates that this CPU beats the computer, I’ve had for 11 months by now, with 3,6 %.

Intel is not beating any one, they just managed to catch up.

The even funnier part is that my computer (M1 mini) has a cheap ass solid aluminium block with fins as a cooler and can sustain that performance indefinatley without the fan going BRRR. I bet my ass the cooler needed for the intel is a fair amount more expensive.

DISCLAIMER of course the intel will beat my machine in MT it has more cores.

 

[UBS28]

Exactly. The M2 should be based on the A16. The A16/M2 performance cores will likely be 20-30% faster than the A14/M1 cores. Even at just 20% faster at the same power still gets a single core Geekbench of 2100. That likely a good deal faster than any other CPU in 2022.

Intel can easily make a faster version because Apple uses 8 high performance cores while Intel only used 6 in this version.

 

[UBS28]

In your wording you hint at why it's not as hypocritical at you make it out to be. The Geekbench score shows Intel performing only marginally better than the M1 Max, with a chip that isn't out yet, that mostly likely will consume almost twice the power. For the kind of thermal difference we're going to see here I would expect better than maybe at most a 10% increase in performance.

This will be addressed when Intel switches to 3nm using TSMC in the near future.

Intel is still on 10nm at the moment.

 

[machinesworking]

This will be addressed when Intel switches to 3nm using TSMC in the near future.

Intel is still on 10nm at the moment.

How do you see that playing out? like everyone is mentioning Apple aren't standing still here. I think we will see the same outcome over and over for a while, Apple release a low power CPU with the best performance, then AMD and/or Intel catch up but require much more power. Hop skip but the same energy in/out deficiency from x86. At some point Microsoft is entering the build your own Arm chip market, they announced it, and it's not because they blindly copy Apple, it's because they copy them when they're on to something. I don't think this will change magically when the whole industry moves to 3nm, since Apple will also be there with the whole SOC thing and Arm difference to propel better per watt performance.

 

[UBS28]

FYI, in specialized tasks, the M1 Max Pro will be faster than this Intel CPU, as the M1 Max Pro has specialized chips for DSP, video encoding / decoding, and so on ….

But Intel is probably aware Apple has these specialized chips so they might “copy” this in future generation of Intel chips perhaps.

 

[UBS28]

How do you see that playing out? like everyone is mentioning Apple aren't standing still here. I think we will see the same outcome over and over for a while, Apple release a low power CPU with the best performance, then AMD and/or Intel catch up but require much more power. Hop skip but the same energy in/out deficiency from x86. At some point Microsoft is entering the build your own Arm chip market, they announced it, and it's not because they blindly copy Apple, it's because they copy them when they're on to something. I don't think this will change magically when the whole industry moves to 3nm, since Apple will also be there with the whole SOC thing and Arm difference to propel better per watt performance.

What makes ARM attractive, is that companies can custom design their own chips to their own needs at a much lower price.

So a company that provides cloud services for Machine Learning to it’s customers, will design ARM server chips that focusses on specialized compute units aimed at machine learning tasks.

For anything else, I don’t see it especially with software compatibility problems related to ARM.

 

[Krevnik]

FYI, in specialized tasks, the M1 Max Pro will be faster than this Intel CPU, as the M1 Max Pro has specialized chips for DSP, video encoding / decoding, and so on ….

But Intel is probably aware Apple has these specialized chips so they might “copy” this in future generation of Intel chips perhaps.

What do you mean copy? Intel has been shipping dedicated encoder/decoder blocks for video since 2011: https://en.wikipedia.org/wiki/Intel_Quick_Sync_Video

Features like using AirPlay to mirror your Mac screen have been dependent on these hardware blocks for performance.

 

[crazy dave]

This will be addressed when Intel switches to 3nm using TSMC in the near future.

Intel is still on 10nm at the moment.

Just to make sure we’re all on the same page: Intel’s “10nm” is more like TSMC’s “7nm” maybe even more like “7nm+”. So they’re currently only a node behind. This is both good and bad news for Intel given where their chips fall relative to AMD and especially Apple. Good: Intel foundries don’t have as far to go to catch up. If they finally keep their schedule Intel could catch up to (gain parity with) TSMC within the next 5 years roughly - faster if TSMC slip up. The Bad: Intel are maybe slightly further behind in CPU core design than some fan boys on any side of the argument realize. Having said that, *some* of those design issues may be due to the delays in manufacturing forcing design delays. Unclogging that may help. Still, they have a looooong way to go.

FYI, in specialized tasks, the M1 Max Pro will be faster than this Intel CPU, as the M1 Max Pro has specialized chips for DSP, video encoding / decoding, and so on ….

But Intel is probably aware Apple has these specialized chips so they might “copy” this in future generation of Intel chips perhaps.

Yes Intel is planning on greatly expanding it’s custom chip making for customers. However Intel already puts dedicated hardware on their CPUs. That alone won’t be new. AMD already does a lot of custom chip designs - obviously the biggest customers for them utilizing those services being the console makers.

 

[machinesworking]

What makes ARM attractive, is that companies can custom design their own chips to their own needs at a much lower price.

So a company that provides cloud services for Machine Learning to it’s customers, will design ARM server chips that focusses on specialized compute units aimed at machine learning tasks.

For anything else, I don’t see it especially with software compatibility problems related to ARM.

Hmm. you're adding in new negative elements there, taking us down another path. Apple are using Arm for general purpose computing chips, and have been for years and years. So pretty much everything you wrote here is wrong. Unless you think iPhones, iPads, and the new M1 chips are all terrible, which is ridiculous.

 

[KPOM]

Here’s the deal, what people aren’t talking about is that the Gracemont E-cores from Intel are equal to that of Kabylake at 70% less power.

So the first batch of Alderlake is 8P/8E. So we’re taking an 11th gen Tigerlake-H and adding two 7700Ks on top of it. We’re talking about the same processor that was used in the MBP16 of not even 2 years ago but literally the same one used in the current generation +2 more cores.

Meteorlake/Raptorlake adds 8 more E cores with Meteor adding a Xe Apu bringing the count to 24 total cores (8P/16E) that is going to be really tough to beat and I don’t think Apple can without pushing their core count to 32+.

For time reference Raptor will launch beside the M2 Max next year and if the rumors about the M2 being just a clock bump doesn’t look good as Gracemont matches the M1 in power efficiency.

I’ll believe it when I see it.

 

[ratspg]

Intel lost. They sat on their laurels for so many years and didn’t take advantage to continue innovating. Apple caught up and took it to the next level. Anyways who cares about Intel’s 12th gen chips. They won’t be running MacOS. ?? Intel can try and win back Apple in forever from now when they catch up. It is incredible as a professional to have a system top to bottom hardware to software all developed in house for optimal performance. If people want openness go use Linux, Windows even. It’s there for that. Apple Silicon is here to stay and I welcome this new era.

 

[leman]

What kind of stats do you do, @leman? I'm a cognitive science grad student, so I've been getting into a lot of Bayesian mixed-effects models using RStan and brms. A friend of mine who upgraded from a ~2015 MBP to an M1 Air said his Stan models that used to take 3+ hours now only take 20-30 minutes, so I'm extremely curious to see if the improved memory bandwidth with the M1 Pro/Max will push that even further. Do you have an M1 machine to compare to your 16" yet, or have you been holding out for the M1 Pro?

I am usually working with tabular data transformations in R, but I occasionally run stan programs (phylogenetic fitting etc.) developed by my colleagues. I briefly had an M1 MacBook Pro to test out some software before committing our research group to buying more, and that thing was impressive for my workflows. In (multicore) stan and R it easily matched the performance of my i9, and sometimes even outperformed it. I have little doubt that the M1 Pro/Max will do very well for data science and software development, they have massive cache and everyday FP advantages that is not reflected by microbenchmark like Geekbench.

Wow, the Mac now offers type 1 hyervisor support for native GPU passthrough and bare metal performance?! I was only aware of Parallels and VMware which are type 2 hypervisors which are explicitly NOT native, do NOT offer GPU passthrough, and take a rather substantial performance hit! That's really cool that I'm wrong, I wasn't aware of this new support! How are they doing the passthrough with only a single iGPU?

Sorry, what do you want with the bare metal GPU virtualization if your client OS has no drivers for that GPU? I definitely agree that one thing lacking from macOS is support for PCIe passthrough to the VM, that would give a viable option to folks that rely on CUDA.

Windows is kosher to run too, right? Microsoft OK'd it?

No, they didn't. Like they never OKed to run it on Bootcamp.

Maybe so, but that makes it even worse for someone pining for ultra ultra thin pro Macs.

But seriously, if you are taking issue with this (an i9 11900H computer that sustains load):

My problem with ROG Zephyrus M16 is not nessesarily the size (although I doubt it will fit into my 13" backpack — a 16" MBP does fit). It's the poor thermals (20C hotter chassis under load compared to even the Intel 16"), loud fans, poor battery (especially if you are doing any work), poor performance when unplugged, poor ergonomic in cramped spaces (all laptops that have bottom airflow cutout cannot be properly used on the lap/bed etc., which is one of the reasons Apple has always used side vents for airflow) and poor connectivity (one USB 4 port? this is a joke for a work laptop).

These are the flaws shared by virtually all gaming and workstation laptops, which is why I use Apple. They understand that computer design should be functional. Especially in regards to cooling solutions. I have no interest in using a laptop with direct bottom heat exhaust that reaches over 50C under operation. That is enough to burn my wooden desk (not to mention my knees).

Nothing wrong with being a skeptic here and being upset at Apple's business led decisions that negatively affect the direction of our software lives. People will be mad at you because they need to justify their purchase or they feel the need to defend their company out of some narrowminded view of tech.

I am not mad because I need to justify my purchase (there is nothing to justify, all my software and workflows are already ARM-native). I am mad because outlandish arguments (e.g. 32bit software compatibility) are being used to ridicule my purchase choice or technology which is probably the most significant development in the CPU history since the Intel P6 core.

For all the good M1 brings, it's also another step towards locking down the Apple ecosystem even further and making it harder and harder for outsiders to compete or for customers to make choices about how they want their digital life to be structured.

I do not understand this comment. Appel Silicon is a closed hardware platform, that is sure, but it's a fully open software platform.

Here’s the deal, what people aren’t talking about is that the Gracemont E-cores from Intel are equal to that of Kabylake at 70% less power.

So the first batch of Alderlake is 8P/8E. So we’re taking an 11th gen Tigerlake-H and adding two 7700Ks on top of it. We’re talking about the same processor that was used in the MBP16 of not even 2 years ago but literally the same one used in the current generation +2 more cores.

Meteorlake/Raptorlake adds 8 more E cores with Meteor adding a Xe Apu bringing the count to 24 total cores (8P/16E) that is going to be really tough to beat and I don’t think Apple can without pushing their core count to 32+.

For time reference Raptor will launch beside the M2 Max next year and if the rumors about the M2 being just a clock bump doesn’t look good as Gracemont matches the M1 in power efficiency.

Intels strategy with Gracemont cores is not a step forward, it's more of a step to the side. Intel is unable to make fast energy-efficient cores, so they decided to add some slow energy-efficient cores into the mix for throughput-oriented workloads. Now, this is an effective strategy, but it is also limited. Not all workloads scale well with CPU count, and Intel lacks bandwidth on it's consumer platform to properly utilize 30+ cores. Just adding more and more Gracemont cores will reflect well in microbenchmark, but one will quickly run into diminishing returns in the real world.

You say that Gracemont has matched M1 in power efficiency. That is not entirely so. Gracemont matches M1 in power consumption, with 1/3 lower performance. From the benchmarks I have seen Gracemont is somewhere around A15's Blizzard, maybe a bit faster. Apple Silicon cores are sill much more efficient in terms of performance/watt as well as sustained performance per core. If leaked Geekbench benchmarks are to be trusted, 8+2 M1 is within 5% of the of the line mobile 6+8 Alder Lake (and M1 is likely to be faster in more complex workloads). So something like 12+4 an M3 should be more than sufficient to take on a 8+16 Meter Lake.

 

[Andropov]

Intels strategy with Gracemont cores is not a step forward, it's more of a step to the side. Intel is unable to make fast energy-efficient cores, so they decided to add some slow energy-efficient cores into the mix for throughput-oriented workloads. Now, this is an effective strategy, but it is also limited. Not all workloads scale well with CPU count, and Intel lacks bandwidth on it's consumer platform to properly utilize 30+ cores. Just adding more and more Gracemont cores will reflect well in microbenchmark, but one will quickly run into diminishing returns in the real world.

I'd say it's also limited in the sense that it's an improvement that can only be done once. Even if this year they got a huge boost by adding the E-cores, next year all the performance/efficiency gains will have to come from improving the μarch of the existing cores. There's also the option of adding more efficiency cores into the mix, or a new tier of even slower cores, but that's not scalable for the reasons you stated. So we'll see how they do next year.

 

[Serban55]

"I couldn't find any link on the OP-article about the source of this geekbench score for 12900HK. Is it confirmed in any way?"

No, nothing is confirmed since we need to see also in what device is in it..i mean this could be the top maximum it can go inside an 17" MSI massive laptop
SO we need to see that cpu how performs in different enclosures

 

[Serban55]

My problem with ROG Zephyrus M16 is not nessesarily the size (although I doubt it will fit into my 13" backpack — a 16" MBP does fit). It's the poor thermals (20C hotter chassis under load compared to even the Intel 16"), loud fans, poor battery

wait, so let me get this clear...that laptop is big but still has a lot of chassis heat with poor thermals and poor battery life??
I dont understand...why they made it this big then?
I mean MSI GE66 its big, but it has good thermals and its not so hot to the touch

 

[leman]

wait, so let me get this clear...that laptop is big but still has a lot of chassis heat with poor thermals and poor battery life??

It's not really big, it's only 20% larger than the new 16" MBP. In the PC gaming world that's still considered petite.

I dont understand...why they made it this big then?
I mean MSI GE66 its big, but it has good thermals and its not so hot to the touch

The MSI GE66 will still heat up to 50C under load... this is not the question of size, really, it's the question of how the cooling system is designed how much heat the system outputs. These laptops have to dissipate roughly 3 times more heat than the M1 Max, and the scalding hot air needs to go somewhere. This is why they rely on noisy fans and large air vents. Apple has always designed their laptops that the hot air does not blow on your legs (or burns your desk) — it is ejected behind the laptop. Appel has also always restricted the maximal thermals of the laptop to make them more manageable on the go (something that has been frequently criticized). With Apple Silicon, MacBook Pros can finally match (or even surpass) the performance of these hot, loud gaming laptops while still retaining the more ergonomic cooling system as well as reasonable total power dissipation of ~80watts under full load.

 

[vir]

I'm often wrong.

 

[leman]

I also said I had the 15" MBP alongside my M16 in person, right here, right now, and the difference in thickness is exactly the same as the difference in thickness the new 2021 MBP has with the 15". I am not wrong about that. You can stare at numbers on websites all you want, but that is not a substitute for real-world experience. Who knows how those numbers are being calculated (probably to include the rubber feet).

Actually you are, by roughly 2mm (which to you must be a big deal seeing how you see 1mm as "so much bigger") And of course one counts the feet. The feet don't magically disappear when you have to stove your laptop away.