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

[alexqndr]

Speed estimated outside a thermal envelope doesn’t mean absolutely anything.

Furthermore M1 Max is Mobile chip. Comparing it against a workstation chip it’s nonsense.

M1 Max is the fastest mobile chip on the market, period.

 

[edfoo]

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”

LOL what do you expect? This is an Apple fanboi forum.

 

[hagjohn]

That's me. TDP means very little, raw benchmarks mean more, and what software it runs and if it's fast enough for the job is what I look for.

I really, really, don't understand why TDP is even talked about that much around here. It's a total yawn for me.

TDP is important because a CPU (or other hardware) will throttle if there is too much heat, so u will never see the maximum output of the CPU or other components. It is brought up a lot because it matters a lot.

 

[falainber]

Speed estimated outside a thermal envelope doesn’t mean absolutely anything.

Furthermore M1 Max is Mobile chip. Comparing it against a workstation chip it’s nonsense.

M1 Max is the fastest mobile chip on the market, period.

Intel H series chips are mobile chips (their highest performance mobile chips)

 

[falainber]

TDP is important because a CPU (or other hardware) will throttle if there is too much heat, so u will never see the maximum output of the CPU or other components. It is brought up a lot because it matters a lot.

It usually means a lot more for Apple hardware - especially in the past - because historically they were trying to make the laptops thinner and quieter. This was done at the expense of cooling. If cooling is adequate then there won't be any throttling. That's not to minimize the technical challenges involved and obviously higher performance at the same power consumption level is a good thing.

 

[anshuvorty]

Well, Apple isn't going to change course anytime soon, so how does this matter to Mac users? If you are a Mac user, you are stuck with Apple Silicon for good or bad....

In other words, if Intel miraculously comes out with a chip that beats Apple Silicon in performance/watt, what are you going to do about it? Jump ship and start using WinTel?

 

[Rendom]

Does anyone else find it suspicious that there is not link to the actual Geekbenck entry, only a screen shot? Tried to find it myself without any luck. Not to mention the "max frequency" of 371Mhz... Might just be me being a noob though! ?

 

[falainber]

A fast CPU also needs to be paired with a capable sub-system. I'm still trying to understand Apple's decision to put 200 - 400 GB/s of bandwidth into a notebook form factor. This is essentially the bandwidth you get at the workstation and server category. Not even the Intel Mac Pro has such high memory bandwidth.

I'm expecting real world tasks to be jaw droopingly fast for these M1 Pro and M1 Max MacBooks.

M1 needs this bandwidth because it has integrated graphics. Graphics needs it. The bandwidth in Mac Pro is for CPU. It's much less critical for CPU. The laptops where H series CPUs are used are gaming (and other high performance) laptops. They use discrete graphics which has its own memory and bandwidth.

 

[falainber]

Intel is possibly going to announce the chip this month, but I wouldn’t be surprised if it wasn’t available in a real product before 2022. Intel has a strong track record of underdelivering the last few years. https://www.anandtech.com/show/12271/intel-mentions-10nm-briefly

Alder Lake chips go on sale November 4th. The first SKUs are expected to be for desktops through.

 

[quarkysg]

M1 needs this bandwidth because it has integrated graphics. Graphics needs it. The bandwidth in Mac Pro is for CPU. It's much less critical for CPU. The laptops where H series CPUs are used are gaming (and other high performance) laptops. They use discrete graphics which has its own memory and bandwidth.

Don’t forget that traditional dGPUs need to get fed data via the PCIe bus, and it maxes out at 32GB/s.

Memory is also important for CPUs, not solely for GPUs. This especially true for high core count CPUs. Otherwise those fast cores will be spending a lot of their time waiting for data to process. There’s a reason why servers are faster at what they do while spotting lower max frequencies, as servers usually have large data pipes feeding the CPU cores.

I have a sneaky suspicion that the M1 Pro and Max Cinebench scores will be impressive relative to the core count increases, and that’s probably due to the bandwidth increase.

 

[profcutter]

Intels new chips certainly are not vapor:

A Lucky Customer Just Got Intel's New 12th-Gen Chips Way Ahead of Their Official Release

A Reddit user says they managed to purchase not one but two Intel Core i9-12900K processors for just $610 each, weeks before they officially go on sale.

gizmodo.com

 

[Taz Mangus]

A fast CPU also needs to be paired with a capable sub-system. I'm still trying to understand Apple's decision to put 200 - 400 GB/s of bandwidth into a notebook form factor. This is essentially the bandwidth you get at the workstation and server category. Not even the Intel Mac Pro has such high memory bandwidth.

I'm expecting real world tasks to be jaw droopingly fast for these M1 Pro and M1 Max MacBooks.

And Apple is using an insane 16-Channel memory architecture. This stuff is mind blowing in a laptop.

 

[whcchief]

The latest release from Intel/AMD/Apple should be compared to the latest release from their competitors. I don't think there is really another fair way to compare them. People can't buy "tomorrow's processor", they can only buy what's available today.

Yeah but I might buy tomorrows processor...tomorrow!

 

[anshuvorty]

And Apple is using an insane 16-Channel memory architecture. This stuff is mind blowing in a laptop.

My insane bet? It's to sustain the development of AR apps when the much rumored AR headset is released.

In order to build apps for an AR headset, you will be need an insane amount of compute and graphical power. This is just the beginning of that roadmap. It seems like plenty for today's workload, but maybe in 5 years time or whenever the headset is released, we will thank Apple for building computers that are able to power the development of high quality apps for the headset. Who knows!

 

[Analog Kid]

Well, Apple isn't going to change course anytime soon, so how does this matter to Mac users? If you are a Mac user, you are stuck with Apple Silicon for good or bad....

In other words, if Intel miraculously comes out with a chip that beats Apple Silicon in performance/watt, what are you going to do about it? Jump ship and start using WinTel?

Things were a lot more interesting back in the PowerPC days when people could analyze the differences between it and the x86 architectures and there was a perpetual horse race. When Apple went Intel, it was much closer to what you're describing-- we get what Intel makes, just like every other platform in the world. Ho hum. I wonder if the new Macbook will have one or two ports on the left this time.

Now we've got multiple design philosophies again, and it's going to make each hardware release much more exciting.

 

[bobcomer]

TDP tells you its sustained load as the power states programmed by Intel is limited to its TDP at full sustained loads

That's still pretty irrelevant to me.

 

[yitwail]

A fast CPU also needs to be paired with a capable sub-system. I'm still trying to understand Apple's decision to put 200 - 400 GB/s of bandwidth into a notebook form factor. This is essentially the bandwidth you get at the workstation and server category. Not even the Intel Mac Pro has such high memory bandwidth.

I'm expecting real world tasks to be jaw droopingly fast for these M1 Pro and M1 Max MacBooks.

I’m no engineer so my uneducated guess is that the bandwidth was implemented to improve GPU performance. I should go read what anandtech had to say about it, if anything.

ETA: here’s a statement from anandtech article on M1 Pro/Max:
“This kind of bandwidth is unheard of in an SoC, but quite the norm in very high-end GPUs.”
Source: https://www.anandtech.com/show/1701...m1-max-giant-new-socs-with-allout-performance

 

[leman]

If these numbers are real, I am a bit relieved. I was worried that Alder Lake would be faster. But this again confirms that Apple’s move to drop Intel was the correct one. Few considerations:

- that particular benchmark result is likely overclocked
- that particular i9 will be only released in very small numbers while the M1 10-CPU cluster will ship with practically every MBP
- Adler Lake will consume at least 2x the power of M1 10-CPU cluster (this particular i9 is probably at least 60-80W)
- the performance will be severely diminished on battery, while M1 will still offer full speed
- M1 has a decisive cache and memory bandwidth advantage that will make it faster in many demanding real-world tasks


At any rate, if this is the best Intel can do with their 6+8 config against Apple’s 8+2 config, it doesn’t look too rosy for Intel. But they should be able to take the sustained crown from AMD

 

[bobcomer]

Package power doesn't stay that much different from TDP. Unless you put that 12900HK processor on an unlocked BIOS with can force that P0 state indefinitely with overclock to boot under full load (maybe in laptops that are like 1.5 to 2" thick like those Eurocom laptops), but for apples to apples comparison, a similar multimedia laptop with that chip and its package power under sustained load will definitely throttle to either base speed or 0.1-0.3 GHz above it (pegging its package power to ~45-50W) and now it's a lot slower than M1 Max at full sustained load (with the assumption that M1 Max can run at full speed indefinitely). After looping Cinebench for 100 times, you can see how much performance degradation that Intel chip has (maybe 15-20% down) compared to M1 Max which would still be the same as the first test (no thermal throttling)

I haven't met a laptop that never thermal throttled. It's whether it stays acceptable in the speed-wise range, and I would expect that of both Alder Lake and the M1 Pro/Max. (unless it's passively cooled like my M1 MBA, worse throttling laptop I've ever owned.)

 

[Bug-Creator]

Few considerations:

- Alder Lake is a CPU with a display-out GPU, M1Max is mostly GPUs so running a CPU benchmark isn't really telling us anything

 

[leman]

- Alder Lake is a CPU with a display-out GPU, M1Max is mostly GPUs so running a CPU benchmark isn't really telling us anything

That is a very weird thing to say. I can’t use a GPU to build software. Of course the CPU performance of Apple Silicon matters and of course running a CPU benchmark tells us a lot.

 

[bobcomer]

TDP is important because a CPU (or other hardware) will throttle if there is too much heat, so u will never see the maximum output of the CPU or other components. It is brought up a lot because it matters a lot.

Not to me it doesn't. Whether it runs the software I need and runs it acceptably is WAY more important.

 

[Gnattu]

Not to me it doesn't. Whether it runs the software I need and runs it acceptably is WAY more important.

Laptops are typically both thermal and power constrained, so you many need a powerful desktop because laptops have to compromise in some way(and be more expensive).

 

[vir]

You lost so much.

 

[bobcomer]

Laptops are typically both thermal and power constrained, so you many need a powerful desktop because laptops have to compromise in some way(and be more expensive).

I'm typing on an i9 non mobile processor right now and throttling is not an issue, no matter what I do, but I also have a couple laptops not counting my M1 MBA, and throttling really hasn't been a problem for me, they stay in the acceptable range speed-wise. (Though curiously, the thinner and lighter of the 2, an X1 Carbon, seems to throttle less than my Dell XPS15, Both i7's a generation apart.)