Question Alder Lake - Official Thread

[Hulk]

With the release of Alder Lake less than a week away and the "Lakes" thread having turned into a nightmare to navigate I thought it might be a good time to start a discussion thread solely for Alder Lake.

 

[IntelUser2000]

AT has a review, and so does NBC on the MSI Raider. Battery life is pretty bad, perhaps unsurprisingly with 110 W PL1 12900HK and a 175 W 3080 Ti.

https://www.notebookcheck.net/MSI-GE76-Raider-Laptop-Review-With-Alder-Lake-and-RTX-3080-Ti-to-gaming-heaven.595391.0.html

Regression in NBC's review isn't too bad. It goes from 328 minutes in the 11th gen GE76 to 306 minutes. Although both are a regression from Cometlake at 375 minutes.

We will see how the U and P does but it does show why Intel didn't say anything about battery life. Like I said Tigerlake doesn't improve things over Ice or Cometlake.

-H could have been better with the on package PCH but it didn't materialize. We will see if that's unique with this laptop.

 

[Mopetar]

7-zip decompression test is a weird anomaly. Why is 5900HX so great at that? And so bad at compression? Perplexing.

7-zip is mainly an integer-based workload and Zen has usually done well in that area. I think a big part of that just comes down to the wider design and being able to execute more of those per clock cycle.

The large L3 cache probably helps out a lot as well. If I'm recalling correctly, the original Zen and Zen+ CPUs weren't as good and that may have been due to the smaller and split cache layout.

 

[ondma]

Understatement. Intel should be ashamed of itself for producing such a laptop cpu.

And is anyone really going to use a high end gaming laptop for any extended time on battery? I dont really think Intel should be "ashamed" of building such a cpu. I one doesnt want such a power hungry cpu, they can certainly buy a different model. I guess in you mind choice is not a good thing?

 

[igor_kavinski]

If 6900HX improves on performance without sacrificing power efficiency, Intel's "victories" here would be short-lived.

 

[uzzi38]

And is anyone really going to use a high end gaming laptop for any extended time on battery? I dont really think Intel should be "ashamed" of building such a cpu. I one doesnt want such a power hungry cpu, they can certainly buy a different model. I guess in you mind choice is not a good thing?

This same chip will also be used in thin and light laptops too, in case you forgot. This is sending out bad signals for what's to come for that segment, especially considering that ADL-P will be compared to TGL-U and TGL-H35, both of which already had the chipset on package.

 

[uzzi38]

If 6900HX improves on performance without sacrificing power efficiency, Intel's "victories" here would be short-lived.

Sacrificing? It's the same CPU core and refined everything else. Of course power efficiency isn't being sacrificed lmao. The question has always been: how much does it improve, not does it improve

 

[eek2121]

And is anyone really going to use a high end gaming laptop for any extended time on battery? I dont really think Intel should be "ashamed" of building such a cpu. I one doesnt want such a power hungry cpu, they can certainly buy a different model. I guess in you mind choice is not a good thing?

Not that one, no. I have a gaming laptop I use to play games plugged in, and use for other work when unplugged. It has good battery life.

If 6900HX improves on performance without sacrificing power efficiency, Intel's "victories" here would be short-lived.

The 6900HX will IMPROVE power efficiency, not sacrifice it.

Sacrificing? It's the same CPU core and refined everything else. Of course power efficiency isn't being sacrificed lmao. The question has always been: how much does it improve, not does it improve

It's a refined CPU core. Mostly the same, but changes have been made. It is on a new process, for starters.

 

[dullard]

And is anyone really going to use a high end gaming laptop for any extended time on battery? I dont really think Intel should be "ashamed" of building such a cpu. I one doesnt want such a power hungry cpu, they can certainly buy a different model. I guess in you mind choice is not a good thing?

Quite honestly, I think it is a bad use case. If you want to have the most power, use a desktop. If you want portability, these are barely portable at 6.4 lbs. Yet, for some reason there is a sizable enough market for that.

This same chip will also be used in thin and light laptops too, in case you forgot. This is sending out bad signals for what's to come for that segment, especially considering that ADL-P will be compared to TGL-U and TGL-H35, both of which already had the chipset on package.

Um, the 12900HK certainly is NOT the same chip for thin and light laptops. It isn't even the same die if you talk about the U-series thin and light laptops. You could have said that the same cores will be used.

https://images.anandtech.com/doci/17215/2022_CES_Pre-Brief_Dec%2015%20Final-page-005.jpg

 

[jpiniero]

Um, the 12900HK certainly is NOT the same chip for thin and light laptops. It isn't even the same die. You could have said that the same cores will be used.

https://images.anandtech.com/doci/17215/2022_CES_Pre-Brief_Dec%2015%20Final-page-005.jpg

H and P use the same die.

 

[igor_kavinski]

6900HX's performance could be an improvement over 5900HX, or it could be over 12900HK. The latter might entail sacrificing power efficiency due to required higher clocks. But I wager AMD will stick to keeping its chips power efficient rather than beating its rival at any cost.

 

[IntelUser2000]

The battery life is much better on Linus' test so let's wait a bit more.

But yea Alderlake potentially regresses on the battery life department.

@igor_kavinski

I don't think RMB-H will improve performance much. ADL-H is still going to hold a significant performance lead.

Remember where each company touted their horns on - Intel for -H CPU performance, and AMD for -U CPU performance and iGPU performance and battery life. Everything that both are quiet on they will not do that well.

To reiterate:
-Intel will have the fastest -H CPU by a significant margin
-AMD will have the leadership integrated graphics, better battery life, and significant CPU performance advantage in most of the -U and -P category.

 

[igor_kavinski]

Everything that both are quiet on they will not do that well.

Good point!

 

[IntelUser2000]

7-zip is mainly an integer-based workload and Zen has usually done well in that area. I think a big part of that just comes down to the wider design and being able to execute more of those per clock cycle.

Golden Cove is wider than Zen 3 in every category.

Here's what one AT review said:

Decompression is an even lower IPC workload, as it is very branch intensive and depends on the latencies of the multiply and shift instructions.

 

[uzzi38]

The battery life is much better on Linus' test so let's wait a bit more.

But yea Alderlake potentially regresses on the battery life department.

@igor_kavinski

I don't think RMB-H will improve performance much. ADL-H is still going to hold a significant performance lead.

Remember where each company touted their horns on - Intel for -H CPU performance, and AMD for -U CPU performance and iGPU performance and battery life. Everything that both are quiet on they will not do that well.

To reiterate:
-Intel will have the fastest -H CPU by a significant margin
-AMD will have the leadership integrated graphics, better battery life, and significant CPU performance advantage in most of the -U and -P category.

For the MSI laptop reviews are split on battery life.

AT and Hardware Canucks got serious degradations.

LTT and PC World got serious improvements.

Dave2D got margin of error level differences (at idle, but under load significant improvement)

Notebookcheck got small degradation over TGL-H.

Then there's HotHardware's Alienware testing which showed abysmal battery life, and Nadalina's testing of the Strix Scar G17 which had better battery life than other TGL devices, but lost to the previous year's Scar 15 with 5900HX.

 

[uzzi38]

Um, the 12900HK certainly is NOT the same chip for thin and light laptops. It isn't even the same die if you talk about the U-series thin and light laptops. You could have said that the same cores will be used.

https://images.anandtech.com/doci/17215/2022_CES_Pre-Brief_Dec%2015%20Final-page-005.jpg

P-series is also for thin and light laptops.

 

[IntelUser2000]

Then there's HotHardware's Alienware testing which showed abysmal battery life, and Nadalina's testing of the Strix Scar G17 which had better battery life than other TGL devices, but lost to the previous year's Scar 15 with 5900HX.

The Ryzen H chips were always better than Intel's, because Intel treats the H chips like the desktop platform and the resulting notebooks weren't optimized for power efficiency at all.* So Nadalina's testing isn't something we can't take seriously.

But overall it's showing a degradation.

Do you know for sure if the P is exactly same as the H? Because I know currently the U chips are different from the H chips. The H chips require less voltage for the same clock speed as the U chips - I assume the U chips are made for low leakage because it starts to matter there.

*Actually there's a notebookreview thread that shows how to optimize for the best battery life and some with H chips literally triple their browsing battery life. At least for Skylake-based chips even with the H you can take it under 1W package power when idle, which is a fraction of what it measures when you take the laptop out of the box.

 

[Exist50]

The Ryzen H chips were always better than Intel's, because Intel treats the H chips like the desktop platform and the resulting notebooks weren't optimized for power efficiency at all.* So Nadalina's testing isn't something we can't take seriously.

But overall it's showing a degradation.

Do you know for sure if the P is exactly same as the H? Because I know currently the U chips are different from the H chips. The H chips require less voltage for the same clock speed as the U chips - I assume the U chips are made for low leakage because it starts to matter there.

*Actually there's a notebookreview thread that shows how to optimize for the best battery life and some with H chips literally triple their browsing battery life. At least for Skylake-based chips even with the H you can take it under 1W package power when idle, which is a fraction of what it measures when you take the laptop out of the box.

Even desktop Alder Lake uses the low leakage process, though not for lack of trying. But P and H do share a die anyways.

 

[hemedans]

Sacrificing? It's the same CPU core and refined everything else. Of course power efficiency isn't being sacrificed lmao. The question has always been: how much does it improve, not does it improve

Some reviewers like Linustechtips got better battery life in gen 12 compare to gen 11, we need real life test to know for sure, some synthetic benchmarks may use bigger cores to measure battery life and ignore Efficiency cores.

 

[uzzi38]

Some reviewers like Linustechtips got better battery life in gen 12 compare to gen 11, we need real life test to know for sure, some synthetic benchmarks may use bigger cores to measure battery life and ignore Efficiency cores.

1. The entire point of thread director is to avoid the situation where non-ideal cores are used.

2. Using the efficiency cores is actually detrimental in this case, because the only situation in which they're more power efficient than the P cores is at extremely low power. Ideally you want to finish tasks in as quickly as possible on battery so that you can power off as much of the SoC as possible again. It's called race to idle.

https://twitter.com/i/web/status/1486022083775590402

 

[IntelUser2000]

Ideally you want to finish tasks in as quickly as possible on battery so that you can power off as much of the SoC as possible again. It's called race to idle.

This isn't always the case by the way. It depends on if the CPU or the system power is dominant. In the case where system power is dominant, you'll actually save power if you have a low TDP CPU. In the case where CPU power is dominant, yes the race to idle matters.

The reason the E cores don't improve battery life is because whatever core needs to be used, it powers up the uncore. When the E core is active, the uncore is quite dominant, so any power efficiency each core has is dampened. When the P core is active, the uncore power addition stays similar, so the faster performance makes it look more efficient than it actually is.

For example-
X cores: 20+5 active = Absolute value 25
Y cores: 20+20 = Absolute value 40

Y=2X performance

X core system level relative value: 25
Y core system level relative: 20

So yes the E core is efficient, but efficient in the sense of allowing the whole chip to achieve higher performance with lower power required than if they were all P cores. But in case of battery life, it doesn't do much.

 

[uzzi38]

This isn't always the case by the way. It depends on if the CPU or the system power is dominant. In the case where system power is dominant, you'll actually save power if you have a low TDP CPU. In the case where CPU power is dominant, yes the race to idle matters.

Sure, but for the tasks that are usually used for battery life testing - web surfing and video playback - both of which are very light loads on the SoC in general. The latter especially barely hits the CPU at all and is more reliant on the hardware decode engine present on-chip, whilst the former is a pretty lightly threaded task.

 

[IntelUser2000]

Sure, but for the tasks that are usually used for battery life testing - web surfing and video playback - both of which are very light loads on the SoC in general. The latter especially barely hits the CPU at all and is more reliant on the hardware decode engine present on-chip, whilst the former is a pretty lightly threaded task.

Yea so we can see the addition of the E cores are entirely for performance reasons. "Efficient performance" should be the actual name.

Some reviewers like Linustechtips got better battery life in gen 12 compare to gen 11, we need real life test to know for sure, some synthetic benchmarks may use bigger cores to measure battery life and ignore Efficiency cores.

The E cores won't improve battery life. Because whether you are talking about an E core, or a P core, when it's at idle, it's close to zero. And if either one is active, you need to fire up the uncore, and the PCH. So you might as well go use a P core and ramp down faster.

E core helps in constant load scenarios or when you need absolute performance, since it adds performance in a more efficient manner(both power and die area) in comparison to adding P cores.

So the hybrid implementation won't improve battery life. What will is improved power management, and most importantly for Intel, go Foveros to have an effective on-die PCH, and that won't happen until Meteorlake. Raptorlake might help a bit with tricks, but the real potential biggie is MTL.

 

[Abwx]

12900HK review at Computerbase :

Intel Core i9-12900HK im Test

Mit Alder Lake-H bietet Intel wieder die schnellste CPU für Notebooks: Der Core i9-12900HK schlägt die gesamte Konkurrenz im Test.

www.computerbase.de

 

[Zucker2k]

1. The entire point of thread director is to avoid the situation where non-ideal cores are used.

2. Using the efficiency cores is actually detrimental in this case, because the only situation in which they're more power efficient than the P cores is at extremely low power. Ideally you want to finish tasks in as quickly as possible on battery so that you can power off as much of the SoC as possible again. It's called race to idle.

https://twitter.com/i/web/status/1486022083775590402

Race to idle doesn't trump inherent efficiency, though, does it? If you're browsing the web, for example, would you rather use low power cores or performance cores? I think the criteria for determining energy efficiency is also very important. Testing chips in a constrained environment, like laptops have, introduces caveats and nuances that could impact certain chips favorably or unfavorably depending on codes in use.

 

[Abwx]

Race to idle doesn't trump inherent efficiency, though, does it?

Actually that s a scam more than anything else.

Optimum efficency is when the CPU consume the same as the rest of the system, from here any turboing is a compromise between efficency and acceptable usability, because no one like an item that is not responsive enough in respect of our perception speed.