News Panther Cove will reportedly arrive with big IPC improvements, support for Intel APX

[bit_user]

There are multiple battery life tests where the Intel cpus (with smaller battery than the competiton) are on the top of the chart. I posted this a few days ago

While battery life is obviously quite relevant to laptop users, I think it's important to point out that it's rather different that compute-efficiency of the CPU, itelf. That's the aspect I was trying to address.

A blanket statement like "the most efficient x86 CPU" means we need to fully explore the topic of compute efficiency, and that's also best done using an external monitor, because if we're really testing the CPU itself, then we don't want things like variability between laptop screens cluttering up the measurements. Notebook Check did exactly that - their data I quoted was collected using an external screen (and presumably the systems running on AC power).

 

[bit_user]

thought it was supposed to be a standard convention in any writing to spell it out at first use.

Given the liberties people have taken with the term, I'm not sure spelling it out would've been that much more helpful. I think that, if they want to make their articles more accessible, they should probably have included a link to explain the subject.

 

[TheHerald]

While battery life is obviously quite relevant to laptop users, I think it's important to point out that it's rather different that compute-efficiency of the CPU, itelf. That's the aspect I was trying to address.

A blanket statement like "the most efficient x86 CPU" means we need to fully explore the topic of compute efficiency, and that's also best done using an external monitor, because if we're really testing the CPU itself, then we don't want things like variability between laptop screens cluttering up the measurements. Notebook Check did exactly that - their data I quoted was collected using an external screen (and presumably the systems running on AC power).

It's not different though. If 2 laptops do the same thing (eg video playback) then the one that dies last is more efficient. Now yes, there is definitely laptop variability, but you don't get 4 hours extra battery life with 10% smaller battery due to laptop variability.

 

[bit_user]

It's not different though.

It most definitely is. Battery life tests usually don't run the CPU at max utilization and also use different frequency governor settings.

The most you can say is that a more efficient CPU will correlate with better battery life, but there's a lot else that goes into each category of compute efficiency and battery life than just that.

 

[TheHerald]

It most definitely is. Battery life tests usually don't run the CPU at max utilization and also use different frequency governor settings.

The most you can say is that a more efficient CPU will correlate with better battery life, but there's a lot else that goes into each category of compute efficiency and battery life than just that.

Well okay, so the only other explanation is that every laptop company made much better laptops today than they were making 6 months ago. Come on now....

And ignore AMD, look at the difference between lunar and meteor lake. Same laptop (zen book 14) lunar lake got 6 hours extra battery life with a smaller battery on local video playback. That's 70% extra battery life. That's not down to laptop variability. Especially the local playback isn't using the wifi chips and all that, it's basically just the SOC and the ram (and of course the screen)

 

[bit_user]

Same laptop (zen book 14) lunar lake got 6 hours extra battery life with a smaller battery on local video playback.

If video playback were a perfect test of CPU efficiency, then why isn't it the industry standard? Why doesn't TechPowerUp use it for their CPU efficiency metrics?

I'm telling you: a lot more goes into the efficiency of a CPU than what's measured by battery life tests. That's why Notebook Check looked at the two areas separately.

 

[TheHerald]

If video playback were a perfect test of CPU efficiency, then why isn't it the industry standard? Why doesn't TechPowerUp use it for their CPU efficiency metrics?

If any other test is the perfect test why isn't it the industry standard etc. That's not a great argument my man. Maybe TPU is wrong for not using it. Btw it is the industry standard for laptops (not even joking, pretty much all laptop reviews are using local and online video playback).

Nobody said it's the perfect test. There is not a single test that is the perfect one but local video playback isn't the only one that lunar lake excels at. The videos I posted cover local, online playback, gaming and light workloads (excels, powerpoints, browsing etc.) which is what the target for thin and light laptops is. And lunar lake excels in all of those.

 

[bit_user]

it is the industry standard for laptops (not even joking, pretty much all laptop reviews are using local and online video playback).

It's not the only way people test battery life. I've seen some reviewers use a scripted loop of office apps or web browsing. The point is to try and give prospective buyers relevant, real-world data on how different laptops will run on battery. That's it.

It's not designed to tease out efficiency differences between different CPU microarchitectures or models. There are much better tests for that.

local video playback isn't the only one that lunar lake excels at. The videos I posted cover local, online playback, gaming and light workloads (excels, powerpoints, browsing etc.) which is what the target for thin and light laptops is. And lunar lake excels in all of those.

Notebook Check also tested some benchmark suites and web usage, as mentioned in my post. Performance-wise, it lost to at least one of the AMD AI 9 HX models in every one of those: PCMark 10, CrossMark, WebXPRT 4, WebXPRT 3, Jetstream 2, Speedometer 3.0.

Again, people should check it out for themselves:

https://www.notebookcheck.net/Intel...its-everyday-efficiency-is-good.893405.0.html
​

 

[TheHerald]

It's not the only way people test battery life. I've seen some reviewers use a scripted loop of office apps or web browsing. The point is to try and give prospective buyers relevant, real-world data on how different laptops will run on battery. That's it.

It's not designed to tease out efficiency differences between different CPU microarchitectures or models. There are much better tests for that.


Notebook Check also tested some benchmark suites and web usage, as mentioned in my post. Performance-wise, it lost to at least one of the AMD AI 9 HX models in every one of those: PCMark 10, CrossMark, WebXPRT 4, WebXPRT 3, Jetstream 2, Speedometer 3.0.

Again, people should check it out for themselves:

​

https://www.notebookcheck.net/Intel...its-everyday-efficiency-is-good.893405.0.html​

​

I don't think you are being fair (oh well, what a surprise). Yes - it lost in PCmark10 performance by ~10% to the amd chip, but the amd chip consumed 60% more power....it's right there on the article you just linked man.

 

[-Fran-]

Video playback is usually handled by dedicated pipelines nowadays, unless you explicitly disable those and, for laptops in particular: battery size, screen type and even RGB stuff will affect it. Given the power envelope of mobile platforms, all the additional power draw from satelite components matters, so isolating "the CPU" in a laptop test is incredibly difficult when you're talking <35W.

The only way to test it would be to completely isolate the CPU or find a way to properly normalize the results in a credible manner.

EDIT: I forgot to mention that they're close (LNL and StrixH). Close enough to really not matter that much. That is good IMO. It means Intel has finally managed to close an important gap, but it's not quite there yet, I'd say.

Regards.

 

[bit_user]

I don't think you are being fair (oh well, what a surprise).

The data is quite clear. Perhaps you're just confused?

Yes - it lost in PCmark10 performance by ~10% to the amd chip, but the amd chip consumed 60% more power....it's right there on the article you just linked man.

You had switched from talking about efficiency to talking about performance. Now, you're switching back to efficiency when the Lunar Lake CPU lost on performance?

Anyway, I'll grant you that the article only looked at one Lunar Lake model, and it wasn't the top end. So, it surely could've seen better performance, but then the efficiency metrics would also suffer.

 

[TheHerald]

The data is quite clear. Perhaps you're just confused?


You had switched from talking about efficiency to talking about performance. Now, you're switching back to efficiency when the Lunar Lake CPU lost on performance?

Anyway, I'll grant you that the article only looked at one Lunar Lake model, and it wasn't the top end. So, it surely could've seen better performance, but then the efficiency metrics would also suffer.

Where did I talk about performance? The whole argument is about whether lunar lake is efficient or not. The notebookcheck article you referenced has lunar lake curbstomping AMD on efficiency by more than ~50% perf / watt.

 

[bit_user]

Where did I talk about performance?

You're saying these claims were all in reference to efficiency?` So, they measured perf/W or Joules that you can quote for us?

The videos I posted cover local, online playback, gaming and light workloads (excels, powerpoints, browsing etc.) which is what the target for thin and light laptops is. And lunar lake excels in all of those.

The notebookcheck article you referenced has lunar lake curbstomping AMD on efficiency by more than ~50% perf / watt.

LOL, no. I broke it down quite clearly, in Post #23. The only place where the 258V beat the best Ryzen AI HX they tested on efficiency was CB24 single-threaded, where it had 33.7% better points/W.

However, when it came to multi-threaded, the best AMD CPU they tested delivered 11.3% more points/W, both running at stock power. In their most efficient configurations tested, the gap increased to 30.6% more points/W, in AMD's favor.

 

[TheHerald]

You're saying these claims were all in reference to efficiency?` So, they measured perf/W or Joules that you can quote for us?

Yes, video playback is a reference to efficiency. What kind of actual performance would you measure from video playback? They all deliver the same performance, the only variable is how much power were they drawing, which directly coresponds to battery life

LOL, no. I broke it down quite clearly, in Post #23. The only place where the 258V beat the best Ryzen AI HX they tested on efficiency was CB24 single-threaded, where it had 33.7% better points/W.

However, when it came to multi-threaded, the best AMD CPU they tested delivered 11.3% more points/W, both running at stock power. In their most efficient configurations tested, the gap increased to 30.6% more points/W, in AMD's favor.

You need to work on your reading comprehension.

Man you referenced PCmark 10, Webmark etc.

So according to the test you linked, in PCmark10 the ryzen part is consuming 60% more power for 11% more performance. They even have a huge graph with CPU package power during the whole PCmark run. What the heck are you even talking about? They even have a whole paragraph named "POWER CONSUMPTION DURING EVERYDAY USE" where they clearly state how much more efficient lunar lake is. That's from the test YOU posted to show that it is not the case.

Spoiler

So far, we have only compared the performance/efficiency under full load, which of course isn't representative of the everyday requirements of the majority of users. However, this is also different for every user, which is why we decided to compare the power consumption during a PCMark 10 test (duration ~22 minutes). We compared the CPU package power of the Core Ultra 258V with the Core Ultra 7 155H and the Ryzen AI 9 HX 370, each with the standard power limits. Although the Core Ultra 7 258V was at a slight disadvantage here due to its RAM, its average power consumption remained significantly lower than that of the other two competitors. For the Lunar Lake chip, we noted just under 9 watts, for the AMD Zen 5 processor it was just over 14 watts and the Meteor Lake processor required almost 16 watts.

Goddamn man

 

[bit_user]

Yes, video playback is a reference to efficiency.

Funny that you suddenly stopped talking about the other workloads you mentioned, which I quoted, and went back to video playback.

What kind of actual performance would you measure from video playback?

As Fran said, video playback isn't measuring CPU performance, since it uses hardware video decoders. So, we're really not talking about CPU performance or efficiency. Since the person I quoted was specific to say "x86", I think it was really in reference to the CPU cores and not the media playback engine.

Man you referenced PCmark 10, Webmark etc.

So according to the test you linked, in PCmark10 the ryzen part is consuming 60% more power for 11% more performance. They even have a huge graph with CPU package power during the whole PCmark run. What the heck are you even talking about? They even have a whole paragraph named "POWER CONSUMPTION DURING EVERYDAY USE" where they clearly state how much more efficient lunar lake is.

The graph was kind of unintelligible, due to so many CPUs being plotted, and I didn't see the averages listed in the text. I wonder why they had no bar chart with the averages, as if the only power figures that mattered were the 258V and the HX 370. Would've been interesting to see what the HX 365 burned, especially given that it even edged out the HX 370 in PC Mark 10 performance.

I have to say that average power during PC Mark 10 isn't the same as seeing the average power during some of the web benchmarks, because some parts of PC Mark are bottlenecked on I/O and therefore we're not really seeing the efficiency of either single-threaded or multi-threaded.

 

[TheHerald]

Funny that you suddenly stopped talking about the other workloads you mentioned, which I quoted, and went back to video playback.

I was talking about efficiency in every single post. Every single one...

As Fran said, video playback isn't measuring CPU performance, since it uses hardware video decoders. So, we're really not talking about CPU performance or efficiency. Since the person I quoted was specific to say "x86", I think it was really in reference to the CPU cores and not the media playback engine.

Hardware video decoders huh? I wonder where those are located at. A device called iGPU which is part of the CPU huh? Interesting

The graph was kind of unintelligible, due to so many CPUs being plotted, and I didn't see the averages listed in the text. I wonder why they had no bar chart with the averages, as if the only power figures that mattered were the 258V and the HX 370. Would've been interesting to see what the HX 365 burned, especially given that it even edged out the HX 370 in PC Mark 10 performance.

I have to say that average power during PC Mark 10 isn't the same as seeing the average power during some of the web benchmarks, because some parts of PC Mark are bottlenecked on I/O and therefore we're not really seeing the efficiency of either single-threaded or multi-threaded.

How surprising, the link you yourself posted to show otherwise now isn't enough. Man, there are 50 reviews out there, all of them have the Intel cpu topping battery and efficiency charts where it matters for thin and light laptops.

 

[bit_user]

I was talking about efficiency in every single post. Every single one...

Then you should use clearer wording. It seems to happen a lot, where you make some vague statement and then later claim to be talking about something else. If you'd be more careful in your wording, I think some of the back-and-forth could be avoided.

Hardware video decoders huh? I wonder where those are located at. A device called iGPU which is part of the CPU huh? Interesting

First of all, no - it is technically not in a CPU, but rather a SoC. People have informally referred to them as CPUs, but the CPU is only a subset of the SoC.

Furthermore, as I mentioned, the iGPU has nothing to do with what instruction set the cores are executing. If someone is talking about efficiency as an x86 CPU, then they should be talking about the CPU cores and not the SoC writ large.

How surprising, the link you yourself posted to show otherwise now isn't enough. Man, there are 50 reviews out there, all of them have the Intel cpu topping battery and efficiency charts where it matters for thin and light laptops.

Like I've already said, it topped none of the charts in Notebook Check's testing! The most efficient was Apple M3! In the case of multi-threaded, even AMD Ryzen AI 9 370 beat out the 258V.

 

[rtoaht]

I didn't see anything in the video to substantiate that claim. I went to the section marked "Efficiency Testing", but I don't see any graphs or data being presented. Please provide timestamps of anything in that ~1 hour video you'd like us to see, because there's no way I'm watching the whole thing and even trying to search through and parse the transcript is a bit of a chore, when I don't even know what I'm looking for.

The link I provided was already timestamped. If you just clicked, it would have taken you to 44:33 within that video. Not sure how you missed that.