Discussion Intel current and future Lakes & Rapids thread

[SiliconFly]

Intel's 'Meteor Lake' CPU power will be controlled by AI

An AI-controlled algorithm will govern how Intel's upcoming Meteor Lake chip will swing between high-power processing states and low-power sleep states, making it more responsive and saving power.

www.pcworld.com

By only implementing advanced power management, MTL achieves 10~20% long battery life.
There is no room for doubt that MTL has great competitiveness.

MTL has been designed from ground up to be extremely power efficient.

(1) Intel recently confirmed that Intel 4's transistor library itself was designed with efficiency rather than performance in mind. Thats why the boost clocks aren't that high.

(2) Leaks suggest MTL has DLVR for 15% to 20% more efficiency.

(3) Intel now confirmed MTL includes a brand new AI power management tech for better efficiency.

(4) The LP E-cores in SoC tile can completely turn off the CPU tile & GPU tile while idling with massively power savings.

All these put together, MTL may comfortably DOUBLE laptop battery life!

Thats what Ultra means I think. An energy-efficient beast.

 

[Geddagod]

(1) Intel recently confirmed that Intel 4's transistor library itself was designed with efficiency rather than performance in mind. Thats why the boost clocks aren't that high.

hard to believe that with the lack of HD cells

(2) Leaks suggest MTL has DLVR for 15% to 20% more efficiency.

Where? And at what power levels?

All these put together, MTL may comfortably DOUBLE laptop battery life!

It kinda needs to lol

Thats what Ultra means I think. An energy-efficient beast

Literally just branding but ok

 

[Doug S]

All these put together, MTL may comfortably DOUBLE laptop battery life!

Even if the CPU doubles energy efficiency (color me highly skeptical) that won't come remotely close to doubling battery life, because in normal usage scenarios the display uses more power than the CPU. If you're hitting the CPU hard that's not true of course, but no laptop lasts more than a few hours with the CPU/GPU pegged so you are going to be tethered to an outlet in that scenario regardless.

 

[SiliconFly]

Even if the CPU doubles energy efficiency (color me highly skeptical) that won't come remotely close to doubling battery life, because in normal usage scenarios the display uses more power than the CPU. If you're hitting the CPU hard that's not true of course, but no laptop lasts more than a few hours with the CPU/GPU pegged so you are going to be tethered to an outlet in that scenario regardless.

Not really. I think Intel's just trying to follow M2 macbook's footsteps. Also, MTL mobile might just be well ahead AMD's mobile offerings. Thats what I mean.

 

[uzzi38]

MTL has been designed from ground up to be extremely power efficient.

(1) Intel recently confirmed that Intel 4's transistor library itself was designed with efficiency rather than performance in mind. Thats why the boost clocks aren't that high.

(2) Leaks suggest MTL has DLVR for 15% to 20% more efficiency.

(3) Intel now confirmed MTL includes a brand new AI power management tech for better efficiency.

(4) The LP E-cores in SoC tile can completely turn off the CPU tile & GPU tile while idling with massively power savings.

All these put together, MTL may comfortably DOUBLE laptop battery life!

Thats what Ultra means I think. An energy-efficient beast.

I'm so sorry for your loss.

(again)

 

[igor_kavinski]

Thats what Ultra means I think. An energy-efficient beast.

Going all in on Intel's marketing Kool-Aid, before they have even begun marketing MTL!

 

[igor_kavinski]

(4) The LP E-cores in SoC tile can completely turn off the CPU tile & GPU tile while idling with massively power savings.

I foresee a bug that happens under certain conditions where the CPU is unable to wake up the tiles

Rich people will enjoy the netbook experience too then.

 

[SiliconFly]

I'm so sorry for your loss.

(again)

Maybe it's time (again) for cat food challenge. I can send them to you for free!

 

[Khato]

Not really. I think Intel's just trying to follow M2 macbook's footsteps. Also, MTL mobile might just be well ahead AMD's mobile offerings. Thats what I mean.

I really don't understand why there's this impression that x86 - Intel in particular - battery life sucks. In their review of the Lenovo X13s - https://www.notebookcheck.net/Lenov...e-Qualcomm-Snapdragon-8cx-Gen-3.665008.0.html - notebookcheck has a nice sampling of 13" 1920x1200 x86 designs in addition to the 2560x1664 Macbook Air M2. The 'good' examples of x86 designs are within +/-10% of the ARM designs on their WiFi v1.3 battery life metric. (Load battery life isn't a particularly interesting metric as it ignores performance.) Meanwhile the 'bad' examples of x86 designs come in at 1/3 to 2/3 the battery life.

Simple fact is that even if MTL halves SoC power consumption in normal usage (web browsing and media playback) it's only going to result in a 10-20% increase in battery life.

More importantly, all it takes is one bad component choice or firmware issue on the part of a laptop OEM to cut battery life in half.

 

[Doug S]

I really don't understand why there's this impression that x86 - Intel in particular - battery life sucks. In their review of the Lenovo X13s - https://www.notebookcheck.net/Lenov...e-Qualcomm-Snapdragon-8cx-Gen-3.665008.0.html - notebookcheck has a nice sampling of 13" 1920x1200 x86 designs in addition to the 2560x1664 Macbook Air M2. The 'good' examples of x86 designs are within +/-10% of the ARM designs on their WiFi v1.3 battery life metric. (Load battery life isn't a particularly interesting metric as it ignores performance.) Meanwhile the 'bad' examples of x86 designs come in at 1/3 to 2/3 the battery life.

Simple fact is that even if MTL halves SoC power consumption in normal usage (web browsing and media playback) it's only going to result in a 10-20% increase in battery life.

More importantly, all it takes is one bad component choice or firmware issue on the part of a laptop OEM to cut battery life in half.

I think it is because of exactly the reason you say. You have to cherry pick a particular model,a nd make sure it comes the right 'U' CPU or whatever to come close to matching Macbook battery life. Do you think the average PC buyer wants to dig through reviews to find a particular model, then make sure they get the RIGHT model, selecting the correct CPU in the order configuration and so on?

With a Macbook you just buy a Macbook and you get the great battery life without the bother. Having a lot of choice can be a good thing, but humans are paralyzed by choice in things they don't consider important. Most people don't know what the hell toothpaste to buy with all the options so they just whatever they bought last time or if they can't remember just pick one that seems to do a lot of good things. It is probably similar for the average person buying PCs. They aren't following the market like people who post in Anandtech forums do, even if they get pointed at the most power efficient laptop someone else might point them to the highest performing laptop, a third person at the lightest laptop, and another person at the laptop with the best display. And there's always the draw of low price.

A lot of people hate on Apple for having so few choices, but Steve Jobs understood the psychology of the public a lot better than the beancounters in charge of the laptop lineup at Dell, etc. You know you will get a laptop with good performance, long battery life, that's light, and yeah it isn't cheap but you can't have everything. It makes the process of buying a lot less painful for the average person who doesn't want to deal with so many models each having a bunch of choices for the CPU, storage, RAM even colors in the case and software options that pay to be included at checkout like tax software or whatever. That's incredibly frustrating and overwhelming for a lot of people, which is why they ask people like us for guidance when they are forced to buy lol

So that's why Macs have that reputation, because it is true of EVERY Mac, not just if you know enough to select the right handful of models out of hundreds.

 

[SiliconFly]

I really don't understand why there's this impression that x86 - Intel in particular - battery life sucks. In their review of the Lenovo X13s - https://www.notebookcheck.net/Lenov...e-Qualcomm-Snapdragon-8cx-Gen-3.665008.0.html - notebookcheck has a nice sampling of 13" 1920x1200 x86 designs in addition to the 2560x1664 Macbook Air M2. The 'good' examples of x86 designs are within +/-10% of the ARM designs on their WiFi v1.3 battery life metric. (Load battery life isn't a particularly interesting metric as it ignores performance.) Meanwhile the 'bad' examples of x86 designs come in at 1/3 to 2/3 the battery life.

Simple fact is that even if MTL halves SoC power consumption in normal usage (web browsing and media playback) it's only going to result in a 10-20% increase in battery life.

More importantly, all it takes is one bad component choice or firmware issue on the part of a laptop OEM to cut battery life in half.

Thats not a direct comparison. A M2 macbook actually beats almost all the Intel & AMD processors listed in your link by a wide margin in almost all aspects. A good intel/amd laptop cpu & gpu comparable to or better than a M2, tend to draw a lot more power in reality. Here's a simple breakdown:

In a typical laptop, the CPU, GPU & Display are the main culprits. Things like RAM, SSD, Keyboard, Touchpad, Wifi, Bluetooth, Speaker, Mic, Camera don't drain the battery much (RAM, SSD, WIFI tend to draw some power, but only when stressed, esp. the last two).

Just for the sake of argument, under light load, if the CPU & GPU & Display don't draw any power from the battery, the laptop will run for many days on a single charge! Thats how much power these 3 components draw.

If AMD & Intel can significantly reduce the power draw of CPU+GPU, the laptop battery life increases significantly. Like from just 10 hours to all day battery life like a M2 Mac (assuming normal usage with reasonable display brightness level).

 

[Khato]

Thats not a direct comparison. A M2 macbook actually beats almost all the Intel & AMD processors listed in your link by a wide margin in almost all aspects. A good intel/amd laptop cpu & gpu comparable to M2, tend to draw a lot more power in reality. Here's a simple breakdown:

In a typical laptop, the CPU, GPU & Display are the main culprits. Things like RAM, SSD, Keyboard, Touchpad, Wifi, Bluetooth, Speaker, Mic, Camera don't drain the battery much (RAM, SSD, WIFI tend to draw some power, but significantly less).

Just for the sake of argument, if the CPU & GPU & Display don't draw any power from the battery, the laptop will run for many days on a single charge! Thats how power these 3 components draw.

If AMD & Intel can significantly reduce the power draw of CPU+GPU, the laptop battery life increases significantly. Like from just 10 hours to all day battery life like a Mac (assuming normal usage with reasonable display brightness level).

Heh, the same test protocol applied to different laptops isn't a direct comparison? What is then? The one point I will agree with is that the M2 macbook air is at a notable disadvantage due to the higher resolution display.

Rather than talking in the abstract, why not use actual numbers? To obtain the ~16 hour run times reported for each with their ~50Wh batteries requires an average power consumption of approximately 3W. Now taking that Lenovo X13s as an example, the review unit is using a BOE NV133WUM-N61 panel, which according to panelook specifications has a maximum backlight power of 2.62W for 300 nits brightness. Scaling that down to 150 nits likely brings you to the vicinity of 1.25W for just the panel backlight. If you go with all other system power draw at 0.25W (low estimate) that'd leave 1.5W average for the SoC. What happens if that SoC power consumption drops by a third? Well, then total power consumption goes down to 2.5W and a 16 hour battery life increases by 20% to 19.2 hours.

Majority of the x86 laptops with 'bad' battery life either have firmware that's causing the SoC to not idle correctly, cheap screens that draw twice the power, or poor implementations of other hardware in the system that cause constant parasitic consumption. Switching from RPL to MTL in a laptop design that gets 10 hours on a 50Wh battery isn't going to magically fix the other problems and result in a 16+ hour battery life. It's just going to reduce the fraction of power that goes to the SoC.


On a related subject, I wonder if Intel's going to get notably stricter with the Evo requirements now? That general strategy is the best bet x86 has to simplify choices for the consumer, it's just that previous iterations weren't really that challenging for the OEMs to meet.

 

[SpudLobby]

Well we'll be able to put this one to the test.

Sort of, anyway.

Assume this is referring to the MTL refresh on Intel 3, or the ARL U series with RWC on Intel 3 right?

 

[SiliconFly]

Heh, the same test protocol applied to different laptops isn't a direct comparison? What is then? The one point I will agree with is that the M2 macbook air is at a notable disadvantage due to the higher resolution display.

Rather than talking in the abstract, why not use actual numbers? To obtain the ~16 hour run times reported for each with their ~50Wh batteries requires an average power consumption of approximately 3W. Now taking that Lenovo X13s as an example, the review unit is using a BOE NV133WUM-N61 panel, which according to panelook specifications has a maximum backlight power of 2.62W for 300 nits brightness. Scaling that down to 150 nits likely brings you to the vicinity of 1.25W for just the panel backlight. If you go with all other system power draw at 0.25W (low estimate) that'd leave 1.5W average for the SoC. What happens if that SoC power consumption drops by a third? Well, then total power consumption goes down to 2.5W and a 16 hour battery life increases by 20% to 19.2 hours.

Majority of the x86 laptops with 'bad' battery life either have firmware that's causing the SoC to not idle correctly, cheap screens that draw twice the power, or poor implementations of other hardware in the system that cause constant parasitic consumption. Switching from RPL to MTL in a laptop design that gets 10 hours on a 50Wh battery isn't going to magically fix the other problems and result in a 16+ hour battery life. It's just going to reduce the fraction of power that goes to the SoC.


On a related subject, I wonder if Intel's going to get notably stricter with the Evo requirements now? That general strategy is the best bet x86 has to simplify choices for the consumer, it's just that previous iterations weren't really that challenging for the OEMs to meet.

Like you mentioned in the 1st statement, these aren't direct comparisons. Mac has a higher resolution display & more importantly mac cpu+gpu has higher performance than intel/amd counterparts in your list. They shouldn't even be compared in the first place.

Performance wise, if you want to compare a m2 mac with an equivalent intel laptop, we should probably compare it with a 13th gen i7 mobile at least for the sake of equality. This is where the mac truly shines in power efficiency.

But MTL has the potential to change the game. We'll know soon enough.

 

[Henry swagger]

Intel Discloses 8 Core & 528 Thread CPU Based on RISC Architecture, 66 Threads Per Core

During Hot Chips 2023, Intel showed off a brand new CPU design featuring 8 cores but a massive 528 threads based on RISC.

wccftech.com

Impressive risc 5 cpu by intel but manufactured on tsmc 7nm node.. smt on steroids 😄

 

[DrMrLordX]

If it is indeed the case, that Arrow Lake will use Intel 3 in one of its versions instead of 20A, it makes a lot of sense to me. In fact, I thought it was dumb for Intel to pin its roadmap on a node that is so far out, instead of a node that should be a lot more solid at the anticipated time of the release.

Seems more likely that there will be TSMC N3E Arrow Lake rather than Intel 3.

This slide should put to rest the theory that Sierra Forrest will have multiples of the 144 cores. This tweet says that 288 cores would be available - in 2 socket servers:

https://twitter.com/x/status/1696315800792309934

Wait what? Intel allegedly already canned the ~300c and ~500c Sierra Forest parts. Now they're killing 288c Sierra Forest-AP?

That leaves only the 144c part remaining.

 

[mikk]

Intel is going to need to like 80% more battery life to match AMD. There's plenty of room to doubt.

How do you get 80%? Are you saying AMD has a 80% better battery life when compared with two comparable devices? Let's say Intel gets 10 hours of battery life and AMD gets 18 hours, correct?

 

[SiliconFly]

How do you get 80%? Are you saying AMD has a 80% better battery life when compared with two comparable devices? Let's say Intel gets 10 hours of battery life and AMD gets 18 hours, correct?

I think it's more a lighthearted statement than actual numbers.

 

[Henry swagger]

Intel To Utilize AI To Make Next-Gen Meteor Lake CPUs More Power Efficient & Faster

Intel Meteor Lake CPUs aren't just going to make use of AI to make user experiences better but also improve its own internal capabilities.

wccftech.com

With meteor lake intel will increase laptop market share to 95%

 

[igor_kavinski]

With meteor lake intel will increase laptop market share to 95%

Oh no. Gotta sell all my belongings to afford a Meteor Lake laptop or I will be left behind and shamed by my peers for not being current. This is most terrible, most frightening, most urgent. I might need to open a GoFundMe page

 

[SiliconFly]

Intel To Utilize AI To Make Next-Gen Meteor Lake CPUs More Power Efficient & Faster

Intel Meteor Lake CPUs aren't just going to make use of AI to make user experiences better but also improve its own internal capabilities.

wccftech.com

With meteor lake intel will increase laptop market share to 95%

Actually I thought it was a joke too. But reality is a very different beast.

Considering Intel already has 88% laptop market share & if MTL delivers as promised, Intel might be able to do just that actually. No kidding.

 

[Doug S]

Actually I thought it was a joke too. But reality is a very different beast.

Considering Intel already has 88% laptop market share & if MTL delivers as promised, Intel might be able to do just that actually. No kidding.

I assume that's 88% Windows PC market share because I'm pretty sure Apple is at or pretty close to 12% share of laptops all by itself given that the Mac market has a far higher share of laptops vs desktops than the Windows PC market.

 

[A///]

very cool high thread count processor. I'd love to see it against a similar true core product in a head to head benchmarking suite in future if it's ever produced.

That leaves only the 144c part remaining for them to cancel.

yep

 

[A///]

I assume that's 88% Windows PC market share because I'm pretty sure Apple is at or pretty close to 12% share of laptops all by itself given that the Mac market has a far higher share of laptops vs desktops than the Windows PC market.

Apple has the reins on university students. My niece who I've mentioned to you before mentioned once to me in passing that about 96% of the student she saw at school used MBP or Airs. Her colleagues at TI seem to prefer MBP outside of work, too.

I like Windows but a MBP is a much better experience and feels more versatile even with the new ARM design than a Windows laptop. Let's not ignore it looks damn good compared to 99% of windows laptops.

 

[SiliconFly]

I assume that's 88% Windows PC market share because I'm pretty sure Apple is at or pretty close to 12% share of laptops all by itself given that the Mac market has a far higher share of laptops vs desktops than the Windows PC market.

x86-64