Intel "Bartlett Lake" Appears as a P/E-Core Hybrid, P-Core Only CPUs Could Soon Follow

[JustBenching]

That's because e-cores aren't efficient with power, only with die area. One p-core takes up the same space as a cluster of 4 e-cores, so e-cores are a nice way to increase core count and MT performance, but they throw power economy out of the window. I have a feeling that the 12 p-core design will consume less power than Raptor Lake does, unless Intel decides to overvolt it to the moon.

This gets a lot of people confused. When they say ecores are not efficient, what they mean is at ISO power P cores are faster. Therefore more efficient. That doesn't take into account that there are 4 as many ecores at the same space, and that ecores are clocked considerably lower. There is no chance in hell the 12P core chips will be faster / more efficient than the 14900k.

 

[AusWolf]

I think Intel is reaching our for gamers and the performance crown again. Take what was Intel's for so long before AMD took it.

They probably are reaching out for gamers, but not for the performance crown. If they did, they'd put BTL under the Core Ultra branding.

This gets a lot of people confused. When they say ecores are not efficient, what they mean is at ISO power P cores are faster. Therefore more efficient. That doesn't take into account that there are 4 as many ecores at the same space, and that ecores are clocked considerably lower. There is no chance in hell the 12P core chips will be faster / more efficient than the 14900k.

Yes, but the 4 e-cores consume more power, too, at least that's what I suspect.

 

[JustBenching]

Yes, but the 4 e-cores consume more power, too, at least that's what I suspect.

Than a single Pcore, i'd say probably yes, especially in RPL when they are pushed to 4.4ghz. But they are also faster than 1 pcore. If I remember the numbers correctly, on alderlake they were pulling around 80w for a score of 9k in CBR23. Which sounds absolutely horrible on paper (that's 5600x territory if im not mistaken), but then if you take into account that 2P cores would never reach that score even at 500 watts, puts that into perspective.

 

[chrcoluk]

That's because e-cores aren't efficient with power, only with die area. One p-core takes up the same space as a cluster of 4 e-cores, so e-cores are a nice way to increase core count and MT performance, but they throw power economy out of the window. I have a feeling that the 12 p-core design will consume less power than Raptor Lake does, unless Intel decides to overvolt it to the moon.

I expect them to be aggressive, but I will just tone it down to something sensible. e-cores are great for MT productivity stuff, but it is difficult to completely overcome scheduling problems. I think this 12 core design will also have more L3 cache as Intel boost it for every p-core.

 

[JustBenching]

I expect them to be aggressive, but I will just tone it down to something sensible. e-cores are great for MT productivity stuff, but it is difficult to completely overcome scheduling problems, and probably is as you said overall less power efficient. I think this 12 core design will also have more L3 cache as Intel boost it for every p-core.

If they indeed increase the cache, that 12p core part will be insane for games, especially the core heavy types (TLOU, CP etc.).

 

[badreligionx]

Was hoping for bartlett lake but was kind of worried they would not come to life anymore. Really happy about this extended life for the 1700 platform and will get one for sure. 12 P-Cores and hopeffully a bit more cache and gaming performance will be amazing. The newer architecture cores from arrow lake would be great of course but it almost sounds too good to be true.

Intel seems to wake up finally so I guess the pressure from AMD is finally doing its job now

 

[Dr. Dro]

Was hoping for bartlett lake but was kind of worried they would not come to life anymore. Really happy about this extended life for the 1700 platform and will get one for sure. 12 P-Cores and hopeffully a bit more cache and gaming performance will be amazing. The newer architecture cores from arrow lake would be great of course but it almost sounds too good to be true.

Intel seems to wake up finally so I guess the pressure from AMD is finally doing its job now

Raptor Cove will hold up. It'll be more interesting to see if they leave the AVX-512 capability enabled now that there are no E-cores in the picture. Certainly a worthy upgrade

 

[AusWolf]

Than a single Pcore, i'd say probably yes, especially in RPL when they are pushed to 4.4ghz. But they are also faster than 1 pcore. If I remember the numbers correctly, on alderlake they were pulling around 80w for a score of 9k in CBR23. Which sounds absolutely horrible on paper (that's 5600x territory if im not mistaken), but then if you take into account that 2P cores would never reach that score even at 500 watts, puts that into perspective.

Exactly. That's why I'm not interested in e-cores. They're good for MT performance, but with 8+ p-cores, I don't need more MT, especially not at such insane power costs. I'd much rather have 10-12 more conservative p-cores and a bit lower power consumption. That design would still be good for a bit of headroom for future gaming or background tasks. Not to mention the software scheduling nightmare which is also a big no-no for me.

Just really, shame that it came too late and I'm already happy with a new AMD platform. I would have definitely bought BTL a year or two ago, but not now.

 

[JustBenching]

Exactly. That's why I'm not interested in e-cores. They're good for MT performance, but with 8+ p-cores, I don't need more MT, especially not at such insane power costs. I'd much rather have 10-12 more conservative p-cores and a bit lower power consumption. That design would still be good for a bit of headroom for gaming.

Just really, shame that it came too late and I'm already happy with a new AMD platform. I would have definitely bought BTL a year or two ago, but not now.

Well that's kinda confusing though. If one does not need more MT performance why do you need 12P cores?


The only benefit that more Pcores have is the odd application (maybe games included) that scale above 8 cores but not above say 14. In those cases, a 12P core cpu will be faster than an 8+16. But that's a fringe scenario, in most cases the 8+16 will be faster. A lot faster. While pulling less power. It's still a long time till reviews are out but I can guarantee you - lock them both to same power and the 8+16 config of the 14900k will absolutely scorch a 12P core CPU. 12P core will probably end up sandwiched between a 13700k and a 14700k in MT performance.

 

[AusWolf]

Well that's kinda confusing though. If one does not need more MT performance why do you need 12P cores?

I don't need them, but it doesn't hurt to have them. Unlike e-cores which require special scheduling.

The only benefit that more Pcores have is the odd application (maybe games included) that scale above 8 cores but not above say 14. In those cases, a 12P core cpu will be faster than an 8+16. But that's a fringe scenario, in most cases the 8+16 will be faster. A lot faster. While pulling less power. It's still a long time till reviews are out but I can guarantee you - lock them both to same power and the 8+16 config of the 14900k will absolutely scorch a 12P core CPU. 12P core will probably end up sandwiched between a 13700k and a 14700k in MT performance.

The benefit of more p-cores is more MT performance without the need for software scheduling. Sure, 4 e-cores are faster than a single p-core, but the tradeoff of a purely p-core design is something I can live with.

 

[Onyx Turbine]

Realistically speaking, a 12P+0E design will be a regression in multithreaded performance against the existing 8P+16E processors in controlled, benchmarked environments. However, the extra high performance execution units should prove very useful in many more real-world scenarios, without the drawbacks of the 12-core Ryzen processors' split CCDs or their X3D counterparts' imbalanced hardware configuration.

E-cores also come in clusters of 4, which roughly take the same die area as one P-core. I believe you wanted something like the Royal Core design, where there are P-cores and a couple of super big cores. This was supposed to be Beast Lake and to be marketed as Core Ultra series 4 (after Panther Lake/series 3, which would be Arrow Lake's refinement still on LGA 1851), but it was unfortunately canned last year.

probably the reason they reinvigorate 1700 socket is that their consumer data shows, a vast amount of people ranging around enthusiast consumer denominator are on this socket. Also the current growth rate on the socket in terms of sales might even be higher than on the arrow lake one.
So i think nobody minds if they would even try out different chiplet designs, as realistically speaking they have the opportunity to make a legendary cpu for this platform.
If the performance is good a fair consumer price could be as high as ~300 Dollar/Eur. Releasing the most potent chip i would definitely suggest try to delay well into 2026,
as coming 2 years actually will be interesting to gauge how long current cpu computing power availabilities are reasonably priced or affordable for consumers for gaming in triple aaa games. Was kind of surprised to read that for certain upcoming titles for a minimum of 60fps a 13700k is recommended.

For the one who noticed the reason i mention an octa p core as i think this is the consumer cpu what will work well into 2035+, as we now are abandoning for gaming 4corep and can do well with souped up 6core p, the 8core p will be very solid base to improve game design programming performance etc.

 

[fry]

Well that's kinda confusing though. If one does not need more MT performance why do you need 12P cores?


The only benefit that more Pcores have is the odd application (maybe games included) that scale above 8 cores but not above say 14. In those cases, a 12P core cpu will be faster than an 8+16. But that's a fringe scenario, in most cases the 8+16 will be faster. A lot faster. While pulling less power. It's still a long time till reviews are out but I can guarantee you - lock them both to same power and the 8+16 config of the 14900k will absolutely scorch a 12P core CPU. 12P core will probably end up sandwiched between a 13700k and a 14700k in MT performance.

Well that's kinda confusing though. If one does not need more MT performance why do you need 12P cores?


The only benefit that more Pcores have is the odd application (maybe games included) that scale above 8 cores but not above say 14. In those cases, a 12P core cpu will be faster than an 8+16. But that's a fringe scenario, in most cases the 8+16 will be faster. A lot faster. While pulling less power. It's still a long time till reviews are out but I can guarantee you - lock them both to same power and the 8+16 config of the 14900k will absolutely scorch a 12P core CPU. 12P core will probably end up sandwiched between a 13700k and a 14700k in MT performance.

You're right, a 14900k would scorch a 12P CPU in multicore, which is one of the only reasons you'd need the 16 e-cores anyway. Still, 12P cores would not only be more power efficient than 8P 16E at lower power levels, but it would avoid the rare P/E core conflicts that can occur on certain programs and operating systems. It depends on preference, but the drop in multicore performance might be palatable to people like Auswolf for increased performance at lower power draw and mildly improved reliability.

Also, do keep in mind that leaks also claimed the newly released chips (201E, 211E, 251E) would be Raptor Lake S, and that the Q3 P-core only variants might house Lion Cove cores instead of Raptor Cove cores. I have no idea how this would improve efficiency or performance (or whether it would cause Arrow Lake style regression), but it is something to consider when comparing hypothetical 10/12P chips with existing 12-14th gen chips.

 

[JustBenching]

Still, 12P cores would not only be more power efficient than 8P 16E at lower power levels,

Mega doubt. We already have data on a power limited 14900k. If w1z does the same for the 12P I guess we will see what's what, but I don't see how eg. at 100w the 12P core will be faster than the 14900k. it will still be slower

 

[AusWolf]

Mega doubt. We already have data on a power limited 14900k. If w1z does the same for the 12P I guess we will see what's what, but I don't see how eg. at 100w the 12P core will be faster than the 14900k. it will still be slower

It will be slower in MT, not so much in gaming, but it will also eat less power. How it'll work out for actual efficiency is a guess on anyone's part. I'm quite curious.

 

[JustBenching]

It will be slower in MT, not so much in gaming, but it will also eat less power. How it'll work out for actual efficiency is a guess on anyone's part. I'm quite curious.

Why do peoplethink it will eat less power? Like in what way? You mean it's going to be locked to a lower power limit?

 

[AusWolf]

Why do peoplethink it will eat less power? Like in what way? You mean it's going to be locked to a lower power limit?

No, I mean even when fully unlocked it will eat less power. I think that because while a single e-core eats less power than a single p-core, there's 4 of them in a cluster which throws the picture off balance.

For example, if we suppose that an e-core eats as much as 0.4 p-cores, then 0.4 * 4 = 1.6, that means an e-core cluster eats 1.6 times more than a single p-core.

It's just my theory, though (based on the 12-13-14900K's insane unlocked MT power consumption figures), but I'm curious to find out.

 

[JustBenching]

No, I mean even when fully unlocked it will eat less power. I think that because while a single e-core eats less power than a single p-core, there's 4 of them in a cluster which throws the picture off balance.

For example, if we suppose that an e-core eats as much as 0.4 p-cores, then 0.4 * 4 = 1.6, that means an e-core cluster eats 1.6 times more than a single p-core.

It's just my theory, though (based on the 12-13-14900K's insane unlocked MT power consumption figures), but I'm curious to find out.

But cores (whether we talking about e or p ) don't need a certain amount of power. They can work with however much power intel decides to give them. When someone says "it will need less power", I think of 2 things, either you are claiming that Intel will lock them to lower power / lower clocks (which we have no indication of), or that at the same power it will be faster.

Also even if it is the case and it eats less power, the problem is - it's still slower. Why does it matter that it eats less power when you can just lock the 900k to the same power and still end up faster?

 

[AusWolf]

But cores (whether we talking about e or p ) don't need a certain amount of power. They can work with however much power intel decides to give them. When someone says "it will need less power", I think of 2 things, either you are claiming that Intel will lock them to lower power / lower clocks (which we have no indication of), or that at the same power it will be faster.

They can't decide to give them infinite power. Every chip has its limitations. Even with no power limits, they have to be locked by clock speed and voltage to make sure the chip doesn't destroy itself.

What I'm saying is, you take an unlocked CPU, give it a MT task, and you'll see lower power consumption on it than you would on a 14900K. Sure, lower performance, too. How much lower is the true question.

Also even if it is the case and it eats less power, the problem is - it's still slower. Why does it matter that it eats less power when you can just lock the 900k to the same power and still end up faster?

Yes, because of the mess around thread scheduling which I'd prefer to avoid, even if it comes with lower overall performance.

 

[chrcoluk]

Why do peoplethink it will eat less power? Like in what way? You mean it's going to be locked to a lower power limit?

Less complexity.

The default scheduler is also a bit whacked, so e.g. if any p-core is awake, it locks e-cores to max turbo clock which also raised v-core. Luckily I figured that out early, and my CPU isnt doing it now as I changed the hidden scheduler setting for it.

My plan is to keep some things I learned and have all of the background service stuff on 2-4 cores between them, so 8 cores will be idle for gaming use. Similar to what I am doing now with e-cores but without the occasional scheduling headache.

As an example some apps are heavy on background cycles, but also are interactive, so my preference is to make sure they on an e-core, but then the interactive side feels slower, that headache is gone on a 12 p core chip.

 

[JustBenching]

They can't decide to give them infinite power. Every chip has its limitations. Even with no power limits, they have to be locked by clock speed and voltage to make sure the chip doesn't destroy itself.

What I'm saying is, you take an unlocked CPU, give it a MT task, and you'll see lower power consumption on it than you would on a 14900K. Sure, lower performance, too. How much lower is the true question.


Yes, because of the mess around thread scheduling which I'd prefer to avoid, even if it comes with lower overall performance.

Well sure, there is an upper and a lower limit on what you can do with power, but I can easily see the 12p core pulling 300 watts if Intel decides to give it that much. Especially since it will be stupid easy to cool due to 12 big cores.

As you know from my previous posts the last couple of years, power itself is irrelevant for me. I wanna know what it can do with the same amount of power. And I think in this one the 900k will be the obvious winner.

There aren't any particular thread scheduling issues to speak off. Since 2021 there was just one very old game not playing ball and another one that had worse performance with ecores on (Warhammer) That's it as far scheduling is concerned. You'll see much more scheduling issues caused by the presence of HT actually than by ecores. You can even test it on your 7800x 3d, despite the lack of core counts, a lot of games will run faster with HT off.

 

[AusWolf]

Well sure, there is an upper and a lower limit on what you can do with power, but I can easily see the 12p core pulling 300 watts if Intel decides to give it that much. Especially since it will be stupid easy to cool due to 12 big cores.

That's what I doubt. Every core has a maximum clock speed and voltage that it can comfortably maintain, and I don't think it's gonna be so high to make the CPU eat 300 W.

There aren't any particular thread scheduling issues to speak off. Since 2021 there was just one very old game not playing ball and another one that had worse performance with ecores on (Warhammer) That's it as far scheduling is concerned. You'll see much more scheduling issues caused by the presence of HT actually than by ecores. You can even test it on your 7800x 3d, despite the lack of core counts, a lot of games will run faster with HT off.

I'd still like to keep my CPU as simple in principle as possible. And I can't test HT on and off in games because I run a 6750 XT - it wouldn't make much of a difference.

My biggest scheduling issue is that you need software to make it work better - software made for Windows (I'm on Linux). Not to mention, CPUs should be controlled by the BIOS, not software, imo.

 

[JustBenching]

That's what I doubt. Every core has a maximum clock speed and voltage that it can comfortably maintain, and I don't think it's gonna be so high to make the CPU eat 300 W.

300 isn't even that high tbh. That's 25w per core. That's akin to pushing 200w on a 8p core chip, even my 9800x 3d is hitting 160 and it's temps that are stopping it. With how easier intel is to cool, 25w per core would be a low figure

 

[AusWolf]

300 isn't even that high tbh. That's 25w per core. That's akin to pushing 200w on a 8p core chip, even my 9800x 3d is hitting 160 and it's temps that are stopping it.

And my 7800X3D eats 80 W on the whole package. Around 7-8 W per core. Thermals are fine, too, the only thing stopping it is the max boost clock.

With how easier intel is to cool, 25w per core would be a low figure

Easier, sure, but that much easier? Hmm... I don't know... Still sounds excessive to me. 300 W needs a large AIO, or custom water even on Intel.

 

[JustBenching]

And my 7800X3D eats 80 W on the whole package. Around 7-8 W per core. Thermals are fine, too, the only thing stopping it is the max boost clock.


Easier, sure, but that much easier? Hmm... I don't know... Still sounds excessive to me.

The 9800x 3d is running around 30c higher than the 12900k at same power using the exact same cooler and case. The 12p core part should be even easier than my 12th gen due to bigger die.

 

[AusWolf]

The 9800x 3d is running around 30c higher than the 12900k at same power using the exact same cooler and case. The 12p core part should be even easier than my 12th gen due to bigger die.

Maybe. I just want to see BTL at safe temps with a reasonably-sized cooler. We'll see when it comes out, I guess.