News Intel Panther Lake processors could pack up to 16 cores, maximum of four performance cores according to leak

[bit_user]

Anyone trying to claim these E-cores are faster than Zen4 are either a TROLL or ignorant as can be.

Intel tells us they've disproportionately improved the E-cores, in the Skymont generation (of Lunar & Arrow Lake). I think the claim is probably a stretch, but not as big as you're suggesting.

Source: https://www.tomshardware.com/pc-com...pc-gain-for-e-cores-16-ipc-gain-for-p-cores/3
​

Basically, they claim a 38% improvement in integer IPC and 68% improvement in floating point over Crestmont, which was already a double-digit improvement (IIRC) over Gracemont. So, Intel has really been narrowing the gap between P-cores and E-cores. I think not as close as AMD's C-cores are to their full cores, but coming close to it.

AMD has yet to launch its x3d zen5 chips

The main reason X3D models are more efficient than the X models is because their clock speed is kept down in the range where they run more efficiently. Since we're talking about laptop CPUs, here, their clocks will already be more constrained than the desktop X-models and therefore we shouldn't see as much efficiency benefit from the 3D cache version.

while zen4 is way more efficient than Intel 13th and 14th gen

Because Intel was stuck on their 7 nm-class node, Raptor Lake focused on frequency-scaling (deja vu, anyone?). AMD has been able to go to smaller nodes, which has enabled them to derive more of their performance improvements from IPC gains. That said, the Zen 4 X-models got stupid with power limits to milk the last few drops of performance via the last couple hundred MHz. When you look at how much more efficient the non-X models are, by comparison, it really puts this into perspective.

 

[thestryker]

Also slightly off the mark, since those aren't complemented by 8x E-cores, but rather 8x Zen 5C cores. The distinction being that the latter have virtually the same IPC as the regular Zen 5 cores and feature SMT.

No it's absolutely on the mark as complaining about only getting "4 P-cores" on the face of it is a stupid argument which is the point I am making. It doesn't matter how many "P-cores" there are only the absolute performance of the whole and its capability. This type of arbitrary complaining is the same thing as saying 6 cores aren't good enough when 6 core parts can easily be faster than 8+ core predecessors.

 

[kjfatl]

Intel Lunar Lake compute tiles are built buy TSMC just like AMD. There won't be any competitive advantage node wise for either company. If Intel's fabs fail, you will all get your fanboy products built in the same factory for the foreseeable future . Call it a Toyota or BMW, it will be pretty much the same thing. AMD rode TSMC to success, and Intel's fab issues have held them back. StRaIgHt fActS, mIC dRoP!

The best thing for AMD is to have Intel fabs be successful, and to have Intel build some of their parts. The only players hurt in this scenario are Samsung and TSMC. It will kill TSMC's margins while increasing the profitability of both AMD and Intel. Competition is a good thing for the consumer. This will also keep TSMC from getting lazy, closing the door to Chinese domination of semiconductor products. (Last year China invested more in chip manufacturing equipment that TSMC and Intel combined).

 

[bit_user]

What she has done with AMD is nothing short of remarkable. Give that woman a raise. I've ran Intel processors since 1997 but when it came time for this new build AMD was the obvious choice.

Lisa deserves a ton of credit for many things, including figuring out how to keep the lights on and focus the teams on the right things in those few years before Zen launched, but she didn't initiate the Zen project or even bring back Jim Keller. Those both happened under her predecessor.

 

[bit_user]

I can give you a hint of what to expect.
They will be power limited.
More p cores means more power throttling. Not much of a point in a mobile chip.

The question isn't whether they hit the power (or thermal) limit before peak clocks, because it's quite likely even a 2P + 8E configuration would. The question is how performance would be impacted.

I see a lot of complaints about CPU performance for the thing and light models and no mention of battery live. Most laptop users would give up performance for battery life. Personally, I would prefer a model with a '24 hour rated battery.

Most laptop reviews cover battery life, although it's significantly impacted by things like what kind of screen they use (OLED is traditionally much less power-efficient).

 

[bit_user]

Hello experts, A question about x86 multicores these days. It used to be that if one core suffered an uncorrectable machine check such an SEU, there were so many opcodes in flight, and cache entanglements it was impossible to restart or resume a core without rebooting the whole processor. Is this still the case now that core counts are on the rise?

Well, I'll try to answer that without answering that, because it's not really my area of expertise. I think it depends on what the core was doing. If it occurs while the core is in user space, then you can probably assume only that process is affected and simply abort it with probably a SIGBUS. At that point, the kernel could conceivably just take the core offline and move on. However, if it's executing in the kernel, then I think you can't safely assume the kernel's state hasn't been invalidated and should probably just raise a kernel panic.

That would be pretty safe, unless the error were something like an uncorrectable error in the cache hierarchy, at which point you don't know how it maps to a process. The most sensible thing to do would probably just be to kernel panic. If this sort of thing starts happening often enough, devising better strategies would be warranted. If not, then it's not worth the trouble and complexity of doing anything more fancy (IMO).

For mission-critical workloads, it's already common practice to do things like mirroring and checkpointing. So, these sorts of strategies aren't as problematic as they might seem.

 

[bit_user]

Intel Lunar Lake compute tiles are built buy TSMC just like AMD. There won't be any competitive advantage node wise for either company.

Lunar Lake uses TSMC 3B (pretty sure; but could be 3E), while AMD is using N4P. So, Intel has a real node advantage between Lunar Lake and Strix Point.

If Intel's fabs fail, you will all get your fanboy products built in the same factory for the foreseeable future .

TSMC doesn't currently have the volume to take on all of Intel's CPU business. It would take them several years, at least, to add enough capacity.

Call it a Toyota or BMW, it will be pretty much the same thing. AMD rode TSMC to success, and Intel's fab issues have held them back. StRaIgHt fActS, mIC dRoP!

Your Toyota vs. BMW analogy strikes me as weird, given the point you seem to be making. Manufacturing node isn't entirely determinitive of CPU performance or efficiency, although it's clearly a factor. One example would be to look at how much Raptor Lake improved over Alder Lake, even though the manufacturing nodes were nearly identical.

 

[Thunder64]

I see a lot of complaints about CPU performance for the thing and light models and no mention of battery live. Most laptop users would give up performance for battery life. Personally, I would prefer a model with a '24 hour rated battery.

How often to you really need 24 hours of battery life? I'm genuinely curious. It can't be all that often, or am I wrong? More is better sure, and maybe I am lucky to have power readily available most of the time.

 

[rluker5]

The question isn't whether they hit the power (or thermal) limit before peak clocks, because it's quite likely even a 2P + 8E configuration would. The question is how performance would be impacted.

I agree and the performance in question is the balance between single core and well multithreaded workloads.
If the scheduling works well enough to put the high performance at low # of core use tasks on the P cores I bet those 4 cores could eat up the entire power budget just fine on their own. Give it 8 P cores and they will probably just split the load and run slower and give worse single core performance for what? The 12 more E cores should already give all of the multicore performance that one could reasonably expect out of a mobile chip.

It seems like adding more wheels to a motorcycle. If you want a pickup truck to do some heavy towing it is going to take a lot of gas. If you want a zippy efficient motorcycle it isn't going to tow as large a boat as a pickup. Those 2 types of vehicles perform better at their tasks when they are designed for them. Some 3 wheeled pickup truck motorcycle hybrid isn't good for most users.

But of course I could be wrong. Reviews will clarify whether the 4 P core is a good design or not. It seems likely to me, but something could certainly not work as planned. We will see more with LNL since that also has the 4P thing going on.

 

[kjfatl]

How often to you really need 24 hours of battery life? I'm genuinely curious. It can't be all that often, or am I wrong? More is better sure, and maybe I am lucky to have power readily available most of the time.

The 24 hour life quoted will be under the best of conditions with a new battery. In real-life conditions you will see something like 11 hours. When the battery gets near it's end of life it will be less.
It would be nice to have a laptop that can be used during a long day at work packed with meetings.

On the other side of things are the Dell portable workstations. The Precision 7780 w/ 64G RAM can burn through its battery in 45 minutes. They are far from 'thin and light', but will benefit when Intel moves its processors to 18A.

 

[bit_user]

I agree and the performance in question is the balance between single core and well multithreaded workloads.
If the scheduling works well enough to put the high performance at low # of core use tasks on the P cores I bet those 4 cores could eat up the entire power budget just fine on their own. Give it 8 P cores and they will probably just split the load and run slower and give worse single core performance for what?

Sorry, but what does the number of P-cores have to do with ST performance? We've seen plenty of examples of CPUs with low base clocks, which is the lower-bound on all-core clocks, and much higher single-core maxes (plus turbo tables with high boost frequences for smaller numbers of cores).

The 12 more E cores should already give all of the multicore performance that one could reasonably expect out of a mobile chip.

For whom, though? Look at Meteor Lake, where Intel decided to use a 6P + 8E + 2LPE configuration, yielding 22 threads. This Panther Lake will have 4P + 8E + 4LPE, yielding 16 threads (not 20, due to the lack of HT in the P-cores). Intel has a recent track record of delivering more than 16 threads for laptop users, even in their mid-tier models.

A lot of software developers would tell you they can use more than 16 threads for building and testing their code. My dev box at work has 24 threads and I could certainly use more.

It seems like adding more wheels to a motorcycle. If you want a pickup truck to do some heavy towing it is going to take a lot of gas. If you want a zippy efficient motorcycle it isn't going to tow as large a boat as a pickup. Those 2 types of vehicles perform better at their tasks when they are designed for them. Some 3 wheeled pickup truck motorcycle hybrid isn't good for most users.

Sure, not every laptop CPU needs a ton of threads, but just because you have a bunch of cores doesn't mean it's inefficient. Look at this perf/W curve for Redwood Cove vs. Lions Cove:

​

The data is not presented with units, but perf/W curves like this are the norm. Performance always tapers off, as you near the high end of the power envelope. In this case, their hypothetical curve has the new P-cores delivering 84.3% as much performance at half of the peak power. So, if you start with a number of P-cores that can all reach peak GHz within the power envelope and then double the number, without increasing the power budget, you end up with 68.6% more performance. That's an artificial example, I know, but the phenomenon is real and explains how server CPUs are able to scale performance to hundreds of cores while only consuming a few times more power than the TDP of performance desktop CPUs.

I want to clarify that I'm not taking a firm stand that "6 P-cores good; 4 P-cores bad" (with apologies to George Orwell), but I think it's interesting and will be looking closely at their scaling data to see if they did it really because the marginal benefit of more P-cores would be negligible or maybe more for area/cost reasons. For one thing, I'm dying to know just how the area of Skymont cores compares to Lions Cove.

P.S. a thought that's been kicking around my head, for a while, is that maybe one of the reasons they decided to grow Skymont cores was for thermal density reasons. It'll be really interesting to see how much power they can each burn, once we know their area, and compare that to other CPUs on a similar node.

 

[Thunder64]

The 24 hour life quoted will be under the best of conditions with a new battery. In real-life conditions you will see something like 11 hours. When the battery gets near it's end of life it will be less.
It would be nice to have a laptop that can be used during a long day at work packed with meetings.

On the other side of things are the Dell portable workstations. The Precision 7780 w/ 64G RAM can burn through its battery in 45 minutes. They are far from 'thin and light', but will benefit when Intel moves its processors to 18A.

And there is no access to power or in between meetings? For me it used to be lack of power on aeroplanes but that hasn't been a problem in some time. I see it as a "nice to have" but not necessary.

 

[usertests]

Per leaks, Panther Lake will be for mobile, probably the successor to Lunar Lake. Nova Lake will be for desktop and mobile, replacing Arrow Lake. In between is Arrow Lake Refresh.

https://videocardz.com/newz/intel-n...-desktops-panther-lake-reportedly-mobile-only

This is critical in understanding the story, and to prevent panic. Lunar Lake only has 4 P-cores. Although there being a Panther Lake-U/H/HX??? all alongside Arrow Lake mobile is confusing when Lunar Lake is just one thing.

Based on the July story, Panther Lake-U/H get 4 Xe3 "cores", which could make it a sidegrade/downgrade in graphics performance from Lunar Lake's 8 Xe2 cores. Intel has way too many product lines, as usual. At least now they are assembling them from "tiles" instead of making a bunch of different monolithic dies.

 

[TheHerald]

zen4 while zen4 is way more efficient than Intel 13th and 14th gen

Yeah, not true. ISO watts intel is more efficient in most segments in all types of workloads (ST / MT / Gaming / Idle). If you don't like ISO comparisons and wanna go out of the box then intel wins in every segment, not just most of them.

Why people like you claim otherwise, I really don't understand. The heck

As of today, 13700 (291$) vs 7700x (281$), at same power (~125w) the 13700 is around 25% faster and 25% more efficient in something like cinebench, corona etc. But sure, let's pretend otherwise, why not.

 

[bit_user]

Yeah, not true. ISO watts intel is more efficient in most segments in all types of workloads (ST / MT / Gaming / Idle).

If you don't like ISO comparisons and wanna go out of the box then intel wins in every segment, not just most of them.

Only when you cherry-pick the models and the power levels to make Intel look good.

AMD's 7900 and 8000G series show the potential of Zen 4 to be very efficient, indeed.

​

As of today, 13700 (291$) vs 7700x (281$), at same power (~125w)

The 7700X has a TDP of 105 W, not 125 W. You can't go around making a big fuss about iso-power and then just disregard a discrepancy of 19%, Not only that, but the idea of taking two CPUs and running them both above TDP, as if such a comparison tells us anything meaningful shows you're just trying to clown us.

Don't try stunts like that, if you don't want to look desperate.

 

[MacZ24]

I see a lot of complaints about CPU performance for the thing and light models and no mention of battery live. Most laptop users would give up performance for battery life. Personally, I would prefer a model with a '24 hour rated battery.

And I do not understand the need of castrated cores in a desktop CPU where power efficiency is not really an issue.

We have all our problems, it seems.

 

[bit_user]

And I do not understand the need of castrated cores in a desktop CPU where power efficiency is not really an issue.

Whether it's an issue depends on what you're doing. If it's something that's heavily multi-threaded, then you will be either power-limited or thermally-limited. In such cases, the CPU able to deliver more perf/W will end up delivering more perf, period.

For some of us, efficiency is either an issue due to expensive electricity, heat output, or the noise/cost/complexity of the requisite cooler. For me, the room where I work gets hot in the summer. I don't like to run the air conditioner, because that cools the entire house, which is expensive. My electricity isn't very cheap, either. And I don't like really noisy PCs, but prefer not to deal with water cooling.

 

[TheHerald]

Only when you cherry-pick the models and the power levels to make Intel look good.

AMD's 7900 and 8000G series show the potential of Zen 4 to be very efficient, indeed.

​

​

​

​

​

The 7700X has a TDP of 105 W, not 125 W. You can't go around making a big fuss about iso-power and then just disregard a discrepancy of 19%, Not only that, but the idea of taking two CPUs and running them both above TDP, as if such a comparison tells us anything meaningful shows you're just trying to clown us.

Don't try stunts like that, if you don't want to look desperate.

TDP of 105 results in 135w power draw. I was generous and dropped it by 10. Still it gets owned by the 13700 at the same wattage. Please, move along sir

Eg1. Just check TPU and indeed the 7700x is drawing 135w. You seem desperate.

 

[bit_user]

TDP of 105 results in 135w power draw. I was generous and dropped it by 10. Still it gets owned by the 13700 at the same wattage. Please, move along sir

The i7-13700's PL2 is 219W. So, you're being less than generous, when you cut it down to 125W. It's misleading for people who fall for your ruse and believe you're doing quite the opposite.

Eg1. Just check TPU and indeed the 7700x is drawing 135w. You seem desperate.

Quite projecting.

Speaking of the 7700X, that's just your usual M.O. of drawing sweeping conclusions from cherry-picked, rigged benchmarks.

 

[MacZ24]

In such cases, the CPU able to deliver more perf/W will end up delivering more perf, period.

I understand the argument of the price of electricity or heat output. But that can be resolved by lowering the SKU of CPU (taking something less powerful).

I don't agree with the rest of your argument. CPU cores are not infinitly powerful. An e-Core capacity will stop well before the p-Core capacity is reached. You are not limited by the power sent to the CPU but by what the CPU cores are capable of.

It doesn't really matter if your cores are extremely efficient if they can only run at 1 Mhz and achieve little (reductio ad absurdum)

 

[TheHerald]

The i7-13700's PL2 is 219W. So, you're being less than generous, when you cut it down to 125W. It's misleading for people who fall for your ruse and believe you're doing quite the opposite.


Quite projecting. And cherry-picking, as usual.

What's wrong with comparing both at 125w?

 

[bit_user]

What's wrong with comparing both at 125w?

It's not how normal people actually use these CPUs, so the conclusion isn't directly useful. Then, to draw a sweeping conclusion about Zen 4 vs. Raptor Lake from those data points is also misleading.

I think there are points you could make on this subject, but I have yet to see you make a well-defended point without sweeping overstatements. Its your tendency to do that, which undermines your credibility and support with many of us.

 

[TheHerald]

It's not how normal people actually use these CPUs, so the conclusion isn't directly useful. Then, to draw a sweeping conclusion about Zen 4 vs. Raptor Lake from those data points is also misleading.

I think there are points you could make on this subject, but I have yet to see you make a well-defended point without sweeping overstatements. Its your tendency to do that, which undermines your credibility and support with many of us.

It is how normal people that care about efficiency use their cpus. If they don't care about it sure, I agree with you, but why does it matter how people that don't care about efficiency use their cpus when the discussion is about efficiency? Those people are irrelevant here.

If a friend asks for a very efficient MT cpu at 300$ obviously I'd strongly strongly advice him to buy an Intel chip (and then cap it to whatever power he feels appropriate) than to buy the amd chip (and cap it to whatever power he seems appropriate) cause the intel chip will be literally (literally literally) 2 generations ahead in both perofmenace and efficiency. The Zen 5 ryzen 7 (which is also more expensive) hasn't caught up with a 2022 i7, and probably neither will zen 6. But hey, let's pretend amd is more efficiency, God knows why.

There is no world we're you'll end up buying an amd chip cause of its MT efficiency (excluding the 7950x), that's a laughable argument and you simply don't really care about efficiency if you do that. Which is perfectly fine, my point is stop pretending you care about it when you are buying chips that are insanely inefficient

It's not even about Intel vs amd, there are people who bought a 5950x over a 7950x because "it's more efficient" and I laughed at them too. The 7950x is leagues ahead of the 5950x. Right here you are basically arguing that the 5950x is more efficient than the 5950x which is a joke.

 

[bit_user]

I understand the argument of the price of electricity or heat output. But that can be resolved by lowering the SKU of CPU (taking something less powerful).

When one opts for a less power-hungry CPU, one typically wants to sacrifice as little performance as possible. This is where efficiency comes into play.

I don't agree with the rest of your argument. CPU cores are not infinitly powerful. An e-Core capacity will stop well before the p-Core capacity is reached. You are not limited by the power sent to the CPU but by what the CPU cores are capable of.

It doesn't really matter if your cores are extremely efficient if they can only run at 1 Mhz and achieve little (reductio ad absurdum)

They don't have to be capable of burning an infinite amount of power, but look at how much power a single core can actually burn!

​

Even at stock, you can't scale 8 P-cores @ something like 44W each! Even at 35W, 8 P-cores is 280W. If you replaced the i9's 4x e-core clusters with more P-cores, you'd be at 420 W, with less overall performance than Raptor Lake's 8 + 16 configuration.

Intel's hybrid strategy is also about delivering more compute performance per mm^2 of silicon, which means better perf/$.

This is old data, but it shows that Alder Lake's E-cores can perform about 65% as fast as one of its P-cores on integer workloads and about 54% as fast on floating point. Consider that they do this while using about 28% of the area and about 25% of the power as a P-core.

​

That's why Intel did it. So they could deliver more MT performance than if they used the equivalent number of P-cores within the same die area and ran them inside the same power budget.

 

[MacZ24]

That's why Intel did it. So they could deliver more MT performance than if they used the equivalent number of P-cores within the same die area and ran them inside the same power budget.

It makes no sense on a desktop.

Why then just no pack the die with e-cores ? Why bother with p-cores at all ? If you get more performance per area, that what you should do. Obviously, they don't.

Computers are really running either a < 5% load (more like 1% load), a single core 100% load or at 100% load. In the first case there is not need for an hybrid architecture; in the second case you would only need 1 p-core, and in the last case you would need only a ton of e-core following your logic.

There is no reason for having an hybrid achitecture on a desktop. Either all e-cores or all p-cores. Having an hybrid achitecture means really that you CPUs can't be fully utilised when you want to be fully utilised. If you have more time-sensitive threads than you have p-cores, your p-cores will act as e-cores because they will be waiting for them. Or you don't use the e-core at all in this case.

It makes sense on a laptop, because of energy use and battery usage. The reason this exists on a desktop is because they are not designing SKUs for desktop but for laptops, which is a much bigger market. The rest is mostly marketing.