Intel Lunar Lake Technical Deep Dive

[Daven]

So the E-core now has about the same IPC as the P-core. They have been able to duplicate the same IPC between AMDs regular and compact cores.

 

[SL2]

AMD isnt offering different SKUs for different groups of notebooks, its all single CPU which can go into ultrathin to full fat desktop replacements so for them killing support for user replaceble memory is misstep.

They do. Stop spreading FUD.

Fire range is bascally a mobile variant of Granite ridge, just like Dragon range is a mobile Raphael. Both have basic graphics and are intended for "full fat desktop replacements"

While I do appreciate tech deep dives like this, I do have to question the choice to make Lunar Lake the subject, given that the audience of this site is overwhelmingly users who care about desktop, not ultraportable.

I thought that was obvious, there's not much Arrow lake info for now, so what you're suggesting isn't possible at the moment.

 

[Denver]

I'm sorry, but it stinks... they haven't presented a single benchmark even showing the real performance of the iGPU or CPU. Everything shown is theoretical performance in the best possible scenario.

 

[Daven]

Last sentence in the conclusion:

“If you want to see Lion Cove, Skymont, Xe2 Battlemage, and NPU 4 in a more familiar package, you should look out for Arrow Lake, which not just covers other mobile form-factors, but also desktop.”

So where is Arrow Lake? Did Intel make one mention of it?

 

[londiste]

I'm sorry, but it stinks... they haven't presented a single benchmark even showing the real performance of the iGPU or CPU. Everything shown is theoretical performance in the best possible scenario.

SPEC, CineBench, Geekbench, WebXPRT?

 

[SL2]

So, is WIFI 5 faster than 6? I'm not up to date..

 

[Denver]

SPEC, CineBench, Geekbench, WebXPRT?

Show me the numbers, I haven't seen them.

 

[phints]

Hoping to see a new Intel HEDT CPU this year. An actual new one on a new lithography. It's been ages.

 

[Philaphlous]

I'm sorry, but it stinks... they haven't presented a single benchmark even showing the real performance of the iGPU or CPU. Everything shown is theoretical performance in the best possible scenario.

Throw some money at Intel... a substantial amount and they might give you some benchmarks....
Nothing is going to blow peoples minds away with benchmarks ATM. Intel innovation has gone down the tubes and at best you'll get modest improvements over prior gen. Apple on the other hand, while they've matured out the M series chips, they're lightyears ahead of intel and even AMD... passive cooling and amazing perf/watt that neither AMD nor Intel can touch is really important to take note, especially in the mobile segment.

 

[Daven]

Throw some money at Intel... a substantial amount and they might give you some benchmarks....
Nothing is going to blow peoples minds away with benchmarks ATM. Intel innovation has gone down the tubes and at best you'll get modest improvements over prior gen. Apple on the other hand, while they've matured out the M series chips, they're lightyears ahead of intel and even AMD... passive cooling and amazing perf/watt that neither AMD nor Intel can touch is really important to take note, especially in the mobile segment.

I think AMD is getting closer:

9700x 65W 5.5 Ghz 4 nm vs 7700x 105W 5.4 Ghz 5 nm

Apple is on 3 nm so they still have a process node advantage. If Strix Halo is using 5, 4 and 3 nm then it might equal the M3 Max.

 

[Wirko]

So the E-core now has about the same IPC as the P-core. They have been able to duplicate the same IPC between AMDs regular and compact cores.

Now AMD only needs to shrink their 4c core in half (we don't know enough about 5c yet) and they will be on par.

Don't forget, AMD's Zen 4c cores are 2/3 the size of Zen 4 cores. Intel's E cores are 1/3 the size of P cores. That's my rough measurement from the slides, but I wish Intel were less mysterious about the physical sizes of the cores.

@W1zzard Did Intel reveal anything about HT in Arrow Lake? I find it unlikely they're planning to abandon it across their entire range of products, all the way up to workstation and server chips.

 

[Adrian Andrzejewski]

There seems to be a typo in the text. It should say "Pentium 200 MMX" instead of "Core Ultra 200 MX" ;-)

 

[Daven]

Now AMD only needs to shrink their 4c core in half (we don't know enough about 5c yet) and they will be on par.

Don't forget, AMD's Zen 4c cores are 2/3 the size of Zen 4 cores. Intel's E cores are 1/3 the size of P cores. That's my rough measurement from the slides, but I wish Intel were less mysterious about the physical sizes of the cores.

@W1zzard Did Intel reveal anything about HT in Arrow Lake? I find it unlikely they're planning to abandon it across their entire range of products, all the way up to workstation and server chips.

The Skymont E-core versus Zen 5c comparison will be interesting but at the end of the day its all about the power. Chip space on package doesn't affect purchasing decisions but less size sometimes means less TDP.

But we do have some preliminary information to look at. For instance the Epyc 9754 which uses 128 Zen 4c cores with hyperthreading has a TDP of 360W. The new Xeon 6700E which uses 144 Skymont E-cores with no hyperthreading has a TDP of 330W.

Intel Xeon 6700E "Sierra Forest" CPUs Launched: Up To 144 E-Cores, 330W TDP, 34% More Efficient Versus AMD EPYC Bergamo (wccftech.com)

I'm not seeing a big difference here in TDP and potential performance between these two chips and the Epyc 9754 has been on the market for a year.

 

[SL2]

Don't forget, AMD's Zen 4c cores are 2/3 the size of Zen 4 cores. Intel's E cores are 1/3 the size of P cores. That's my rough measurement from the slides, but I wish Intel were less mysterious about the physical sizes of the cores.

Here are some numbers for Meteor.

 

[G777]

While I do appreciate tech deep dives like this, I do have to question the choice to make Lunar Lake the subject, given that the audience of this site is overwhelmingly users who care about desktop, not ultraportable. I also have to question the wisdom of using the numbers from Intel's marketing slides, as every company lies, but Intel lies more.

It's Techpowerup, not DIYPCpowerup.

Also, Lunar Lake is relevant for gaming handhelds, which I assume would garner some interest with the gaming-oriented audience. Plus, the P and E core architectures will be used in Arrow Lake.

 

[tfp]

Both tiles are made by TSMC and nothing by Intel Foundry apart from packaging with Foveros? Looks like it's going to be expensive.

How is this any more expensive then any other CPU designer that doesn't own their own FABs? If anything it should be the same price or cheaper with Intel doing it's own packaging. While not ideal neither is Intel's current node or that nodes capacity.

 

[CosmicWanderer]

I doubt this will be able to match the battery life that devices with the Apple M3 and Snapdragon X can achieve.

Time will tell. Looking forward to the benchmarks.

 

[ncrs]

How is this any more expensive then any other CPU designer that doesn't own their own FABs? If anything it should be the same price or cheaper with Intel doing it's own packaging. While not ideal neither is Intel's current node or that nodes capacity.

That's the point - it moves them to the "fabless" category, so they can't leverage in-house manufacturing to compete on price. They can't make their silicon at cost - they have to pay TSMC a premium for it. TSMC's total capacity is limited. The more clients it has competing for it the more expensive everything becomes to end consumers.

 

[JustBenching]

Throw some money at Intel... a substantial amount and they might give you some benchmarks....
Nothing is going to blow peoples minds away with benchmarks ATM. Intel innovation has gone down the tubes and at best you'll get modest improvements over prior gen. Apple on the other hand, while they've matured out the M series chips, they're lightyears ahead of intel and even AMD... passive cooling and amazing perf/watt that neither AMD nor Intel can touch is really important to take note, especially in the mobile segment.

Are you basing this of off Geekbench?

 

[HR_The_Butcher]

While I do appreciate tech deep dives like this, I do have to question the choice to make Lunar Lake the subject, given that the audience of this site is overwhelmingly users who care about desktop, not ultraportable.

Perhaps I'm in the minority, but I'm on the lookout for a fanless ultraportable so I was quite happy when I discovered the article.
Snapdragon X Plus looks promising, but I haven't seen any fanless models yet.

 

[Neo_Morpheus]

Fool me once and stuff….

I simply do not believe one word coming out from intel marketing team, given that they always lie, so until i see proper reviews from unbiased reviewers, they can shove it.

 

[AnotherReader]

Unless they double or quadruple the caches & make the clocks much higher no way Skymont comes close to RPL. Those IPC claims sound more than just dubious!

That claim for equivalent IPC is probably for low clocks. Skymont is unlikely to clock as high as Raptor Cove. Given that they are using TSMC's N3 process for Lunar Lake, the claim of equivalent IPC is plausible.

 

[kondamin]

While I do appreciate tech deep dives like this, I do have to question the choice to make Lunar Lake the subject, given that the audience of this site is overwhelmingly users who care about desktop, not ultraportable. I also have to question the wisdom of using the numbers from Intel's marketing slides, as every company lies, but Intel lies more.

I'd love a low power full x86 mini desktop like those found with the n100

So, is WIFI 5 faster than 6? I'm not up to date..
View attachment 350033

wifi6 ax is the faster standard, something probably went wrong when generating the graph.
if you are looking to upgrade your wireless infrastructure it's OK to skip it

 

[tfp]

That's the point - it moves them to the "fabless" category, so they can't leverage in-house manufacturing to compete on price. They can't make their silicon at cost - they have to pay TSMC a premium for it. TSMC's total capacity is limited. The more clients it has competing for it the more expensive everything becomes to end consumers.

So they are no different then AMD or any of their competitor for this processor? AMD is highly profitable now, creating the full chip at TSMC only is a problem for Intel if they can't sell the chips at the right price and volume. That's not to say it wouldn't be better for Intel to have better nodes and more capacity but it's not that critical as a stop gap. It might not be critical in the long run as some people believe intel should spin off it's Fabs, which I'm not a big fan of them doing but we'll see.

As for the consumers, none of these companies are going to lower prices for our well being. As long as there is demand they will charge what the consumer can bare just like TSMC is doing to Intel, AMD, Apple, Nvidia, ect.

 

[atomsymbol]

So the E-core now has about the same IPC as the P-core. They have been able to duplicate the same IPC between AMDs regular and compact cores.

Skymont E-cores having the same IPC as Lunar Lake P-cores is most likely a temporary phenomenon. When Intel P-cores adopt the same front-end architecture as Zen5 and Skymont, P-cores will once again be outperforming E-cores in IPC by a significant margin. Note: Zen5 and Skymont actually have similar front-ends on a conceptual level, the only differences being (1) the lack of µop cache in Skymont compared to Zen5 and (2) Zen5 can fetch up to 2 basic blocks while Skymont can fetch up to 3 basic blocks (see https://en.wikipedia.org/wiki/Basic_block).