Intel's Panther Lake CPU Generation on Track for Mid-2025 Release, AI Capabilities to See Significant Boost

[AleksandarK]

Intel's CEO, Pat Gelsinger, has confirmed that the upcoming 18A process of the Panther Lake CPU generation is on schedule for a mid-2025 release, which aligns with the initial projection. This development marks a significant milestone in the company's ongoing efforts to integrate AI capabilities into its processors. The mid-2025 release date is expected to follow the debut of Intel's Arrow Lake process in late 2024 or early 2025, a release that holds the promise of significant advancements in AI computing. During Intel's Q1 2024 Quarterly Results, Gelsinger expressed confidence in the company's AI capabilities, stating that the Core Ultra platform currently delivers leadership AI performance and that the next-generation platforms, Lunar Lake and Arrow Lake, will launch later this year, tripling AI performance. He also mentioned that the Panther Lake generation, set to release in 2025, will grow AI performance up to an additional 2x.

The Panther Lake generation represents the culmination of three generations of work in a short time and is expected to continue Intel's iterative approach. This transition is marked by a shift from a hybrid architecture, a combination of different types of processors, to a disaggregated die, where different components of the processor are separated, as AI computing becomes increasingly prominent. This strategic move is aimed at optimizing AI performance and flexibility. This marks the third generation of the Intel Core Ultra series, following Ultra 100 (Meteor Lake), Ultra 200 (Arrow Lake), and Lunar Lake (200V). Intel's release strategy mirrors the pattern set by the Hybrid Architecture, with Alder Lake debuting in 2021, followed by Raptor Lake in 2022, and a refreshed Raptor Lake released last year to bridge the gap until LGA 1851 was ready. However, Intel's roadmap has seen adjustments in the past, such as the initial promise of an Arrow Lake release before the end of 2024, which was later retracted. The mid-2025 release of Panther Lake aligns with rumors of Arrow Lake's late 2024 or early 2025 debut, suggesting that the 18A process CPU generation could debut several months after Arrow Lake.



View at TechPowerUp Main Site | Source

 

[Dristun]

Nothing exciting for the home enthusiast market, I gather...

 

[Assimilator]

If the only thing Pat can sing about is useless "AI" capabilities, Panther Lake is going to be another failure.

 

[Tomorrow]

Considering how much Intel is already downplaying their 20A i dont expect much from Arrow Lake and if im not mistaken then 18A based Lunar Lake is mobile only (like Meteor Lake).

The bigger problem is their new naming. Despite eliminating two zeroes from the name they managed to somehow make it more confusing.
Also supposedly they have bottleneck in somewhere in the manufacturing process that is holding back volume.

Someone else said it best - if they dont have enough volume for their own products then how are they supposed to fulfill orders for others as a foundry?

 

[sethmatrix7]

Considering how much Intel is already downplaying their 20A i dont expect much from Arrow Lake and if im not mistaken then 18A based Lunar Lake is mobile only (like Meteor Lake).

The bigger problem is their new naming. Despite eliminating two zeroes from the name they managed to somehow make it more confusing.
Also supposedly they have bottleneck in somewhere in the manufacturing process that is holding back volume.

Someone else said it best - if they dont have enough volume for their own products then how are they supposed to fulfill orders for others as a foundry?

Their foundry's are 5+ years from completion. Taking orders is years and years away.

 

[phints]

With TechPowerUp showing the 7800X3D being 4-5x the performance per watt as Intel 14th gen in gaming they have a *LONG* way to go. Intel 4 is MIA and 18A can't come soon enough.

 

[Denver]

will grow AI performance up to an additional 2x.

Now everything has to mention AI... let's talk about performance again, please.
Less AI, more IPC and perf/watt.

 

[Daven]

Intel is shifting resources to NPUs and GPUs. I don’t see much in the way of general core improvements for a long time.

 

[thesmokingman]

will grow AI performance up to an additional 2x.

Now everything has to mention AI... let's talk about performance again, please.
Less AI, more IPC and perf/watt.

When you have jack going for you, mention AI x100. Apparently you can't go wrong mentioning AI, lmao.

 

[AnotherReader]

Intel is shifting resources to NPUs and GPUs. I don’t see much in the way of general core improvements for a long time.

Intel is big enough to do more than just NPUs and GPUs. NPUs, in particular, are relatively simple compared even to GPUs, never mind CPUs.

 

[N/A]

Expecting impressive perfwatt with nano sheet ribbon fet and bs power delivery. 1,5 volts for 6.0 Ghz and instability should be a thing of the past. 1,2 volts and I'm buying. Possible drawback is tile to tile latency. But what can we do.

 

[mb194dc]

What "AI" capabilities on your local machine are of the remotest interest to pretty much any user?

The BS is palpable...

 

[Fouquin]

will grow AI performance up to an additional 2x.

Now everything has to mention AI... let's talk about performance again, please.
Less AI, more IPC and perf/watt.

It's all been about AI with Intel since 2018. TOPS/W is a big metric they've been pushing since the announcement of Ice Lake when they debuted GNA, VNNI, and DL Boost.

"'Ice Lake' is [Intel's] biggest step forward in systematically thinking about AI," Ran Senderovitz, Intel VP of mobile platform marketing. (2019)

I know it's hard to believe we've been hearing about AI for this long.

 

[Wirko]

What "AI" capabilities on your local machine are of the remotest interest to pretty much any user?

The BS is palpable...

Well maybe some clever minds will find ways to repurpose those dumb 8-bit vector-matrix multipliers for general computing tasks, such as zipping/unzipping, running javascript, virus scanning, code compilation, and the like. Win-win! We can hope, right?

 

[napata]

With TechPowerUp showing the 7800X3D being 4-5x the performance per watt as Intel 14th gen in gaming they have a *LONG* way to go. Intel 4 is MIA and 18A can't come soon enough.

Not sure what you're comparing but the 13400 is the 4th most efficient gaming CPU of the ones they tested. The big core count CPUs are never really going to be efficient at gaming. The 7950x sucks as well when it comes to efficiency.

 

[ncrs]

Intel is big enough to do more than just NPUs and GPUs. NPUs, in particular, are relatively simple compared even to GPUs, never mind CPUs.

I wouldn't call the Intel Meteor Lake NPU "relatively simple" . It's a pretty complex device containing multiple SPARC (yes, that SPARC) LEON cores and DSPs beside the MAC arrays.

 

[QUANTUMPHYSICS]

I'll wait till I get the 5090 before I build my next desktop.
Not sure if I'll have a 15900k or "16900k" but it'll be just in time for Grand Theft Auto 6.

 

[AnotherReader]

I wouldn't call the Intel Meteor Lake NPU "relatively simple" . It's a pretty complex device containing multiple SPARC (yes, that SPARC) LEON cores and DSPs beside the MAC arrays.

That was a good read; rather ironic to see SPARC in an Intel product. NPUs are simple compared to GPUs and especially CPUs. You also have to consider that a significant part of bringing CPUs to market is validation time and NPUs are far simpler in that way too.

 

[Darmok N Jalad]

Hey look, that AI you don't use will be twice as fast!

 

[hsgawdsaefryjt]

Intel is shifting resources to NPUs and GPUs. I don’t see much in the way of general core improvements for a long time.

This is entirely an assumption, there's nothing to back it up other than claims. Intel for decades has always been working on projects that may end up being a complete hit or a miss. Look at their employee count on top of how many years they've been making microprocessor architectures in general for on top of random projects, it's before AMD and NVIDIA were doing it.

With TechPowerUp showing the 7800X3D being 4-5x the performance per watt as Intel 14th gen in gaming they have a *LONG* way to go. Intel 4 is MIA and 18A can't come soon enough.

Please look at idle power consumption and power consumption in general from the 0-100% "usage" range across various instructions sets FULLY supported and graph it. Since 8-9th gen intel has been idling around 10 watts while AMD has and is doing 30 watts idle roughly, being for both CPU makers on the desktop end across the board, and for the most of it. That being said your cherry picked power benchmarks aren't that relative as nobody turns on their computer, games, and instantly turns it off, people leave their computers sitting for hours if not days on idle. There's more to efficiency outside of just the node being used, if you want to get complex maybe you should try IO looping with AMD CPU's and compare it to Intel CPU's.

 

[phanbuey]

I'll wait till I get the 5090 before I build my next desktop.
Not sure if I'll have a 15900k or "16900k" but it'll be just in time for Grand Theft Auto 6.

Same here -- CPU / platform to feed the GPU as it's already overkill for most tasks.

 

[kondamin]

Would be nice if independent media were to stop going along with the AI bullshit and just listed tops as it would say core voltage or something as a little detail only those looking for it would read.

 

[Onyx Turbine]

Is there no simple intel roadmap available that splits the release of mobile and desktop cpu release? Im only interested in the latter one.
When pairing it with a mid range 5000 series card i think we will see a time of fairly abundant gpu power when not going dual, ultra wide or triple screen and 4k resolution etc.

 

[Sabotaged_Enigma]

This lake, that lake... Intel processors try to consume power like draining a lake?
btw AI sh*t

 

[AnarchoPrimitiv]

The bigger problem is their new naming. Despite eliminating two zeroes from the name they managed to somehow make it more coconfusing

Glad I'm not the only one....whenever an article.mentions a new chip of theirs, I have no idea where is lies in the product stack