Intel Panther Lake to feature up 16 CPU cores and 12 “Celestial” Xe3-Cores

Intel Panther Lake with a powerful next-gen GPU

Rumors suggest it will feature up to 12 Xe3-Cores. 

Over on Twitter, Prakhar Verma made a series of tweets regarding Intel’s next-generation client CPU series, codenamed “Panther Lake.” This person is not known for any previous leaks, but from what we’ve gathered, the recently posted CPU configurations of these Panther Lake CPUs are indeed correct.

Prakhar made several claims regarding the Panther Lake CPU architecture. First and foremost, it is not a desktop architecture, but rather another mobile variant. It would appear as Panther Lake-H (PTL-H) and Panther Lake-U (PTL-U), with the latter being a low-power version succeeding the soon-to-be-released Lunar Lake.

According to the information shared, Panther Lake is 35% more efficient than Lunar Lake. It’s worth noting that Lunar Lake is already claimed to deliver a 50% increase in multi-core performance over Meteor Lake-U. Furthermore, these CPU series are said to feature Xe3-Cores, based on the Celestial architecture. In fact, this is the first known product to feature this GPU architecture.

The rumor also alleges that Xe3 is 40 faster than Xe2 and the GPU alone provides 120 TOPS for AI acceleration and the whole chip is 170 TOPS. The entire chip is said to feature 4 Performances, 8 Efficient and 4 Low-Power Efficient cores. Earlier leaks suggested it is a Cougar Cove and Skymont based architecture.

As discovered by InstLaX64 via Wccftech Intel is seemingly confirming, there are two versions of Panther Lake GPU: the GT2 and GT3. Usually, the higher the number, the more GPU cores there typically are. This would match the rumor shared by Prakhar.

Intel Panther Lake GPU variants, Source: Coreboot/Wccftech

Intel Panther Lake is expected to launch no sooner than 2025. This product would directly succeed Arrow Lake-H in performance mobile tier and Lunar Lake in low-power models. The current rumored lineup of next-gen Intel CPU series might look like this:

Source: Prakhar Verma