A tile designed for twelve large cores

According to Jaykihn’s posts on 15 November, Intel is considering a dedicated tile containing twelve Coyote Cove P-cores. The silicon would be designed for these large cores from the outset. The configuration would therefore rely neither on two tiles with six P-cores each nor on disabling the E-cores of a hybrid tile. Jaykihn’s post.

Jaykihn sur la configuration de bureau envisagée

X recevra des données de connexion. Votre choix autorise le chargement automatique des publications X sur ce navigateur pendant 180 jours. Modifiable dans « Gérer mes cookies ».

Voir la publication sur X

The notation “12 + 0 + 4” describes three groups across the processor. Only the compute section is said to consist exclusively of P-cores. The processor as a whole would still include four very low-power cores.

Reported group

Count

What the leak describes

P-cores

12

Large Coyote Cove cores on the same tile

Standard E-cores

0

None on the dedicated compute tile

LP E-cores

4

Very low-power cores counted separately

Jaykihn adds that, according to his information, this unlocked version would replace an earlier locked “8 + 0 + 4” variant. He associates it with a bLLC, or large-capacity last-level cache. Its size and latency, as well as gaming performance, remain unknown. Jaykihn’s clarification.

Intel Foundry already has an 18A-P process

Intel Foundry has presented 18A-P as an evolution of 18A with compatible design rules and improvements in performance and power consumption. It is Jaykihn who links this process to the twelve-core tile. He attributes other Razor Lake compute tiles to TSMC’s N2P. Intel has published no product specifications for this variant. Intel Foundry’s technical presentation.

The leak adds a theory about the silicon to earlier reports on the proposed desktop platform after Nova Lake. The socket and release date are separate questions from what goes inside the tiles. Our look at Razor Lake-S and LGA 1954.

Key takeaways

A dedicated tile will interest anyone who prioritises large cores. Assessing how it compares with hybrid models will require clock speeds, a power budget and tests on an identified processor.

All news