Back in 2019/2020, Intel launched its Cascade Lake-AP processors to keep up with AMD in the high compute density segment. These were basically two Cascade Lake-SP dies “glued” together to double the core counts. However, a combination of inefficient thermals, power draw, and inflexible socket compatibility made sure that the lineup was rejected by the market. After Cascade Lake-AP, we got Cooper Lake-SP which was limited to 28-core offerings with a monolithic design.
While Intel has successively increased the core counts with Ice Lake-SP (to 28 cores) and Sapphire Rapids-SP (to 56 cores), the chipmaker still lags behind AMD in terms of sheer compute density. The latter offers 64 cores with Rome and Milan, with Genoa and Bergamo slated to increase it to 96 and 128 cores, respectively (courtesy of TSMC’s 5nm EUV process node).
According to Hungarian outlet ProHardver, Intel is planning to strike back in the cloud segment with its Sierra Forest-AP processors. Part of the Birch Stream platform and leveraging the massive LGA7529 socket, these chips will feature up to 128 efficiency cores, and go head-to-head against AMD’s Bergamo lineup.
Much like Bergamo, Sierra Forest-AP will leverage a heavily modified variant of the efficiency “Gracemont” core used in Alder Lake to keep the power draw in check. It’s important to note that the Gracemont core will be tweaked to suit the needs of the cloud segment, with support for additional instructions, including AVX512 and more. As such, the throughput of these cores will be a fair bit higher than the standard Gracemont cores.
Sierra Forest-AP is slated to land in 2023, sometime after Bergamo. It’s unclear which process it’ll leverage but if I had to guess I’d go with the Intel 4 node. This means that the two 128 core x86 cloud lineups will be evenly matched (mostly) in terms of both compute density and process technology.