Intel is reinforced in the multicore race: its new Cascade Lake (still in 14 nm) has 48 cores

The semiconductor giant has several fronts open with its competitors, and server processors are one of them. Now Intel has renewed its catalog with its new Xeon Cascade Lake Advanced Performance (AP) and Xeon E-2100 processors, which move to a multi-chip approach.

Both are the answer to the AMD EPYC processors focused on data centers, and in these new products Intel goes a step beyond what had gone with their Xeon SP. In those monolithic cores it had 28 cores: now it reaches the 48 cores that it accompanies with 12 DDR4 memory channels for each of the two supported sockets.

Doubts about HyperThreading

This new multi-chip design helps not to increase the complexity with increasingly larger chips. In this new approach, yes, Intel has not clarified if HyperThreading technology is supported to avoid certain risks that this type of option seems to generate in the field of security in some specific scenarios.

The processors will incorporate hardware changes to mitigate some of the variants of Specter and Meltdown , something that Intel has already done in other microprocessor families in recent months.

In Intel they say that these new mics offer 20% more performance than the Xeon SP and no less than 240% more than the AMD EPYC. There are also new features in the instruction set, which includes new AVX512 operations useful in calculations related to artificial intelligence processes: there is a 17% improvement in image recognition algorithms, for example.

The Xeon E-2100 are rather more modest , however, and it is about microphones with 6 cores and 12 threads that are not multichip designs and that are basically a special version of the Core for end users that do admit memory ECC. In both cases Intel makes use of a manufacturing process of 14 nm: in processors for servers it will also be necessary to wait to see 10 nm mics.