Intel releases new performance data for cascade lake

About a week ago, Intel provided quite a bit of information about its upcoming Xeon Cascade Lake processors. On Sunday, Intel announced new benchmarks for the performance of these new processors, with numbers from various HPC / AI applications in the real world.

Intel shows off the performance of Cascade Lake

Intel's Cascade Lake chip is a 48-core (2x24C) multi-chip package. Intel UPI merges creation, and there's optimized caching, a deep learning boost through VNNI, and security mitigations for Specter built into these chips.

We recommend reading our article on How to change screen resolution in Windows 10 and lower brightness

On Sunday, Intel revealed more about the performance of Cascade Lake by sharing the performance numbers of various real-world HPC / AI applications. In its attached blog post, Intel highlights the huge increase in deep learning performance compared to its own Xeon Platinum processors (up to 17X).

Then, in five real-world application tests, it compares the 48-core Cascade Lake in the two-socket configuration with a two-socket system based on AMD EPYC 7601. Since both systems run on two sockets, the Intel machine has a total of 96 cores at its disposal and the AMD offers 64 physical cores. On the real-world testing slide, you can see that the Intel system is ahead between 1.5x and 3.1x.

• Linpack: up to 1.21x compared to Intel Xeonalable 8180 processor and 3.4x against AMD EPYC 7601 Stream Triad: up to 1.83x compared to Intel Scalable 8180 processor and 1.3x against AMD EPYC 7601 Deep Learning Inference / IA: Up to 17 times more images per second compared to Intel Xeon Platinum processor.

Intel's Cascade Lake processors will be available during the first half of 2019. Meanwhile, AMD's 7nm Zen 2-based EPYC Rome processors with up to 64 cores are already in lab testing, and will be released in 2019. Also, there are rumors that Zen 2's IPC is up to 29% higher. higher than Zen.

Techpowerup font