Arm’s Cortex X2-based CPUs are 30 percent faster and more efficient

• 

  Speeds and feeds for the three new CPU designs. [credit: Arm ]

Arm has announced its next-generation CPU designs, which will probably hit the market in early 2022. With all-new cores, a new architecture, and the death of 32 bit, they are proving to be among Arm's biggest releases in some time. I don't think ARM published a first-party post with this information, but you can't ask for a better source than the pages and pages of info at Anandtech.

Arm CPU designs include cores in three sizes: "little," "big," and "performance," all of which tackle different workloads at various power-consumption levels. The follow-up to this year's "performance" X1 core is the Cortex X2, the big core is the Cortex A710, and for the first time in four years, ARM is introducing a new "little" core for high-efficiency workloads, the Cortex A510. Devices will finally be rid of the smaller Cortex A55 cores, which were introduced in 2017. Naturally, everything is faster. Compared to current-generation X1-based chips, ARM is promising a 16 percent faster X2 core, a 10 percent faster and 30 percent more efficient "big" core, and a 35 percent faster and 20 percent more efficient "little" core. Every core design is also seeing a 2-3x improvement in machine-learning performance, for whatever that's worth. Put all those digits together and ARM says a CPU cluster with the normal phone layout (one Cortex X2, three A710s, and four A510s) should have 30 percent better peak performance and 30 percent better-sustained efficacy.

While the Cortex X1 and A710 are both based on the previous designs that go back to 2018 Cortex A76 cores, the smaller Cortex A510 is an entirely new design. ARM says this small core is within striking distance of the Cortex A73, which was the "big" CPU core in 2017 flagship smartphones. All three CPU designs use the new Armv9 architecture, which includes a number of security enhancements like the containerized "confidential compute architecture."

Read 3 remaining paragraphs | Comments

https://ift.tt/3flFMyH