The ARM Cortex-X2 is a CPU implementing the ARMv9-A 64-bit instruction set designed by ARM Holdings' Austin design centre as part of ARM's Cortex-X Custom (CXC) program.

It forms part of Arm's Total Compute Solutions 2021 (TCS21) along with Arm's Cortex-A710, Cortex-A510, Mali-G710 and CoreLink CI-700/NI-700.

Architecture changes in comparison with ARM Cortex-X1

The processor implements the following changes:

  • ARMv9.0
  • 10 cycle pipeline down from 11, created by reducing the dispatch stage from 2 cycles to 1
  • Reorder buffer (ROB) increased by 30% from 224 entries to 288
  • dTLB increased by 20% from 40 entries to 48
  • SVE2 SIMD support
  • Bfloat16 data type support
  • Support for Aarch32 removed
  • DSU-110 Up to 12 cores (up from 8 cores) Up to 16M L3 cache (up from 8 MB)
  • CoreLink CI-700/NI-700 Up to 32MB SLC

Performance claims:

  • Comparing the Cortex-X2 to the Cortex-X1 with the same process, clock speed, and 4MB of L3 cache (also known as ISO-process): 16% greater integer performance / IPC 100% greater ML performance
  • 30% peak performance improvement over the Cortex-X1 in smartphones

(3.3 GHz, 1MB L2, 8MB L3)

  • 40% faster than an Intel Core i5-1135G7 at 15W (3.5 GHz, 1MB L2, 16MB L3)

Architecture comparison

"Prime" core

uArchCortex-A78Cortex-X1Cortex-X2Cortex-X3Cortex-X4Cortex-X925
Code nameHerculesHeraMatterhorn-ELPMakalu-ELPHunter-ELPBlackhawk
ArchitectureARMv8.2ARMv9ARMv9.2
Peak clock speed~3.0 GHz~3.3 GHz~3.4 GHz~3.8 GHz
Decode width45610
Dispatch6/cycle8/cycle10/cycle
Max in-flight2x 1602x 2242x 2882x 3202x 3842x 768
L0 (Mops entries)1536307215360
L1-I + L1-D32+32 KiB64+64 KiB
L2128–512 KiB0.25–1 MiB0.5–2 MiB2–3 MiB
L30–8 MiB0–16 MiB0–32 MiB

Usage

See also