AMD CPUInfo: Complete Overview of Features & Specs

AMD CPUInfo Explained: Benchmark Tips and Optimization Tricks

Model identifier: family, model, stepping (e.g., Zen 3, Zen 4 identifiers).
Core and thread counts: physical cores vs. logical threads (SMT).
Base and boost clocks: nominal base frequency and single/multi-core boost ranges.
Cache sizes: L1/L2/L3 per-core or shared.
TDP and package power limits: rated thermal design power and configurable limits.
Microcode/firmware version: affects mitigations and stability.
Supported instruction sets: e.g., SSE, AVX, AVX2, AVX512 (if present), SHA, SVE for relevant architectures.
Integrated GPU details: if APU, iGPU family and execution units.
Platform features: PCIe version, memory channels and ECC support.

Use native tools:
- Windows: CPU-Z, HWiNFO, or PowerShell’s Get-CimInstance win32_processor.
- Linux: lscpu, /proc/cpuinfo, dmidecode, and sudo dmidecode for SMBIOS data.
Check firmware/BIOS: ensure latest AGESA/motherboard firmware for correct reporting.
Cross-check: compare OS-reported values with manufacturer specs on AMD’s product pages.

Isolate variables: close background apps, set power plan to high performance (or manual P-states on Linux), and disable unnecessary services.
Thermal control: ensure adequate cooling and consistent ambient temperature; thermal throttling skews results.
Power limits: note PBO, PPT, TDC, EDC settings—stock vs. autotuned values change sustained performance.
Use multiple benchmarks: single-thread (e.g., Cinebench single-core, Geekbench single), multi-thread (Cinebench R23 multi), and real-world tests (compilation, rendering, encoding).
Repeat runs: run each test 3–5 times and report median or mean plus variance.
Record telemetry: log core clocks, temperatures, package power, and per-core utilization during runs (HWiNFO on Windows, perf + turbostat on Linux).

BIOS/UEFI tuning: update AGESA, enable XMP/DOCP for proper memory speed, set correct memory timings and voltages.
PBO and Curve Optimizer: for Ryzen, use Precision Boost Overdrive and Curve Optimizer to lower per-core voltage for higher sustained boost—test stability with stress and real workloads.
Undervolting: reduce voltage to lower temps and power draw while maintaining boost behavior—prefer per-core adjustments if available.
Memory tuning: faster DDR4/DDR5 and lower CAS improve frame times and some compute tasks; tighten secondary timings cautiously.
Cooling and case airflow: better cooling allows higher sustained clocks; prioritize low delta-T between ambient and CPU.
OS-level: keep chipset drivers updated, enable large page support for certain workloads, and for Linux tune governor (performance or schedutil with tuned profile).
Background task management: pin critical threads to higher-performance cores or adjust process priority for latency-sensitive tasks.

Look beyond scores: correlate clocks, temps, and power to understand whether a run hit thermal or power limits.
Normalized comparisons: compare systems with similar cooling, power limits, and memory configs.
Real-world relevance: prioritize benchmarks that mirror your workload (gaming, content creation, server tasks).

List CPU model, BIOS/AGESA, OS and kernel version, cooling, memory spec, motherboard, power settings