Home /Modules/cpu/TIMING.md

TIMING.md

lib/cpu/amd-threadripper/TIMING.md · 85 line(s) · markdown

AMD Ryzen Threadripper 9000 (Zen 5 x86-64) Timing

Clock model

The model treats the clock as an event source rated at about 5400 MHz max boost, not as "one instruction per clock." Per-SKU base clocks (e.g. ~3.2 GHz on the 9980X, 4.0 GHz on the 9970X, 4.2 GHz on the 9960X) are configurations of the same model via setSpecClock.

Timing records

Each cycle can record:

Instruction timing

Instruction duration depends on addressing mode, operand transfers, read/write sequence, branch result, and interrupt/wait state. Therefore:

instruction latency != clock period

and

instruction cycles * clock period = approximate elapsed instruction time

when no external wait state is inserted.

Circuit timing

Internal datapath transfers are clock-sensitive. SLeeLa records this as observable/derived timing and does not claim an exact transistor simulation.

Sources

Expert timing, voltages, and variable clocks (shared across all CPU models)

The per-CPU timing above is the architectural/behavioral view for this part. The adjustable, quality-first timing system is shared by every CPU model in lib/cpu and is documented once, in the package guide:

Every CPU model (this one included) inherits the clock/overclock surface from SLCPURuntime and can be assembled with those shared expert-timing, voltage, and component options; they are a build-layer concern, not re-declared per CPU.

Threadripper 9000 ancillary presets (board, RAM timings, voltages)

The shared layer is wired to this family's actual operating point in SLAMDThreadripperPlatform. It hands back ready-configured objects so you can build a rig at spec without re-entering raw numbers:

A complete assembly (CPU + board + DDR5-6400 + rails, HEDT and PRO WX) is shown in components/examples/threadripper-at-spec.sleela.