The Energy Cost of Clock Distribution: Practice, a Thermal-Noise Floor, and the Gap Between Them

The clock network of a synchronous processor carries no information other than timing, yet reported figures attribute between one fifth and one half of total chip power to it. For the Intel Core i9-14900K at 253 W maximum turbo power and 5.6 GHz, the central literature fraction of 0.30 assigns 75.9 W to clocking, or 13.55 nJ per cycle. A thermal-noise floor is derived for distributing a timing edge at a stated rms timing tolerance: at 5 ps and 350 K the floor is 1.21e-19 J per receiving node per cycle (25 kT), so practice sits about 5.6e3 (3.7 orders of magnitude) above the floor per node. A wire-capacitance H-tree model shows the global distribution tree is a minority of clock energy; most of the cost is local. Clock gating and multiple clock domains recover substantial but bounded, overlapping shares. Model calculations are SIMULATED, NOT MEASURED. The computation script and its full results are included and produce every number in the paper.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-10
DOI
https://doi.org/10.5281/zenodo.24102331
Primary Topic
Low-power high-performance VLSI design
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

The Energy Cost of Clock Distribution: Practice, a Thermal-Noise Floor, and the Gap Between Them

William Alexander Aldred
Zenodo (CERN European Organization for Nuclear Research)
Low-power high-performance VLSI design
preprint

The Energy Cost of Clock Distribution: Practice, a Thermal-Noise Floor, and the Gap Between Them

William Alexander Aldred
preprint en

Abstract

The clock network of a synchronous processor carries no information other than timing, yet reported figures attribute between one fifth and one half of total chip power to it. For the Intel Core i9-14900K at 253 W maximum turbo power and 5.6 GHz, the central literature fraction of 0.30 assigns 75.9 W to clocking, or 13.55 nJ per cycle. A thermal-noise floor is derived for distributing a timing edge at a stated rms timing tolerance: at 5 ps and 350 K the floor is 1.21e-19 J per receiving node per cycle (25 kT), so practice sits about 5.6e3 (3.7 orders of magnitude) above the floor per node. A wire-capacitance H-tree model shows the global distribution tree is a minority of clock energy; most of the cost is local. Clock gating and multiple clock domains recover substantial but bounded, overlapping shares. Model calculations are SIMULATED, NOT MEASURED. The computation script and its full results are included and produce every number in the paper.

Zenodo (CERN European Organization for Nuclear Research)
Low-power high-performance VLSI design
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon