Spectre: the tax on a trainable branch predictor

entry · p2026-09-06-180 · validated

previously 8-19% - updated w/ retpoline to be 4-8%

tech: learned-branch-prediction
passes the definition as: the same fit-at-runtime predictor, mistrained by an attacker so the CPU speculates down a chosen path and leaks memory through the cache
from: 2018-01-03
region: world
domain: safety
side: loss
people: 7,470,000,000 devices (2025)
effect: 19 percent (2018-01) — measured slowdown on OLTP and high kernel-to-user-transition workloads from the Meltdown/Spectre kernel mitigations (8-19%, improved to 4-8% with retpoline); every machine pays it permanently
reversible: no
who holds the switch: nobody
attribution: named
source kind: primary
the statistic: Red Hat measured the mitigation cost by workload class: OLTP / high kernel-to-user transitions 8-19% (4-8% after retpoline), database analytics and Java VMs 3-7% (2-5%), CPU-intensive/HPC 2-5% (1-2%), kernel-bypass under 1%; across the tested application set the range was 1-20% in January 2018 and 1-8% after retpoline. Article updated 2019-03-07. Reach as above (Ericsson, 2025) is a floor: the class of vulnerability affects essentially every speculating CPU.
source: https://access.redhat.com/articles/3307751
proposed by: researcher (opus), brief 2026-09-06, quiet-adoptions
received: 2026-09-06T09:39:39-04:00
harvest report: impacts/harvest/2026-09-06-quiet-adoptions.md

Rulings, in order

2026-09-06T09:39:47-04:00 · validatorpass — quote found on the fetched page (exact match; HTTP 200 text/html; charset=utf-8)
2026-09-06T09:59:36-04:00 · sessionseverity — a slowdown of a few percent that nobody felt: real, measured, imperceptible

The proposal as it arrived (JSON)

{
 "id": "p2026-09-06-180",
 "received": "2026-09-06T09:39:39-04:00",
 "proposer": "researcher (opus), brief 2026-09-06, quiet-adoptions",
 "kind": "entry",
 "sha": "399846d455dbcbcc",
 "data": {
  "kind": "entry",
  "tech": "learned-branch-prediction",
  "tech_name": "Spectre: the tax on a trainable branch predictor",
  "definition": "the same fit-at-runtime predictor, mistrained by an attacker so the CPU speculates down a chosen path and leaks memory through the cache",
  "from": "2018-01-03",
  "region": "world",
  "domain": "safety",
  "side": "loss",
  "people": {
   "n": 7470000000,
   "unit": "devices",
   "date": "2025"
  },
  "effect": {
   "n": 19,
   "unit": "percent",
   "date": "2018-01",
   "what": "measured slowdown on OLTP and high kernel-to-user-transition workloads from the Meltdown/Spectre kernel mitigations (8-19%, improved to 4-8% with retpoline); every machine pays it permanently"
  },
  "reversible": false,
  "switch": "nobody",
  "statistic": "Red Hat measured the mitigation cost by workload class: OLTP / high kernel-to-user transitions 8-19% (4-8% after retpoline), database analytics and Java VMs 3-7% (2-5%), CPU-intensive/HPC 2-5% (1-2%), kernel-bypass under 1%; across the tested application set the range was 1-20% in January 2018 and 1-8% after retpoline. Article updated 2019-03-07. Reach as above (Ericsson, 2025) is a floor: the class of vulnerability affects essentially every speculating CPU.",
  "source": "https://access.redhat.com/articles/3307751",
  "quote": "previously 8-19% - updated w/ retpoline to be 4-8%",
  "attribution": "named",
  "source_kind": "primary"
 },
 "from_file": "impacts/harvest/2026-09-06-quiet-adoptions.md"
}