| name | slv-performance-boost |
| description | Update the kernel (Ubuntu HWE 6.14+), set `amd_pstate=active` (AMD), and apply cpufreq governor=performance, EPP=performance, cpuidle disable, and `scaling_max_freq=cpuinfo_max_freq` so the CPU stays at maximum boost. AMD nodes on kernel < 6.14 require HWE install + reboot. |
SLV CPU Performance Boost Skill
Keeps Solana validator / RPC nodes running at peak CPU frequency at all times. Runs alongside SMT disable and IRQ tuning as part of the slv v init / slv r init tuning phase.
What it applies
| # | Item | Effect | Reboot |
|---|
| 1 | apt-get update && apt-get upgrade -y | Refresh OS packages and userland | No |
| 2 | Install HWE kernel (linux-image-generic-hwe-24.04) | Enables amd_pstate 6.14+ on AMD | Yes (AMD only when kernel < 6.14) |
| 3 | GRUB amd_pstate=active processor.max_cstate=0 cpufreq.default_governor=performance | Initializes amd_pstate driver in active mode at boot | Yes (AMD only, first apply) |
| 4 | modprobe amd_pstate | Try to load the driver at runtime | No |
| 5 | governor=performance | Lock frequency at maximum | No |
| 6 | EPP=performance | Push the energy/performance preference toward performance | No |
| 7 | boost=enabled / Intel no_turbo=0 | Enable Turbo Boost | No |
| 8 | cpuidle state*/disable=1 | Disable C-states to avoid latency spikes | No |
| 9 | scaling_max_freq=cpuinfo_max_freq | Align scaling ceiling with the hardware ceiling | No |
| 10 | slv-cpu-boost.service (oneshot) | Re-applies items 5-9 on every boot | No |
Related artifacts
ansible/cmn/boost_performance.yml
- Idempotency marker:
/var/lib/slv/.boost_applied
- Persistence script:
/usr/local/bin/slv-cpu-boost.sh
- systemd unit:
/etc/systemd/system/slv-cpu-boost.service
Options
| Variable | Default | Description |
|---|
amd_pstate_mode | active | active (best performance) or passive |
Standalone run
ansible-playbook -i inventory.yml cmn/boost_performance.yml --limit <host>
ansible-playbook -i inventory.yml cmn/boost_performance.yml --limit <host> -e amd_pstate_mode=passive
Kernel update + reboot
When an AMD CPU is detected and uname -r is below 6.14, the playbook installs the HWE kernel, writes the amd_pstate=active GRUB cmdline, and the orchestrator (cmn/optimize_node.yml) detects need_reboot=true and reboots automatically.
Verification
The fastest way is slv check boost:
slv check boost -t validator -n mainnet -p <pubkey>
slv check boost -t rpc -n mainnet -p <pubkey>
This command is a Deno port of the master-api performanceCheckRouter and evaluates every check by category — [OS], [BIOS], [CONTROL_PLANE], [RUNTIME], [INFO] — returning OK / WARN / NG. Exit codes: OK=0, WARN=1, NG=2. When a BIOS blocker is detected, the report prints a friendly hint that lists the BIOS knobs needed to lift the firmware-side ceiling.
Manual verification:
uname -r
cat /sys/devices/system/cpu/amd_pstate/status
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
cat /sys/devices/system/cpu/cpu0/cpufreq/energy_performance_preference
cat /sys/devices/system/cpu/cpufreq/boost
grep 'cpu MHz' /proc/cpuinfo | sort -u
systemctl is-active slv-cpu-boost.service
When BIOS is the blocker
If slv check boost reports a [BIOS] blocker, the OS side is already correctly configured but cpuinfo_max_freq is below the catalog turbo frequency for the CPU model. That means the motherboard firmware / BIOS is capping the hardware ceiling. Typical fixes:
| BIOS setting | Recommended value |
|---|
| Determinism Slider | Performance |
| CPPC / Core Performance Boost | Enabled |
| Global C-States Control | Disabled |
| Power Profile | Maximum Performance / Top Performance |
| P-State / Cool'n'Quiet | Disabled (when amd_pstate=active) |
Rollback
sudo systemctl disable --now slv-cpu-boost.service
sudo rm /etc/systemd/system/slv-cpu-boost.service /usr/local/bin/slv-cpu-boost.sh
sudo sed -i 's/ amd_pstate=[^ ]*//; s/ processor.max_cstate=[^ ]*//; s/ cpufreq.default_governor=[^ ]*//' /etc/default/grub
sudo update-grub
sudo rm /var/lib/slv/.boost_applied
sudo reboot