Compute Units (40 CU)

Compute Units (40-CU)

The BC-250's RDNA 2 GPU physically has 40 Compute Units, but the driver boots with only 24 active. SkillFishOS can run all 40.

What it buys you

Configuration	vkpeak fp32 (GFLOPS)
tkg + governor, 24 CU	6,868
tkg + governor + 40 CU	11,329

≈ 1.8× the compute throughput. More CUs = more graphics/compute power, but also more heat and power draw.

Two ways it's enabled

1. Boot-time (kernel cmdline): amdgpu.bc250_cc_write_mode=3 makes active_cu_number go from 24 → 40 at GPU init. The Control Center flips this for you and reboots. It's a boot-time setting with no runtime toggle by itself.
2. Live, no reboot: SkillFishOS routes up to 40 CUs at runtime through umr, restored at boot by the skillfish-cu.service systemd unit. You can toggle and test CUs while the system is running.

Manage them from the Control Center

The SkillFishOS Control Center's Tuner section has a CU panel: a live grid of the 40 compute units, shown as 20 pairs — green (active), red (off), grey (driver minimum, always on).

• Click a pair to toggle it (CUs are managed 2 at a time = 1 WGP).
• Type any count from 24 to 40 (in pairs), or click cells directly.
• Test CU stresses a pair under vkpeak and checks for GPU errors/hangs.

Or from the CLI: skillfish-cu get / skillfish-cu <mask>.

Silicon lottery

Not every chip is stable with all 40 CUs. Test CU lets you enable each extra pair on its own and stress it, so you can find a bad CU instead of chasing random crashes.

See Benchmarks for what the extra CUs do in real workloads.