Guides

A card is slower than its neighbors

Identical devices in one box should do identical work. When one does not, the reason is usually written down somewhere siltide already reads.

1. See the outlierThe Overview marks a device that trails its siblings. The Thermals and Power tabs show the same fleet by temperature and by draw.
2. Read the throttle reasonsOpen the device. throttle names why the clocks are down: power cap, thermal, hardware slowdown, or an application clock someone set.
3. Compare against the limitstemp_slowdown and temp_shutdown say where the card gives up, and power_cap what it is allowed to draw. A card at its cap is working as configured, not failing.
4. Look backHistory with m on temperature shows whether this started with the current job or has been true since the machine came up.
5. Check the fabricThe Links tab: a device with a link down or a PCIe width below its maximum is slower for a reason that has nothing to do with heat.

Every step here is a tab, a key or a flag siltide already has. Install it, or explore the same views with siltide --demo.