Glossary/CPU/SMT / Hyper-Threading
SMT / Hyper-Threading
SMT lets one physical core run two instruction streams at once, using execution units that would otherwise sit idle while a thread waits on memory.
Also known as simultaneous multithreading, hyper-threading, hyperthreading
Not measurable from a browser
A browser cannot read temperatures, power draw, fan speed, VRAM capacity, clock frequency or native FLOPS. For a direct reading use a native monitor such as HWiNFO64, GPU-Z or lm-sensors. SystemCheck contributes the other half: whether the condition is costing you throughput, measured as decay and consistency.
Simultaneous multithreading is Intel's Hyper-Threading and AMD's SMT under different names. The core duplicates the architectural state needed to track two threads but shares the expensive parts: execution ports, caches, branch predictors. When one thread stalls on a cache miss, the other keeps the pipeline fed.
The payoff depends entirely on the workload. Compilation, rendering, and video encoding gain meaningfully. Latency-sensitive work and workloads whose working set already fills L2 can lose performance, because two threads now contend for the same cache. This is why some competitive gamers still disable SMT and why some server workloads are certified SMT-off.
SMT also changes what a per-thread number means. Throughput per logical processor drops when SMT is on, even though total throughput rises. If you compare a SystemCheck multi-core score before and after toggling SMT in BIOS, expect total throughput to move modestly while any per-thread interpretation shifts a lot. Always compare totals, not per-thread derivations.
Related terms