Effects of Disabling SMT in Hardware Benchmarks

Disabling Simultaneous Multithreading (SMT) or Intel’s proprietary Hyper-Threading significantly alters computer hardware benchmark results depending on the specific workload being tested. This article provides a straightforward look at how turning off these technologies drastically reduces multi-threaded productivity scores, improves or stabilizes gaming and single-threaded metrics, and lowers thermal output to create more headroom for extreme overclocking benchmarks.

Multi-Threaded Workloads

SMT allows a single physical CPU core to process two software threads at the same time. In synthetic benchmarks designed to stress the entire processor—such as Cinebench, Blender, or V-Ray—disabling SMT results in a massive performance drop.

When you disable SMT, the operating system loses half of its logical processors. Consequently, multi-core benchmark scores typically fall by 15% to 30%. Without SMT, the CPU can no longer fill idle execution pipeline stages with secondary instructions, leading to lower overall computational throughput in heavy productivity tasks.

Single-Threaded and Gaming Benchmarks

In workloads that do not utilize many threads, such as legacy software, lightly-threaded applications, and many video games, disabling SMT can actually increase benchmark scores.

When two threads share a single physical core, they must compete for the same execution resources and CPU cache. Turning off SMT eliminates this resource contention. Each active software thread gets full, uninterrupted access to the physical core's L1, L2, and L3 cache. In gaming benchmarks, this often translates to slightly higher average frame rates and noticeably better 1% and 0.1% low frame times, as the system experiences less thread-scheduling latency.

Thermal and Power Consumption Metrics

Forcing two threads through a single physical core requires more power and generates more heat. When benchmarking a system for thermal efficiency or power draw, disabling SMT results in a substantial decrease in both metrics.

A CPU running a heavy stress test like Prime95 or OCCT with SMT disabled will typically run 5°C to 15°C cooler and consume significantly less wattage compared to the same CPU at the same clock speed with SMT enabled.

Overclocking Headroom

Because disabling SMT reduces heat output and voltage requirements, competitive overclockers frequently turn it off to achieve higher physical core clock speeds.

In benchmarks that prioritize raw frequency over thread count (such as SuperPi, Geekbench single-core, or specific 3DMark tests on high-core-count processors), disabling SMT stabilizes the processor. This stability allows the user to push the physical cores 100 MHz to 200 MHz higher than what would be possible with SMT enabled. In competitive benchmarking, this extra frequency often yields a higher score than having double the logical threads running at a lower speed.

Summary of Effects

  • Rendering and Encoding Benchmarks: Scores decrease by 15% to 30%.
  • Single-Core Benchmarks: Scores remain the same or increase slightly due to dedicated cache access.
  • Gaming Benchmarks: Frame times become more consistent, and average FPS may increase slightly in CPU-bound scenarios.
  • Thermal Benchmarks: Temperatures and power consumption drop significantly under full load.
  • Overclocking Benchmarks: Maximum achievable stable clock speeds increase.