AMD has fixed a scaling bug in HandBrake that was leaving high-core-count Threadripper processors underused during video transcoding. The company found cases where adding more CPU resources actually hurt performance, with some workloads dropping by up to 60%. The fix is now merged upstream and ships in HandBrake 1.11.0 and later.

Threadripper and Threadripper PRO chips exist for workloads like video encoding, where dozens or nearly a hundred cores can chew through a render queue. But raw core count only helps if the software can actually keep those cores fed. AMD's engineers traced the problem to how HandBrake handled thread management and job scheduling once systems went past 64 logical processors, where coordination overhead started eating into the CPU time that should have gone to encoding.

What AMD changed

AMD reworked how HandBrake splits transcode jobs across cores, so the software spends less time coordinating small tasks and more time actually encoding. The fix required no changes on the user side. Anyone running HandBrake 1.11.0 or newer gets the improvement automatically, with no new settings or presets to configure.

AMD tested the change on two systems: a Threadripper PRO 9995WX (96 cores) and a Threadripper 7980X (64 cores), both paired with a Radeon RX 9070 XT GPU. To isolate the threading fix from unrelated encoder quality changes, AMD compared HandBrake CLI 1.11.1 against the older 1.6.1 baseline rather than a more recent build.

The numbers

On the 9995WX, gains reached as high as 181% (2.81x), with the biggest jumps on H.264 720p (+181%), HEVC 10-bit 2160p (+151%) and 8K HEVC workloads (+149%). On the 7980X, gains topped out at 215% (3.15x), led by H.264 720p (+215%) and 8K HEVC 8-bit (+203%). Improvements on the 64-core chip ranged from 16% to 215% depending on the specific test clip and codec, which spanned H.264 and HEVC content from 720p up to 8K in both 8-bit and 10-bit color.

Notably, the smaller, lower-resolution clips saw some of the largest relative gains. That tracks with AMD's diagnosis: scheduling overhead hurts proportionally more on quick, small jobs than on long 8K renders where there is more actual encoding work to spread across cores.

Why it matters beyond the benchmark chart

AMD contributed the fix directly to the HandBrake project rather than shipping a special build, so any user gets the benefit through a standard update. It's a reminder that CPU core counts alone don't guarantee performance, software has to be tuned to use them, and that tuning is often invisible until someone goes looking for it. For creators and editors running Threadripper workstations, the fix means real time saved on transcode-heavy workflows, just by updating HandBrake to 1.11.0 or later.