TECH Signal 243 2 feeds carried it
Microsoft Prism binary translation roughly doubles instruction counts and imposes harsh performance penalties on aarch64 PCs
↗
A Chipsandcheese analysis of Windows 11's Prism binary translator finds that running x86-64 code on aarch64 hardware approximately doubles executed instruction counts and produces significant performance penalties across every tested core, including Qualcomm's Snapdragon X2 Elite.
Arm's viability in the PC market depends on how transparently x86-64 software runs on aarch64 hardware, and these results quantify the cost of that translation layer. Engineers evaluating Windows on Arm devices for x86-64 workloads should expect substantial overhead, not parity with native execution. The finding that Prism is optimized for Qualcomm Snapdragon yet shows similar instruction counts on non-Qualcomm cores suggests the optimization benefits are limited in scope.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Geekbench 7 workloads running under Prism execute roughly twice as many aarch64 instructions as native x86-64 instructions, except in the PDF Viewer workload.
Every tested core including Qualcomm's Snapdragon X2 Elite Extreme suffers harsh performance penalties from binary translation.
Prism is tuned specifically for Qualcomm Snapdragon processors but instruction counts remain similar across different aarch64 cores, suggesting the optimization does not reduce translation overhead at the instruction level.
THE CLUSTER