MediaTek benchmark cheating reopens old wounds

JC Torres - Apr 8, 2020, 10:37 pm CDT
0
MediaTek benchmark cheating reopens old wounds

A few years ago, a scandal broke out in the smartphone market on how once trustworthy benchmarking tools have become suddenly unreliable because smartphone manufacturers were apparently cheating their way to high numbers. That resulted in massive changes to those tools and even a few lawsuits. Other than that, though, it seems that nothing has really changed because when MediaTek was reported to be doing this kind of cheating, it gave the same response Huawei did two years ago: everyone’s doing it anyway.

To be fair, not everyone agrees on what should be benchmarked. Some believe tools should recreate and report performance that is as close to real-world usage while others, including MediaTek, believe the numbers should reflect the maximum throughput a phone is capable of. To do that, however, chip and phone makers have to pull out all the stops, effectively cheating the benchmark test.

AnandTech discovered this when a Helio P95 with two-year-old Cortex-A75 cores solidly beat the Dimensity 1000L’s Corex-A77 cores. Long story short, it learned that MediaTek’s silicon identifies certain benchmarking tools and then enables “sports mode” for maximum performance, at the expense of battery life, of course. This is pretty much the same scenario from years ago and is exactly one of MediaTek’s arguments.

MediaTek’s response basically boils down to how it is an industry-wide practice anyway and that phone makers can disable this sports mode if they want to. It even makes a reference to a competitor, which can only be Qualcomm, that uses those same tactics.

These statements, if taken at face value, reveals that nothing has really changed in the industry ever since those benchmarking scandals broke out. MediaTek is also ironically casting doubt on the entire system, which would mean that any claim it makes about its chips’ performance would also be held suspect.


Must Read Bits & Bytes