Samsung accused of cheating benchmarks on Galaxy S4, claims innocence

31 Jul 2013

Samsung has been accused of engineering the benchmark performance of its flagship Galaxy S4 smartphone to appear faster than it would perform in average user conditions, but the South Korean phonemaker claims it has not been involved in any wrongdoing.

Certain Samsung Galaxy S4 models (such as that which is available in Ireland) feature Samsung’s first-generation Exynos 5 Octa system-on-a-chip (SoC), which integrates four ARM Cortex A15 cores clocked at 1.6GHz and four ARM Cortex A7 cores at 1.2GHz. The GPU is a PowerVR SGX 544MP3 which reaches speeds up to 533MHz.

Claims that Samsung only opened up the 533MHz GPU to certain benchmarks while limiting general usage to a lower 480MHz first came from posts on the Beyond3D Forum and a Twitter user by the name of Andrei F., prompting hardware reviews website AnandTech to investigate.

Analysing the smartphone’s GPU frequency during varied workloads, Brian Klug and Anand Lal Shimpi noted a maximum frequency of 480MHz, even with the most demanding games in action. But testing using certain benchmark tools saw frequencies reach 532MHz.

Similar discrepancies were also discovered when testing the CPU, and, digging deeper, Klug discovered strings in the code of an Android application package file specifically listing certain benchmark tools that would clock higher speeds. However, he also notes that some tools not listed also clocked higher frequencies.

The issue arising here is that certain benchmark tools will produce repeatable high CPU and GPU test results, while users will not normally have access to these speeds. Frequency optimisation is not unheard of in the PC industry, and Klug is unsurprised that it is now appearing in the mobile space – though that doesn’t mean he agrees with it.

“The risk of doing nothing is that we end up in an arms race between all of the SoC and device makers where non-insignificant amounts of time and engineering effort is spent on gaming the benchmarks rather than improving user experience,” he writes. “Optimising for user experience is all that’s necessary, good benchmarks benefit indirectly – those that don’t will eventually become irrelevant.”

The Verge cites Korean-language website Samsung Tomorrow’s response to AnandTech’s claims, saying “under normal conditions, the Galaxy S4 operates up to 533MHz at its best performance.” Samsung has also asserted that specific tools were not used on purpose to acheive higher benchmark scores.

Elaine Burke is the editor of Silicon Republic