Xiaomi, caught cheating in the benchmarks

This case, that of Xiaomi, is only the latest in a long list of similar actions, which goes back many years, probably at the time when device performance testing tools (in those days of PCs) began to become popular and, suddenly, rankings began to appear that valued, by performance, the different proposals on the market in processors , graphics cards, laptops, etc. We can say that it was at that time, when these rankings began to be threatened.

Threatened by attitudes that we read in MSPowerUser that Xiaomi would have had. And it is that, by masking a popular smartphone benchmark as another app, the test yielded a significantly lower result who got the same device with the official version of the app. This shows that the phone’s operating system detects if the phone is undergoing a test and, if so, hits the accelerator to get a better grade.

But what exactly is this trick? Is the phone capable of operating above its maximum speed under certain circumstances? Well no. Or, to be more exact, not entirely. In this case of Xiaomi, as in others that we have seen before, the manufacturers determine the maximum performance of the devices, but then they partially reduce it, because in this way lThe life of certain components of the same, such as the battery, is extended in a marked way. It is a small sacrifice of performance, which provides a greater good in return. Nothing to object to this point.

The problem is that, when the Xiaomi device detects that it is going to be subjected to a test, momentarily unlock that “extra” power, to score a few points more in the turn test. Once the test is finished, the device returns to its normal mode, the test result is surely uploaded to an online database, and the manufacturer, in this case Xiaomi, can take pride in the good results of their device. .

And how can this be detected? Is it necessary to decompile all the Xiaomi software to find the part of it that, when detecting a benchmark, activates the turbo mode? No, it’s actually quite simpler. It is enough, as indicated before, to prevent the smartphone from knowing that it is undergoing a test. And in this case, the user who has detected this anomaly did so by running uA special version of Geekbench posing as the Fortnite Android client.

The Xiaomi Mi 11 in the test took the bait and, consequently, the results of the test, compared to it, but when the mobile detected the test, they were much lower. 30% in single-core performance and 15% in multi-core. A difference that, of course, should be checked in what way it affects the classification of this Xiaomi smartphone in the global Geekbench ranking, but we can already imagine that it will also be marked.

Nothing to object to, I repeat, that manufacturers like Xiaomi bet on extending the life and autonomy of their devices. On the contrary, it seems to me a success. The problem is that, in order to obtain a better ranking in the rankings, its mode of operation is modified, implying a performance that is not real. That’s really ugly in this case.

Related Articles

Leave a Reply

Your email address will not be published.