Some Android are cheating on benchmarks. The worrying thing would be not to do them

  • 26

There are those who go around giving it their all in benchmarks, while diminishing performance in the “real world”, everyday use. It was the case of Samsung limiting the performance of certain apps with your GOSand in our recent reviews of the Xiaomi 12 Pro and 12 we have seen how, despite giving very high scores in benchmarksthe throttling it is present even when the mobile is cold, causing its performance to be reduced.

What if the manufacturers were right to limit? We all want a powerful mobile. But we also want the battery to last. And don’t get hot. And may it be long. Perhaps limiting performance makes sense, and the problem is the constant war by the numbers. Hitting each other in the chest to see who stands out the most in the synthetic tests.

Let’s talk about processor efficiency


To understand what is beginning to be seen in this latest high-end generation at the benchmark level, it is necessary to talk about the three major players in the mobile chip market: Apple, Qualcomm and MediaTek. Historically the best Android processors have been in the hands of Qualcomm, away from a (formerly) inefficient MediaTek that had slow and expensive processors. Things are changing.

The high-end reviews of 2022 had a common factor: the heat that the Snapdragon 888 gave off. The amount of energy that is being demanded to increase performance does not bode well

For the 2022 generation Qualcomm introduced the Snapdragon 8 Gen 1, a powerful processor, but it requires a lot of energy. The movement is not new, and it is that the Snapdragon 888 itself was already a processor that overheated the phones and ate a lot. Although Apple plays in another league at the level of mobile chips, technical tests they indicate differences that allow us to perfectly illustrate the problem.

iPhone 13 Pro (peak performance)

Samsung Galaxy S21U (peak performance)

iPhone 13 Pro (after throttling)

Samsung Galaxy S21U (after throttling)

Power consumption (W)





Performance (FPS)





In this table, with data from Anandtech, we see the poor energy efficiency of the Snapdragon 888. There is a very curious fact: the 888, mounted on a Galaxy S21 Ultra at its maximum performance peak, consumes more and runs at less FPS than an A15 Bionic that has already lowered its performance behind the throttling.

You only have to keep a couple of data: how much energy does the processor need to reach a certain level of performance

We do not want to open a battle for power here, but to show the enormous amount of energy that some chips are beginning to demand to be able to move graphics above 30 FPS in tests such as GFXBench (1440p High offscreen), something that ends up being extrapolated to heavy games.

Tg Image 1730151988

Image: Golden Reviewer.

This is even better understood with an image from the Perfdog test. On the left we see an S22 Ultra running Genshin Impact with active Samsung limitation, and to the right the same mobile, the same game, and without limitations. What we see? That to gain 10 FPS the consumption is almost double. If the mobile sustained those 8, 9 or 10W during a game, the battery would simply fly, and the heat would be very noticeable.

Other sourcesmaximizing the performance of the new MediaTek 8100 and 9000, show that it is possible to stay close to Apple in terms of energy efficiency.

Apple A15 Bionic (peak)

Qualcomm Snapdragon 8 Gen 1 (peak)

Qualcomm Snapdragon 888 (peak)

MediaTek Dimensity 9000 (peak)

Apple A14 Bionic

Consumption (W)






Performance (FPS)






The most curious fact here is that, to give 1 FPS more than the MediaTek Dimensity 9000, the Snapdragon 8 Gen 1 needs 3W more, a completely triggered consumption that has to be regulated if we want to have a balanced sustained performance. The jump from five to four nanometers seems to have been better used by MediaTek, although even with the 8100 (5nm) they are achieving great efficiency. Despite this, Apple is the one that consumes the least, without having even dropped to 4nm.

It makes sense that manufacturers are limiting the performance of their terminals. The problem is the benchmarks, where no one wants to be left behind. A war for the numbers, for being above in a ranking that does not reflect real use, and in which they show that nobody is managing to take advantage of the real power of the processors, since they are not sufficiently optimized for it.

There are those who go around giving it their all in benchmarks, while diminishing performance in the “real world”, everyday…

There are those who go around giving it their all in benchmarks, while diminishing performance in the “real world”, everyday…

Leave a Reply

Your email address will not be published.