Ever since Allwinner A80 was released, the Antutu scores did not add up with some devices getting around 30,000 points, while other devices scoring close to 50,000 or even higher. But what about 143,430 points? Now that would be impressive, and that’s exactly what Byron got in his Tronsmart Draco AW80 box, after changing a single parameter.
The screenshot above reports a Nexus 10 because he must be using one of Freatab ROM, where they often change the device name for better Google Play compatibility. Nevertheless what kind of sorcery is that? How could he achieve this?
It turns out build.prop, as a key called ro.sys.hiritsu. It’s set to 30 in the stock firmware, but if you change this to 95, some magic happens, and you get the very high score above. Byron reported his box got pretty hot with this setting, so it’s probably better not to try a high value… Hiritsu (ひりつ) is a Japanese word meaning “ratio, proportion, percentage”, and in this case It looks like it just set the Antutu cheating ratio… Manufacturers may decide to adjust it just a bit to show a very good score, but not too high or it’d look suspicious.
What happens if you remove ro.sys.hiritsu from build.prop? Freaktab member fess tried it, and the score falls down to 36,185, which should be closer to reality, and similar to the score (36,903) I got using Antutu X. Another member try to adjust the setting to 30 and 35 and run both Antutu and GFXBench. Antutu score goes up from one setting to the other, but GFXBench stays put at 476. GFXBench is a graphics intensive score, and maybe a CPU intensive benchmark, might have a slightly higher score, but certainly in the order of magnitude shown in Antutu.
Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.
So… it’s the firmware who cheats, not Allwinner A80, right?
amlogic has back-gourd service that detect running application and when they detect antutu 3dmark,…
they overclock CPU and GPU
funny is that all SOC vendors has opensource SDK so anybody can look and find that code 🙂
Yes. Semantically you are correct.
Since Allwinner SDK has been released/leaked, it should be possible to see what the hiritsu parameters actually does…
@Jean-Luc Aufranc (CNXSoft)
There was a reason Antutu X was made, I believe it was after people found out that Exynos scores were faked on Galaxy devices.
Whoever it was that brought up the Antutu X to fix ridiculous scores on a review. 😉
I bought a Lenovo cellphone linked below for $117. Antutu give 46000+ score. that is incorrect. Other benchmark software told me
phone is based on mk6572 and display has 1/5 the claimed resolution. Actual response and display clearly tells me that phone is not based on mk6592. RAM may not be 4G. But Antutu gives me spec. as claimed by the seller. I am sure Antutu is screwed and not reliable.
The actual specs of your phone is closer to this.
1.3 Ghz MT6582 Quad-Core CPU
1 GB RAM
720p HD Screen
The trick to produce higher scores in Antutu is true, but going to the land of real applications and not simple scores, how many SoCs from China are capable to simultaneously “encode” VP8 and H.264 4K video +25fps and broadcast live simultaneously with WebRTC (Google Web Real Time Comm.) and HLS (Apple live stream).
Even after the simultaneous video encoding and processing + the embedded web servers (HTTP+ webrtc signaling) the CPU cores have plenty of capacity to include additional processing.
Your phone has a 854×480 screen 1GB RAM and 8GB Flash memory.
I had such a phone claiming 2GB RAM, but using some other benchmarking tool revealed the truth: