![]() ![]() Realme has always been committed to bringing users excellent performance, the actual user experience is the top priority for Realme. On the other hand, regarding the test issues doubted by both sides, we are in touch with Antutu to sort it out positively.” The Realme GT was marketed with a score of over 7,50,000 on the AnTuTu version 8. The benchmarking platform said that the Realme GT delayed threads when running the multithreaded test on its benchmark app. This allowed the smartphone to use the fastest CPU cores, resulting in higher benchmark scores, AnTuTu said. Apart from that, AnTuTu also said that the Realme GT modified the reference JPG image used by the platform to reduce processing times and achieve a higher score. The smartphone allegedly used mosaic colour blocks instead of processing the image verbatim genuinely, to reduce its quality and cut down processing efforts, the company said.ĪnTuTu also said that both delaying threads and modifying the reference image are against the purpose of giving fair benchmark results. The benchmark platform also said that it questioned Realme after finding out the discrepancies, but received no response. This is not the first time a smartphone manufacturer has been caught cheating benchmark results. > possible on all platforms, but at least on Intel/AMD > It would be great if the Geekbench tests would also measure the average clock frequency during the test A platform named UL Benchmark even removed Oppo’s Find X and Oppo F7 from its platform over “high and misleading” scores in 2018.> Maynard Handley on Aug12:01 pm wrote: Companies like Huawei, OnePlus, and Oppo have been found cheating benchmark scores in the past. > systems, where the overclocking problem exists, it could be done using one performance counter. > I agree, but this is a difficult problem because of turbo/thermal throttling, etc. > Some good links there, but it is easier on Linux than Windows (big surprise), but it is possible. BENCHMARKS DELETED GEEKBENCH OVER CHEATING WINDOWS > The duration of the test is already measured by any benchmark, you just need one extra number, the count > Where it is possible, it is not difficult. > of CPU cycles from test start to test end. ![]() Dividing the 2 numbers provides the average frequency. > The link from Stackoverflow is obsolete. > where access to the performance counters is prohibited, > On a processor without performance counters or on a system > not worthwhile, because they must be calibrated for each processor model) timing integer divisions or other long instructions are > To obtain the number of CPU cycles, there is no other good way except using one performance > In all modern processors TSC is incremented at a constant frequency, regardless what the processor cores do. > Anyway, these frequency values are interesting especially > the benchmark should not record the average clock frequency, but where possible, it should be recorded. > the CPU cycles with a performance counter, unless there is a security setting > On these processors it is easy to measure the average CPU frequency, by counting > hundreds or thousands of Geekbench results exist, but all are garbage because they are not comparable. ![]() > that prohibits access, when root/admin rights would be required. > For Linux, the required code can be seen e.g. BENCHMARKS DELETED GEEKBENCH OVER CHEATING WINDOWSīENCHMARKS DELETED GEEKBENCH OVER CHEATING CODE.BENCHMARKS DELETED GEEKBENCH OVER CHEATING CODE. ![]()
0 Comments
Leave a Reply. |