When new devices are released, one of the first things we all do is to look for benchmarks to see how the device ranks against the competition. These can be kind of weird and aren’t always accurate, as some OEM’s alter the software to overperform when benchmarks are being taken. However, it’s a necessary evil as there isn’t really any quantifiable way to test your device against other flagships on the market.
Image courtesy of Reddit
Thanks for bringing up this issue to us. Apprantly, something was wrong in Pixel. Getting the sequential read performance of 2.9GB/s is only possible when the data is read from DRAM. During the microbenchmark test, we use DIRECT_IO to bypass the page cache. Hence, we suspect that DIRECT_IO might not work properly in the Pixel device.
Some lucky folks have already received their Pixel and Pixel XL and have been posting some various screenshots of battery usage and benchmark scores. According to the user, when running these benchmarks using the AndroBench application, there are some issues when running tests on the storage read performance.
There are some “issues” when you set up file-based encryption on your device, which is causing the read performance problems in benchmarks. Apparently, this issue is not confined to the Pixel, but also affects the Nexus 5X and Nexus 6P. The developers of AndroBench suggest that the issue lies within the storage/page caching process for Android 7.1 Nougat devices.
We’ll have to wait until more people get their hands on the Pixel and Pixel XL, but Google may have an issue to clear up on these devices in the near future.