The Apple A16 Bionic is a System on a Chip (SoC) from Apple that is found in the iPhone 14 Pro (2022) models. It offers 6 cores divided in two performance cores (codename Everest, up to 3.46 GHz) and four power-efficiency cores (codename Sawtooth, up to 2.02). According to Apple, the most improvements were made to the efficiency cores and the improved memory bandwidth (+50% thanks to LPDDR5). According to the die analysis from angstronomics, the e-core cluster still has the same 4 MB L2 cache, but the p-cores now offer 16 MB L2 cache. The SLC (System Level Cache) however seems to be reduced to 24 MB.
The Neural Engine still offers 16 cores and offers an increased performance of 17 TOPS (vs. 15.8 of the A15) for AI tasks. The integrated graphics engine still offers 5 cores and benefits mostly from the higher memory bandwidth. The new display engine supports adaptive refresh rates down to 1 Hz, the new always-on display, and improved anti-aliasing (for the dynamic island animations).
The CPU performance only shows small improvements to the last generation (Apple A15 Bionic). Still, the A16 is by far the fastest mobile CPU for smartphones and leaves e.g. the Snapdragon 8+ Gen 1 easily behind (in Geekbench 5.4 single- and multi-thread tests).
The chip is manufactured in the modern 4nm process at TSMC and integrates nearly 16 billion transistors (compared to the 15.8 billion of the A15 and 11.8 of the A14). Apple emphasized on efficiency and e.g. mentioned 20% power savings for the big performance cores and class-leading efficiency of the e-cores.
The HiSilicon Kirin 820 is a ARM-based mid-range Octa-Core-SoC for Android based smartphones and tablets. It was announced mid 2020 and contains 8 processor cores. One fast ARM Cortex-A76 core with up to 2.36 GHz, three addition A76 cores at up to 2.22 (as the old Kirin 810) and four efficiency Cortex-A55 cores at up to 1.88 GHz. Furthermore, the SoC integrates a 5G modem and a fast mid-range ARM Mali-G57 MP6 graphics card. For AI acceleration, the Kirin 820 integrates a DaVinci NPU.
The performance of the CPU part in our tests was slightly higher than the old Kirin 810 (thanks to the fast performance core). Huawei advertises 27% performance advantage for the CPU. The NPU and GPU have bigger performance gains with 73% and 38% more power.
The chip should also integrate Bluetooth 5 and WiFi 6.
The SoC is manufactured in the modern 7nm process at TSMC and therefore should be very energy efficient.
The HiSilicon Kirin 990 5G is an ARM based high-end octa-core SoC for smartphones and tablets, which was introduced with the Huawei Mate 30 Pro in late 2019. It integrates eight CPU cores in three clusters. The performance cluster contains two Cortex-A76 cores with up to 2.86 GHz, the balanced cluster contains also two Cortex-A76 cores but with only up to 2.36 GHz and finally the power saving cluster contains four small ARM Cortex-A55 cores with up to 1.95 GHz. Furthermore, the SoC integrates a neural processor (2 + 1 cores) for AI tasks and a ARM Mali-G76 MP16 GPU clocked at up to 700 MHz. Compared to the 4G version, the 5G Kirin 990 integrates a Balong 5G modem and is slightly higher clocked (2 of the 3 clusters and the GPU).
The Kirin 990 5G is produced in the modern 7nm (7FF+ with EUV) process at TSMC, the 4G version is still produced in the slightly older 7nm process as the Kirin 980.
ARM Mali-G52 MP6 GPU, 1x Cortex-A76 up to 2.36 GHz, 3x A76 nup to 2.2 GHz, 4x A55 up to 1.84 GHz, big.LITTLE, 5G TDD/FDD, Bluetooth 5, WiFi 6, AGPS, Glonass, Baidou
Average Benchmarks HiSilicon Kirin 990 5G → 52%n=5
- Range of benchmark values for this graphics card - Average benchmark values for this graphics card * Smaller numbers mean a higher performance 1 This benchmark is not used for the average calculation
v1.28
log 04. 18:45:18
#0 checking url part for id 14602 +0s ... 0s
#1 checking url part for id 12581 +0s ... 0s
#2 checking url part for id 11508 +0s ... 0s
#3 redirected to Ajax server, took 1720111518 +0s ... 0s
#4 did not recreate cache, as it is less than 5 days old! Created at Thu, 04 Jul 2024 05:15:21 +0200 +0s ... 0s
#5 composed specs +0.032s ... 0.032s
#6 did output specs +0s ... 0.032s
#7 getting avg benchmarks for device 14602 +0.001s ... 0.033s
#8 got single benchmarks 14602 +0.006s ... 0.039s
#9 getting avg benchmarks for device 12581 +0.003s ... 0.043s
#10 got single benchmarks 12581 +0.004s ... 0.047s
#11 getting avg benchmarks for device 11508 +0.003s ... 0.05s
#12 got single benchmarks 11508 +0.007s ... 0.057s