The hype of Artificial Intelligence is escalating like never before. You see AI integrated into every piece of technology out there. Not to forget the single most technology that we hold dear – smartphones! Our smartphones are already capable of handling AI related functionalities but with the advancement of AI systems, you’ll need a dedicated hardware to accelerate the process and bring out more features. Today MediaTek has introduced the latest SoC under its P Series at MWC – the Helio P60 that comes with a dedicated Neural Processing Unit (NPU).
The Helio P60 is the successor of P30 and has brought a significant amount of changes which explains the jump in number. It uses the ARM big.LITTLE architecture in an octa-core setup with four Cortex A73 cores (big) and four Cortex A53 cores (little) both clocked at 2.0GHz. It is a big upgrade from the P30 as it only used the Cortex A53 cores which is ideal for small tasks compared to Cortex A73 capable of handling heavy jobs.
The new P60 processor with its Cortex A53 and Cortex A73 combination can switch tasks between big and little cores depending on the processing power it requires which makes it more efficient. Adding to the efficiency chart, it is based on 12nm FinFET technology making it one of the most powerful and power efficient chipset in P-Series. This shift has resulted in 70% boost in CPU performance.
This powerful CPU has been coupled with the Mali-G72MP3 GPU clocked at 800MHz. The GPU has improved clock speeds compared to last gen Mali G71-MP2 used on P30. It has a larger L1 cache (on-chip cache) and larger tile memory buffer (image separation into fine tiles) which helps in enhancing the performance. It has several improvements to refine Machine Learning Inference and depicts core functionality of Machine Learning Algorithms. The GPU overall also has 70% improvement.
The feature highlight of Helio P60 is the dedicated NPU with NeuroPilot AI system at work. This enables deep learning facial recognition, object and scene identification, user behavior-informed performance and other AI and AR application enhancements. It can handle AI operations seamlessly at 280GMAC/s, maintaining twice the power efficiency of GPU-based AI processing. The sole purpose of using a dedicated NPU is to accelerate the AI processes without putting much load on the GPU. The MediaTek’s NeuroPilot SDK supports a vast array of AI frameworks. This makes it easier for developers to bring out AI applications optimized for this hardware.
Coming to the optics support, it can handle a dual camera setup of 20+16MP or a single 32MP sensor. The 32MP sensor can capture full HD videos at 30fps while the use of 16MP sensor will enable 90fps recording. The P60 depth mapping engine that can produce real-time bokeh effect. It uses three Image Signal Processors that equates to 18% power efficiency compared to P30 meaning taking loads of pictures all day won’t be harsh on the battery. It also offers Phase Detection Auto Focus (PDAF), Electronic Image Stabilization (EIS), Real-time HDR recording/viewing, face detection, scene detection, and multi-frame noise reduction.
On the connectivity side, the P60 offers dual 4G LTE which ensures both the SIM used are connected to 4G networks. The LTE technology is optimized to offer stronger connection, faster call setup and better sound quality. For data connectivity, it has Cat-7 (Download) and Cat-13 (Upload). It supports dual-band ac Wi-Fi but unfortunately lacks the Bluetooth 5.0.
Specification of MediaTek Helio P60
- CPU: Octa-core with four ARM Cortex-A73 up 2.0 GHz and four ARM Cortex-A53 up to 2.0GHz
- GPU: Arm Mali-G72 MP3 at 800MHz
- Memory: Up to 8GB, dual-channel LPDDR4x @ 1800 MHz
- Storage: eMMC 5.1 / UFS 2.1
- Camera: Up to 16+20MP or 32MP
- Display: Up to Full HD+ 20:9
- Connectivity: 802.11ac Wi-Fi, Bluetooth 4.2 and FM radio
- Cellular: Cat-7 (DL) / Cat-13 (UL); Dual 4G VoLTE; TAS 2.0; Global IMS
- Multi-GNSS: GPS/GLONASS/Beidou or GPS/GLONASS/Galileo co-receive
- MediaTek CorePilot 4.0: Energy Aware Scheduling, thermal management and UX monitoring
- Fabrication: TSMC 12nm