Similar to the naming convention used for PC computing platforms, Qualcomm's flagship mobile computing platform for the year, unveiled at the Snapdragon Tech Summit 2024, will be officially named Snapdragon 8 Elite. As previously announced, it will feature an Oryon CPU with a fully proprietary architecture, employing a Prime main core and all-large core architecture, touted as the world's fastest mobile CPU.
As a naming method for new product identification
The new naming method is not difficult to guess that it is related to the Oryon CPU, which also adopts a fully independent architecture design. The same naming method is used as the computing platform previously launched for PC devices, so as to use consistent product names for product identification.
Qualcomm explained that the adjustment to the naming of the computing platform is to use the name "Elite" to highlight the Oryon CPU design with its own architecture and the top-level user experience presented by combining multiple Snapdragon brand technologies.
However, Qualcomm hasn't revealed whether, following this change in naming for its flagship computing platform, it will also adjust the naming scheme for its existing Snapdragon 7, Snapdragon 6, and Snapdragon 4 series computing platforms. However, based on Qualcomm's past product planning strategies, this is expected to be the inevitable development direction.
Equipped with the second-generation Oryon CPU design, all large cores and a new computing block design
As for the Snapdragon 8750 Elite model "SM8-AB" launched this time, it mainly adopts the second-generation Oryon CPU design, and increases the cache memory to 24MB to improve data computing throughput efficiency. It also adopts a new block type design in the Adreno GPU and Hexagon NPU, and further improves computing performance, enabling it to exert higher self-generated artificial intelligence computing capabilities. It is also built on TSMC's second-generation 3nm process (N3E).
The CPU's main core operates at a clock speed of 4.32GHz, while the performance cores can reach a maximum clock speed of 3.53GHz. Following MediaTek's lead, the chip eliminates the energy-saving core design, primarily due to lower power consumption achieved through TSMC's second-generation 3nm process. The chip also achieves a balance between performance and battery life through core computing allocation.
This time, a new computing block design is adopted. In addition to increasing the CPU computing block area, it also improves the Adreno GPU computing block. In response to the increasing demand for artificial intelligence computing, the Hexagon NPU computing block is also enlarged. In addition, an image processor designed for artificial intelligence computing needs is also replaced to correspond to smarter camera shooting effects, and can also be used for computer vision-type artificial intelligence experiences.
According to Qualcomm, the Snapdragon 8 Elite boasts a 45% increase in single-core performance and a 45% increase in multi-core performance compared to the previous generation. Web browsing performance also improves by 62%. Regarding power consumption, the Snapdragon 8 Elite reduces CPU power consumption by up to 45%, GPU power consumption by 40%, and AI computing power consumption per watt by 45%. Overall power consumption is approximately 27% lower than the previous generation.
Maintain the optimal configuration of 8 cores
In terms of CPU core configuration, the Snapdragon 8 Elite uses 2 main cores and 6 performance cores to form a "2+6" core combination, which is significantly different from the previous Snapdragon 8 Gen 4's "1+3+4" combination. In addition, unlike MediaTek, it also retains performance cores with lower operating clock speeds in the core combination to achieve higher computing performance.
Qualcomm pointed out that the 8-core configuration design has now become the mainstream in the market. At the same time, years of development experience have confirmed that the 8-core configuration can achieve better performance distribution and multi-core execution efficiency. Therefore, it is believed that the 8-core configuration will be the best combination at present. There are currently no plans to propose designs with fewer or more cores.
Further enhance artificial intelligence computing performance
To meet the growing demand for on-device AI computing, Qualcomm has enhanced the CPU's performance, combined with a GPU and NPU configuration with larger computing block areas, and supported more data perception capabilities and a shared memory architecture. This allows the Snapdragon 8 Elite to improve AI computing performance while delivering lower latency execution efficiency.
配合此次核心架構設計，Qualcomm說明此次推出的Snapdragon 8 Elite可結合NPU中的張量加速器、6核心設計向量加速器與8核心設計純量加速器，搭配運作時脈可達5.3GHz的LPDDR5X雙通道記憶體設計，使得NPU整體效能可提升45%。
此外，Snapdragon 8 Elite除了可支援大型自然語言模型常見量化精度INT4、INT8、INT16與FP16半精度浮點數，更可對應INT8及INT16的混合精度計算，藉此對應市場更廣泛的大型自然語言模型執行運作。
The multimodal AI assistant design integrated through the computing platform can achieve faster execution efficiency in large-scale natural language models and basic AI calculations, and respond to more in-depth personalized interactions. At the same time, it can also respond to long command input and output, thereby speeding up the response performance of AI application services on mobile phones. For example, digital assistant services that interact with voice can respond more quickly, and the response speed of computer vision is also accelerated.
The Sensing Hub design, which responds to various types of data perception, corresponds to a 60% increase in artificial intelligence computing performance and a 34% increase in memory resource usage, making many artificial intelligence calculations smoother.
On the other hand, the image processor, which is very important for computer vision computing, has been equipped with an additional artificial intelligence design, corresponding to a data throughput of up to 4.3GP/s and can respectively support 4800-megapixel lenses. It can also be directly connected to the Hexagon NPU, allowing artificial intelligence visual perception application services to run more efficiently, thereby corresponding to applications such as image recognition and environmental perception.
When applied to image shooting, it can further correctly identify the shooting scene and automatically adjust the camera's optimal shooting parameters. At the same time, it can also make the subject's skin color, ambient light, sky and other details appear clearer and more natural.
More advanced gaming experience
In terms of gaming experience, the newly launched Snapdragon 8 Elite increases GPU computing power by 12% and increases GPU available memory to 40MB, reducing power consumption by 40% and boosting real-time ray tracing performance by 35%. Furthermore, the Qualcomm Adaptive Performance Engine 4.0 technology can enhance the performance of games like Grid Legends.
In addition, the Snapdragon 8 Elite is the first computing platform to support Nanite virtual geometry computing in the Unreal Engine 5.3 game engine. This allows for more realistic image details through artificial intelligence rendering, and can handle large numbers of geometric object operations simultaneously at a single time. It can also be designed with Snapdragon Gaming for a better gaming experience.
5G connectivity and other platform designs
As for 5G connectivity, the Snapdragon 8 Elite integrates the Snapdragon X80 5G data chip, which can correspond to the 5G Advanced-ready design specification and support mmWave and sub-6GHz frequency bands, corresponding to 5G independent networking and non-independent networking connection types, as well as mmWave connection alone or mmWave and sub-6GHz frequency band combined connection mode, up to 10 Gbps download and 3.5 Gbps upload speeds. Among them, mmWave connection supports 8 carrier aggregation and 2×2 MIMO, while the sub-6GHz frequency band corresponds to 4×6 MIMO.
In addition, the Snapdragon 8 Elite also introduces the FastConnect 7900 platform with second-generation artificial intelligence computing technology, and can optimize multi-antenna operation mode. It also adds satellite connection function that supports NB-NTN specifications, and integrates Wi-Fi 5.8 wireless connection specifications with a peak download speed of up to 7 Gbps, as well as Bluetooth 6.0, dual Bluetooth antennas and UWB ultra-wideband network design.
Display output supports up to 4K@60 Hz or QHD+@240Hz specifications, and external displays can output up to 8K@30Hz or 1080@240Hz specifications. It also supports UFS 4.0 storage components, USB 3.1 Gen 2 ports with USB-C interface design, and Qualcomm Quick Charge 5 fast charging technology.
The security design includes the Truepic video content certification system that corresponds to the C2PA specification, as well as a content copyright protection mechanism for artificial intelligence models.
Will cooperate with many OEM companies
Following the announcement of the Snapdragon 8 Elite, Qualcomm has confirmed that brands such as ASUS, Honor, iQOO, Motorola, Nubia, OnePlus, OPPO, Red Magic, Redmi, realme, Samsung, vivo, Xiaomi, and ZTE will launch terminal devices using this computing platform. Xiaomi and OPPO have also confirmed that this computing platform will soon be used in their new mobile phones, and more brands are expected to launch terminal devices using the Snapdragon 8 Elite in the future.
