Snapdragon 8 Elite Ushers in the New AI Era: Qualcomm Showcases the Future of Mobile Chips at Snapdragon Summit 2024

Snapdragon 8 Elite Ushers in the New AI Era: Qualcomm Showcases the Future of Mobile Chips at Snapdragon Summit 2024The 2024 Snapdragon Summit is in full swing, and Lei Technology is on the ground in Hawaii, providing exclusive coverage. From the core content of the summit, it's evident that Qualcomm is eager to show everyone what a true AI era looks like

Snapdragon 8 Elite Ushers in the New AI Era: Qualcomm Showcases the Future of Mobile Chips at Snapdragon Summit 2024

The 2024 Snapdragon Summit is in full swing, and Lei Technology is on the ground in Hawaii, providing exclusive coverage. From the core content of the summit, it's evident that Qualcomm is eager to show everyone what a true AI era looks like. Qualcomm devoted a significant portion of the summit to highlighting its breakthroughs in AI, showcasing how the new Snapdragon chips utilize powerful technology to build an AI ecosystem.

Snapdragon 8 Elite Launched - Performance Rivals Desktop Processors

The event began with a brief review of Snapdragon's key achievements in the past year, including the recently released Snapdragon X series of commercial PC products. Thanks to its NPU, which outperforms its peers, the Snapdragon X series laptop has created a new category: Copilot+APP. Furthermore, the Qualcomm Oryon CPU used within the Snapdragon X series has redefined the benchmark for notebook energy efficiency.

What happens when you put the powerful Oryon CPU into a mobile chip? At the 2024 Snapdragon Summit, Qualcomm unveiled its new Snapdragon 8 Elite mobile platform. The platform's newness is not just in the product naming but also in its architecture, performance, and energy efficiency.

Second-Generation Oryon CPU: A Perfect Balance of Performance and Efficiency

The new Snapdragon 8 Elite features the second-generation Qualcomm Oryon CPU (2+6). It's important to note that this (2+6) configuration is not the traditional P-core and E-core setup found in most mobile processors; instead, it follows a similar "all-large-core" architecture to the first-generation Oryon CPU.

The second-generation Oryon CPU is comprised of two core clusters:

• The first cluster features 2 super cores (Prime) running at 4.32GHz, each equipped with 192KB of L1 cache.

• The second cluster includes 8 performance cores (Performance) running at 3.53GHz, each equipped with 128KB of L1 cache.

The Snapdragon 8 Elite completely removes the design of energy-efficient cores.

To accommodate the high demand for fast data access from AI-powered devices, the second-generation Oryon CPU supports 5.3GHz LPDDR5X memory and is equipped with 24MB of L2 cache. To avoid potential latency issues with the large L2 cache, the second-generation Oryon divides the 24MB L2 cache between the two core clusters (12MB+12MB).

With this unique architecture, the second-generation Oryon delivers performance that surpasses other mobile processors even when implemented in a mobile phone. For example, in the GeekBench6 benchmark, an engineering sample using the Snapdragon 8 Elite achieved a score of 3209/10205 (single-core/multi-core), which is 134% faster than Intel's Core Ultra 9 Gen 2, essentially packing desktop-class performance into a mobile chip.

According to Qualcomm, the Snapdragon 8 Elite achieves a 45% improvement in both single-core and multi-core performance compared to the Snapdragon 8 Gen 3. More surprisingly, even without using energy-efficient cores, power consumption is reduced by 40%.

For mobile devices like smartphones, energy efficiency translates to a significant and comprehensive advantage. This means the phone will experience a significant decrease in heat generation, not only maintaining high performance for longer but also reducing the need for heat dissipation hardware. Lower power consumption also means phones can dramatically reduce the size of the battery without compromising battery life or achieve longer battery life with the same battery capacity, making the phone more streamlined.

Adreno GPU: Unleashing Mobile Gaming Potential

Now let's talk about the GPU. The new Snapdragon 8 Elite features a 40% performance boost compared to its predecessor in terms of GPU performance. The GPU slicing architecture, combined with a 1.1GHz clock speed and 12MB of graphics cache, provides a strong foundation for the Snapdragon 8 Elite's graphics performance.

The Snapdragon 8 Elite's Adreno GPU also supports several graphics and gaming features, such as Snapdragon Game SuperResolution 2.0, FSR, VRR, ray tracing, and intelligent frame interpolation technology that can increase frame rates by 100%, and even the first mobile platform to support the Unreal Engine 5 Nanite technology.

In the mobile phone market, "gaming performance" has been a primary reason for users to choose flagship devices. With the support of the new GPU, the boundaries between mobile and console gaming are becoming increasingly blurred. In 2024, many gamers have achieved "multi-device simultaneous play," and many mobile handheld devices even utilize Snapdragon chips a point highlighted at the Snapdragon Summit keynote. The emergence of the Snapdragon 8 Elite is expected to further promote the integration of mobile and console gaming, and with Snapdragon's signature 5G and Wi-Fi connectivity, gaming will become truly ubiquitous.

Hexagon NPU: A Comprehensive Upgrade to AI Capabilities

Besides being a new starting point for Snapdragon CPUs, the Snapdragon 8 Elite is also a new milestone for Snapdragon NPUs. The Snapdragon 8 Elite features a new Hexagon NPU. The Hexagon NPU includes three different accelerators: Tensor, Scalar, and Vector. All three cores have been enhanced, with the latter two even having additional cores. In terms of AI scheduling, the three accelerators correspond to different AI model calculations, giving the Snapdragon 8 Elite stronger AI model support. Moreover, Qualcomm has adjusted the concurrent performance working mode of the Snapdragon 8 Elite NPU, allowing AI calculations and computer vision (CV) calculations to coexist on the NPU and dynamically allocate the required resources.

The integration of multimodal models has significantly enhanced the Snapdragon 8 Elite NPU's response speed for AI instructions. Taking the commonly used voice interaction as an example, support for LMM (multimodal models) allows the NPU to directly recognize voice content without relying on external plugins to convert it into text first, significantly improving the phone's response speed. Similarly, the addition of LMM enables the NPU to directly understand the content of images, accelerating image translation and enhancing the NPU's understanding of images.

Speaking of images, let's touch upon the NPU's application in imaging. The AIISP in the Snapdragon 8 Elite can now work with the NPU at the architectural level, enabling astounding data transfer capabilities for the Snapdragon 8 Elite a throughput of 4.3GP/s (gigapixels per second) allows the Snapdragon 8 Elite to simultaneously support three 48MP cameras.

Additionally, this allows the NPU in the Snapdragon 8 Elite to directly access AIISP through memory direct connection technology, enabling the phone to leverage the NPU's capabilities to assist ISP in faster segmentation and understanding of individual elements within images, achieving faster "blur" or "object removal."

This design offers more possibilities for the Snapdragon 8 Elite's imaging capabilities, such as adding real-time, adjustable-angle simulated lighting effects to videos or performing AIGC object removal on videos.

The Focus in the AI Era Isn't "Super Apps" but "Connecting Apps"

At the Snapdragon Summit 2024, we also saw a very interesting perspective: the focus in the AI era is not "AI apps" but using universal AI capabilities to connect different apps, integrating AI into various apps, using software to expand the boundaries of AI capabilities while reshaping the app experience.

Simply put, future mobile AI use cases should be divided into four components: AI developers, apps, mobile OS, and NPU. Users interact with apps and make AI requests, apps connect to the NPU through frameworks like ONNX and DirectML along with the Qualcomm AI software stack, the NPU processes the request, and the OS schedules the app and presents the result.

For example, in Honor's demonstration, Dr. Guo Rui, CMO of Honor, showcased how Honor utilizes Qualcomm AIAgent to implement a new AI interaction: the phone preloads training results, and users can directly use voice commands to instruct the AI assistant to order coffee or cancel automatic deductions. Upon understanding the user's instruction, the AI assistant can simulate user clicks based on current elements on the user interface, step by step to complete the task requested by the user, such as opening the food delivery app to order coffee or opening the payment app and summarizing automatic deduction options.

Although the process of AI operating apps currently feels relatively "primitive," Qualcomm says this AI capability has even more possibilities in the future, like bypassing app interaction interfaces to directly acquire the logic behind the app interface for "off-device" calculations, thereby comprehensively improving interaction efficiency.

In the demonstration of the "Naraka: Bladepoint" mobile game, NetEase showcased a brand-new, side-by-side AI teammate powered by the capabilities of the Snapdragon 8 Elite NPU that can understand user voice and text commands. Users can control the AI teammate with in-game shortcuts or give commands as if they were playing with a friend.

The Snapdragon 8 Elite NPU's language comprehension ability can quickly understand user instructions in natural language and provide feedback through the AI teammate.

Is the Future the Era of AI, and also the