Qualcomm’s next flagship processor will be called Snapdragon 8 Gen 4
At the esteemed Snapdragon Tech Summit 2023, Qualcomm unveiled its novel flagship processor, the Snapdragon 8 Gen 3. Through an intricate discourse and ensuing discussions, further nuances were elucidated, notably the processor’s consistent manufacturing process yet a conspicuous enhancement in performance. The dialogue even intimated the nomenclature of the impending processor iteration as Qualcomm Snapdragon 8 Gen 4.
Though it remains nebulous if Chris Patrick, Qualcomm’s Vice President and General Manager of Mobile Handheld Devices, corresponded to the on-site media query regarding the Snapdragon 8 Gen 4 designation, he did concede that the forthcoming year’s processor, crafted with the autonomous Oryon CPU architecture, might witness escalated production costs. Nonetheless, this would be counterbalanced by an exponential surge in computational prowess.
Concerning the Snapdragon 8 Gen 3, even whilst adhering to TSMC’s 4nm fabrication process, its amplified efficacy is attributable to pivotal recalibrations in core configurations and operational frequencies. For instance, the unique “1+5+2” core blueprint was employed, eschewing the erstwhile design boasting four energy-conserving cores, and retaining a mere two cores of quintessential energy conservation, thereby optimizing the processor’s operational efficacy.
A meticulous inspection of the Snapdragon 8 Gen 3’s die design reveals a modicum of augmentation. However, when amalgamated with cutting-edge artificial intelligence methodologies, the processor manifests superior computational capacities at a markedly reduced power expenditure.
Post the anticipated 2024 debut of the flagship processor imbued with the Oryon CPU design, Chris Patrick intimated the launch of products catering to both the mainstream and novice market strata. Concurrent adjustments in AI design will be instated to resonate with the diverse product niches while ensuring a steadfast preservation of device-side AI computational prowess.
The lauded Snapdragon 8 Gen 3 boasts the capability to execute colossal natural language models with an excess of 10 billion parameters. Yet, its versatility permits a recalibration to a 7 billion parameter scale, streamlined through INT4 precision, ensuring seamless functionality even on devices equipped with a basal 8GB memory.
Operationalizing 7 billion parameters via INT8 precision generally necessitates around 7GB of memory. However, when transmuted to the INT4 precision modality, a mere 3.5GB of memory suffices for smooth execution.
From the purview of the Snapdragon 8 Gen 3’s maximal compatibility with 24GB memory specs, using INT4 precision effortlessly addresses computational demands exceeding 10 billion parameters. It also underpins a myriad of device-side generative AI applications, such as the showcased videographic edge expansion capability, swiftly synthesizing potential corresponding vistas through generative AI.