Home Blog Snapdragon 8 Elite and Elite Extreme Gen 6 Launched: The Battle with A20 Pro Begins

Snapdragon 8 Elite and Elite Extreme Gen 6 Launched: The Battle with A20 Pro Begins

TahaSite Editor
Sep 29, 2026
195 Views
Snapdragon 8 Elite and Elite Extreme Gen 6 Launched: The Battle with A20 Pro Begins

Qualcomm is drawing a new boundary among flagship Android chips with the Snapdragon 8 Elite Gen 6 and the Extreme model.

The Snapdragon 8 family has long had more than one member; just last year, Qualcomm offered the Snapdragon 8s Gen 4, Snapdragon 8 Gen 5, and Snapdragon 8 Elite Gen 5 across various segments of the premium market. However, this year, Qualcomm has split the top tier as well; now we have the Snapdragon 8 Elite Gen 6, and above it sits the Snapdragon 8 Elite Extreme Gen 6—a chip that will bring Qualcomm's most complete set of technologies to the most sophisticated Android flagships.

The 'Extreme' suffix was previously used in the name Snapdragon X2 Elite Extreme, Qualcomm's most powerful laptop chips; but now, that same logic has reached the Snapdragon 8 family. Interestingly, 'Extreme' is not just a version with a more power-hungry CPU and a few hundred megahertz higher frequency. The processing cores of both chips actually operate at the same frequency, and both reach the 5 GHz threshold for the first time in the mobile world. The main difference should be sought elsewhere, such as in artificial intelligence, graphics, and image processing—the very areas where Qualcomm apparently sees the future of flagship phones more than anything else.

2nm manufacturing process in both chips

Before the official introduction, there were rumors that Qualcomm might use different manufacturing technologies for its two flagship chips; 2nm for Extreme and 3nm for Elite; but in the end, this did not happen, and both chips are produced using a process that Qualcomm calls an "advanced 2nm process"; therefore, we are likely dealing with TSMC's N2P process; an improved version of N2 that offers about 5 percent higher performance or better efficiency under similar conditions.

The story doesn't just end with shrinking transistors. Qualcomm has also changed the layout of the chip and memory, using a type of Offset Package-on-Package packaging. In this design, the DRAM is no longer placed exactly on top of the computing unit and has been moved slightly to the side to distance itself from areas like the CPU and GPU that generate the most heat; previously, Apple had taken a similar step by moving the DRAM to the side of the A20 Pro computing unit.

5 GHz; a big number that isn't the whole story

معماری پردازنده گرافیکی اسنپدراگون (Qualcomm Adreno) با سه بخش GPU slice با فرکانس 1.45GHz

In the fourth generation of Oryon, Qualcomm has continued with the familiar combination of 2 Prime cores and 6 Performance cores, but the frequencies have increased once again. The 2 Prime cores in both the Elite and Extreme versions can reach a maximum frequency of 5 GHz, and the 6 Performance cores have a frequency ceiling of 4 GHz. For comparison, the Prime cores in the Snapdragon 8 Elite Gen 5 operated at up to about 4.6 GHz.

5 GHz is an impressive number for a mobile chip; however, one should not let this number overshadow the rest of the story. The frequency of the Prime cores has increased by about 8.5 percent, and Qualcomm itself speaks of about a 10 percent increase in CPU performance in Elite and 13 percent in Extreme; therefore, considering that the CPU frequency of the two models is the same and no specific architectural difference between them has been announced so far, the reason for the 3 percent difference in the claimed performance of the two versions is not yet entirely clear.

On the other hand, Qualcomm's claimed numbers regarding CPU performance indicate that we should not expect a major leap in IPC. It seems that a significant portion of the speed increase in this generation has been achieved through increased frequency and peripheral optimizations; a path that Apple has also followed to some extent in recent generations.

Perhaps Qualcomm's more important claim is where it is less seen in the headlines; the American chipmaker says the new Oryon can be up to 37 percent more efficient in some processes. In a chip for a device like a smartphone, which is supposed to rely on battery 99 percent of the time, lower energy consumption is far more important than a few percent increase in performance; because energy consumption and heat ultimately determine how much of the chip's power is actually available to the user.

FlexCache; Qualcomm has redesigned the CPU cache

هسته‌های ماتریس (Matrix Cores) در پردازنده گرافیکی Qualcomm Adreno برای پردازش‌های هوش مصنوعی

One of the interesting changes in the new CPU is not related to frequency or cores at all. Qualcomm has changed the way cache memory is distributed and has created a structure called Oryon FlexCache. In the previous generation, 2 Prime cores had access to 12 MB of L2 cache, and 6 Performance cores had access to a separate 12 MB of cache. In the new generation, these two independent pools have been replaced by a shared 16 MB pool that all cores can use.

At first glance, one might even say that the cache capacity has decreased; however, the advantage of the new design lies in its flexibility; meaning that each Prime core is no longer necessarily limited to its predetermined share and, under appropriate conditions, can access more than 12 MB of cache. Qualcomm is trying to keep data as close to the CPU as possible, because accessing DRAM has higher latency and consumes more energy; therefore, better use of cache can reduce some of the pressure on the main memory.

In the memory section, Qualcomm's sixth-generation chips are also prepared for new standards. Both chips still support LPDDR5X-10600, but the possibility of using LPDDR6-10600 has also been provided; support for UFS 5.0 storage has also been added to the list of features, which increases the raw data transfer rate per lane from 23.2 Gbps in UFS 4.0 to 46.4 Gbps.

Artificial Intelligence; where Qualcomm's main plan reveals itself

نمایش مشخصات فنی اسنپدراگون 8 الیت (Snapdragon 8 Elite) روی صفحه نمایش بزرگ در مراسم معرفی

If in previous generations all companies loved to attach the term AI to every new feature, the popular keyword of this generation is Agentic AI; an artificial intelligence that is not just supposed to take an input and provide a response. Agents must be able to carry out a task in several steps, switch between different tools, remember previous information, evaluate the result of one step, and proceed to the next based on it. Such a pattern naturally puts a different kind of pressure on hardware.

Qualcomm does not leave agentic AI tasks solely to the NPU. The Oryon CPU itself also has hardware accelerators for AI processing, and depending on the type of workload, processing can be shifted between the CPU and other accelerators. In the new Hexagon NPU, a new unit called the Element Accelerator has been added; a complex name for an idea that is not actually that complicated.

Neural networks constantly perform simple operations such as addition, multiplication, comparison, and activation functions on Tensor elements. Previously, such tasks were executed on more general NPU units; now, Qualcomm has dedicated a specific section for them so that these repetitive operations are performed with less overhead and lower energy consumption. The Element Accelerator sits alongside the Scalar and Vector units. Scalar is suitable for processing individual values and control tasks, Vector processes a large amount of data in parallel, and the new unit handles element-wise tasks.

Extreme starts here

نمای ارائه مشخصات فنی تراشه اسنپدراگون ۸ الیت (Snapdragon 8 Elite) روی نمایشگر بزرگ

So far, the similarities between Elite and Extreme are much greater than their differences; but in terms of the NPU, one of the first serious discrepancies between the two chips emerges. Qualcomm has allocated about 50 percent more dedicated memory for the Extreme version's NPU, although it has not announced the exact memory capacity; however, the American chipmaker wants important model data to reside as close as possible to the processing units themselves so that the chip is less frequently forced to access DRAM.

Increasing the available NPU memory becomes even more critical when running language models. The current state of the model, the conversation text, and the KV Cache must all be readily accessible. More local memory can enable a larger context window, faster token generation, and lower latency in multi-step tasks.

The new NPU continues to support INT2, INT4, INT8, INT16, FP8, and FP16 precisions, and Qualcomm specifically notes a 50 percent speed increase in the prefill stage for INT4 models. Prefill is the phase in which the model processes existing input—such as a user query or previous conversation history—all at once before it begins generating a response.

Qualcomm also considers the Extreme version better suited for running Mixture-of-Experts (MoE) models. Under the MoE concept, not all of a model's parameters are active for every request; for example, a model might have 30 billion parameters in total, but only activate the three billion parameters relevant to that specific task when generating a token. The result is the ability to use larger models without a direct and proportional increase in computational cost.

GPU; The most significant gap between Elite and Extreme?

نمای ارائه مشخصات فنی تراشه اسنپدراگون ۸ الیت (Snapdragon 8 Elite) در مراسم معرفی

If the difference in NPU memory is still an abstract concept for many users, the difference between the two chips in the graphics department is more tangible; the Extreme version is equipped with new units called Adreno Matrix Cores, units that do not exist in the regular Elite. Matrix Cores are designed to perform matrix calculations quickly; the very calculations that form a large part of neural network and machine learning processing.

Qualcomm calls the set of capabilities brought to the GPU using these Matrix Core units, Adreno Neural Fusion. Its two most important applications are AI Super Resolution and Frame Generation. For years, companies like Nvidia have been using machine learning models in desktop computers to reconstruct images at higher resolutions or generate intermediate frames, and now this same idea is entering mobile GPUs with greater seriousness.

The advantage of having Matrix Core units inside the GPU itself is that the processes required for upscaling or frame generation do not need to exit the graphics pipeline and be delegated to another accelerator such as the NPU. The AI model runs right where the frames are being rendered. Qualcomm already supports Unity and Unreal Engine, a point that indicates Neural Fusion is not intended to be just a feature for a few demos, and the company wants game developers to truly be able to utilize it.

The core architecture of Adreno is still based on three independent blocks or slices. Each slice has its own shader and ALU resources, and the activity of these various sections can be managed depending on the processing load; therefore, the GPU does not always need to operate at full capacity. There is also 18 MB of dedicated Adreno High Performance Memory (HPM) to reduce the GPU's need to access DRAM. Qualcomm states that memory power consumption has also decreased by 12 percent.

The GPU frequency has also seen a significant increase, rising from about 1.2 GHz in the Snapdragon 8 Elite Gen 5 to 1.45 GHz. Qualcomm claims a graphical performance improvement compared to the previous generation of about 35 percent for the Elite and 44 percent for the Extreme, noting that energy efficiency can be up to 40 percent better.

4K at 240 fps and 8K at 60 fps; Still exclusive to Extreme

نمودار ساختار پردازنده اسنپدراگون ۸ الیت (Snapdragon 8 Elite) و اجزای داخلی آن

The boundary between the two chips continues in image processing. Qualcomm says the data transfer capacity of the new ISP has increased fourfold. The most significant result of this change is seen in the Extreme version, where the chip can support 4K video recording at 240 frames per second and 8K at 60 frames per second. The Elite version remains at the same 4K120 and 8K30 limit.

Of course, ISP capability should not be equated directly with a phone's camera capability. Whether a phone can actually record in 4K240 depends on the camera sensor, memory bandwidth, storage space, heat dissipation, and, of course, the manufacturer's own decision. Qualcomm provides the infrastructure; utilizing it is up to the manufacturer.

The Extreme version supports the H.266 or VVC codec for the first time. VVC is designed for more efficient compression than H.265 and can require approximately 30 to 50 percent lower data rate at the same image quality; a factor that can have a significant impact on both file size and required bandwidth for heavy 4K and 8K videos.

X105; A modem integrated directly into the SoC this time

تراشه اسنپدراگون ۸ الیت (Snapdragon 8 Elite) با فرکانس ۵ گیگاهرتز روی نمایشگر

The new generation of Snapdragon has another important change; unlike separate chips in previous generations, the Qualcomm X105 modem is now integrated inside the SoC. In terms of raw speed, the X105 modem is considered a relatively limited upgrade compared to the X85. The maximum theoretical download speed has increased from 12.5 to 14.8 Gbps, and upload speed from 3.7 to 4.2 Gbps; these figures, naturally, represent the platform's capacity more than the speed that users will experience on real networks.

More interesting changes have occurred in other areas. The modem's AI processor has reached its fifth generation and can monitor signal conditions in real-time; from detecting interference and selecting a more suitable frequency to adjusting the antenna and managing power consumption.

The RF transceiver has also become 15 percent smaller and 30 percent more power-efficient, and satellite communication support has been upgraded from NB-NTN to NR-NTN; meaning from narrowband satellite communications to a platform based on 5G New Radio that has greater capacity for data services.

Quad-frequency positioning

نمایش مشخصات فنی اولین پردازنده موبایل ۵ گیگاهرتزی اسنپدراگون (Snapdragon) روی نمایشگر بزرگ

The X105 modem also supports quad-frequency GNSS for positioning and can utilize the L1, L2, L5, and L6 bands. Having multiple frequencies allows the receiver to better correct some of the errors caused by the ionosphere and reduce the impact of the Multipath phenomenon; the same occurrence that, in streets surrounded by tall buildings, causes the satellite signal to reflect several times before reaching the phone, leading to an incorrect location report.

Qualcomm states that alongside the increased accuracy, power consumption in this section has also decreased by 25 percent. The X105 modem is equipped with a new RF Front-End, and among its changes is the use of Power Class 2 LPAMIDs, which can transmit signals at higher power compared to common Power Class 3 models; the main goal is to increase range and improve connectivity at the cell edge, where the phone is at a distance from the tower and maintaining a connection becomes more difficult.

Extreme is not just a faster Elite

نمای لایه‌بندی اجزای تراشه اسنپدراگون ۸ الیت اکستریم (Snapdragon 8 Elite Extreme Gen 6)
چیدمان Offset Package-on-Package

Putting all the changes together, perhaps the most interesting point about the Snapdragon 8 Elite Extreme Gen 6 is that Qualcomm did not choose the simple path of increasing CPU frequency to build it. The CPUs of the two versions are nearly identical; the same 5GHz Prime cores, the same 4GHz Performance cores, and the same FlexCache. Both are manufactured using a 2nm process, both support LPDDR6 and UFS 5.0, and both include the X105 modem.

"Extreme" truly finds its meaning when we move beyond the CPU. More NPU memory, better readiness for MoE models, GPU Matrix Cores, Neural Fusion, AI Super Resolution, Frame Generation, 4K240 and 8K60 video recording, and H.266 support; these are the capabilities that Qualcomm has reserved for its top-tier model.

This fact largely indicates the direction in which the competition for flagship mobile chips is heading. A few years ago, almost all attention was focused on the CPU, and shortly thereafter, the GPU became the main arena of competition. Now, the distinction between high-end chips must be sought more than ever in specialized accelerators, memory proximity to processing units, and the way artificial intelligence is utilized across various parts of the SoC.

Now, we must wait for the first phones equipped with Qualcomm's sixth-gen chips to see what results these changes lead to outside of the Snapdragon Summit event slides; especially considering how higher graphics performance, a 5GHz frequency, and heavy AI tasks will impact energy consumption and heat during long-term use.

نمایش ویژگی‌های ویدئویی تراشه اسنپدراگون (Snapdragon) روی نمایشگر بزرگ کنفرانس
نمای نمایش معماری تراشه اسنپدراگون ۸ الیت (Snapdragon 8 Elite) روی صفحه نمایش بزرگ
Tags:
Snapdragon 8 Elite Elite Extreme Gen 6 Mobile Processor Smartphone Chipset Tech News
TahaSite

TahaSite

Since childhood, I have always been fascinated by computers and technology. My passion for programming and web development has driven me to constantly learn and create. At TahaSite, my goal is to provide the latest tech insights and build cutting-edge software infrastructure.

Comments

No comments yet. Be the first!