For many, the biggest excitement from Apple’s new content announced this fall is undoubtedly the A10 Fusion chip that powers the iPhone 7. After all, the potential brought by the significant performance enhancement may not be obvious, but the benefits are felt by everyone.

The Most Powerful Mobile Processor A10

Every year, Apple’s new chip launch is a major highlight. Just watching the continuously rising performance curve is enough to excite any smartphone enthusiast. At the fall launch event, Apple introduced the A10 Fusion chip that powers the iPhone 7 and iPhone 7 Plus. Naturally, the performance of the new chip has seen great improvements compared to its predecessor, with a 40% increase that certainly makes one wonder what it can actually do.

Furthermore, the graphics performance of the A10 Fusion chip has improved by 50% compared to the A9. During the launch event, Apple demonstrated the mobile game “Oz: Broken Kingdom”. The game’s stunning magical effects and physical destruction effects left a deep impression, even showcasing an astonishing performance of 400 enemies on screen simultaneously. Interestingly, the chip’s power consumption has decreased by about 30% compared to the A9.

For many, performance comparison is indeed more intuitive through benchmark scores. Let’s take a look at the specific benchmark results provided by foreign media:

GeekBench 3 Single-Core:

The iPhone 7 is approximately 38% faster than the iPhone 6S with the A9, over 100% faster than the iPhone 6, and nearly 150% faster than the 5S.

GeekBench 3 Multi-Core:

The overall performance is basically on par with single-core, crushing the 5S, so it makes sense that Cook mentioned the iPhone 7 aims to make those holding iPhone 5/5S feel the need for an upgrade.

Comparing with Snapdragon 820

Snapdragon 820 is also a quad-core design, but the A10 leads in both single-core and multi-core benchmark scores by about 50%, and it can lead the Snapdragon 810 by nearly 200%. The multi-core gap is smaller but not evenly matched.

Kraken and Google Octane Tests

These two tests are based on browser JS performance, which is closer to the actual CPU performance we use daily, and the iPhone 7’s lead remains significant.

The GPU Performance is Also Powerful

The CPU performance of the iPhone 7 A10 processor is exceptionally robust, with unbeatable single-core scores and multi-core performance that can suppress octa-core. So, what about the GPU performance? Apple claims a 50% improvement over the A9.

Now, foreign media have released the first GPU performance results for the iPhone 7 from the authoritative GFXBench, including T-Rex and Manhattan sub-projects.

According to the horizontal comparison at a unified resolution of 1080p off-screen, the iPhone 7/7 Plus GPU achieved excellent scores of 97FPS for T-Rex and 52-59FPS for Manhattan, which is an improvement of up to 20% and 42% over the iPhone 6S/6S Plus, closely matching Apple’s claims. However, it is somewhat strange that the iPhone 7 Plus is slightly lower.

This is also the first time the Manhattan project has approached the 60FPS mark.

Compared to the strongest Snapdragon 820 Adreno 530 in the Android camp, the iPhone 7 also wins, with the Manhattan project leading by as much as 25%, while T-Rex only leads by 9%.

At the native resolution, the iPhone 6S series has already reached the limit of 60FPS.

Additionally, the GPU architecture in the A10 processor remains unclear, but it is estimated to be the new generation PowerVR 8XT.

Although Apple’s new chip launches are always called “best than ever”, this phrase carries even more meaning for the A10 Fusion. As everyone is very curious about this chip, today we will take a closer look at it and see what we know and what we don’t about the A10 Fusion.

The Source of Power

We can say that the A10 Fusion chip represents the most significant advancement since Apple transitioned its system-on-chip to a 64-bit architecture. The A10 Fusion chip features four cores and 3.3 billion transistors.

Although Apple has not disclosed the transistor count for the A9 chip, we can speculate that this number is certainly between the 2 billion of the A8 and the 3.3 billion of the A10 Fusion. Because Apple has not published specific numbers, the A9’s transistor count should not exceed 3 billion.

Additionally, Apple revealed that its GPU is designed with six cores, but benchmark results suggest that its L1 and L2 cache sizes remain unchanged.

The reason the A10 Fusion chip has over 50% more transistors than the A8 is undoubtedly due to its unique four-core design (even though the additional two cores are quite small). Moreover, the enhanced image signal processor in the iPhone 7 has also played a significant role. If its manufacturing process is the same as the A9, continuing to use TSMC’s 16nm FinFET technology, then it is clear that the A10 Fusion chip is larger than its predecessor.

Why didn’t Apple adopt the more enticing 14nm FinFET process from Samsung? This is mainly due to the complexity it would add to production. In contrast, optimizing chip size and component layout is much more mature and stable.

Interestingly, Apple mentioned that the A10 Fusion chip’s performance can be up to 40% better than the A9, while we know from benchmark tests that the former has a core frequency of 2.33GHz, which is only 25% faster than the latter’s 1.85GHz on paper. This implies that Apple may have gained a significant improvement through architectural optimization.

Despite the only 15% increase in core frequency, considering there has been no change in the chip’s manufacturing process, this is already a significant advancement. Apple likely achieved this due to optimizations in the thermal efficiency of the chip’s information packaging module. This is almost the only possibility since Apple employs a four-core design with two high and two low-performance cores.

A core frequency of up to 2.34GHz allows Apple to approach competitors Qualcomm and Samsung in terms of device specifications. To achieve this, the design of the transistors may have also changed. By increasing the voltage, Apple can achieve higher frequencies. Although this inevitably leads to energy waste, it is not a major issue since the A10 Fusion has those two very impressive cores.

Small is Beautiful

Compared to the performance increase of the iPhone 7, the industry is more concerned about its unique design of high and low-performance core pairing. Through this improvement, Apple has achieved a new dynamic voltage frequency scaling mode. Apple designed a dedicated performance controller for this, which can intelligently manage core load, allowing certain cores to be completely shut down.

This concept seems similar to what ARM has adopted, with designs like Cortex-A15 in 2011 and big.LITTLE in 2012, but Apple’s advantage lies in its ability to redeploy any iOS software interface to the performance controller as needed, making its application range broader.

Regarding these two low-performance cores, the focus is on whether they are Apple’s custom design or entirely derived from ARM, such as Cortex-A53. Although Apple has always preferred fully custom designs, it has not shied away from using existing CPUs, as seen in the first generation Apple Watch, which used Cortex-A7. This suggests that the dual-core chip used in the Apple Watch Series 2 may very well be the two low-performance energy-efficient cores used in the A10 Fusion.

According to Apple’s promotion, the results of adopting this design are astounding. We know that the iPhone 7’s screen brightness has increased by 25%, which theoretically is a “power-hungry” feature. However, under these circumstances, Apple claims the iPhone 7 can still last two hours longer than the iPhone 6s. If the actual effect is as good as claimed, the A10 Fusion chip’s energy management capability can be described as astonishingly efficient.

What’s Next

One critical question is why Apple chose to implement this unique chip design in the iPhone 7. The answer is quite simple: Apple has made extensive optimizations to its main core design, and these improvements have nearly reached their limits, with diminishing returns. To enhance performance, the simplest method is to increase the clock speed, but the issues of energy consumption and heat generation cannot be ignored.

Through design optimization, the A10 Fusion chip’s size has likely increased to accommodate more transistors, but chip size cannot increase indefinitely, and the CPU’s clock speed cannot be infinitely raised either—high-end desktop CPUs have been constrained between 3GHz and 4GHz for the past decade.

Through optimization and redesign, Apple has achieved another performance increase for the A10 Fusion chip, but with even lower energy consumption, which is undoubtedly an innovation. With iOS 10 bringing many new features and functionalities, we are eagerly looking forward to seeing what the A10 Fusion and iOS 10 will bring in terms of software and hardware integration.

It is noteworthy that Apple has created the W1 chip for AirPods, and executive Phil Schiller stated that this is the company’s first wireless chip, hinting that more is to come in the future. Perhaps, Apple’s ambitions are about to expand again.

There are still many mysteries surrounding the A10 Fusion chip, and we need more disassembly, testing, and analysis to understand it. What else will Apple bring us next? It is truly exciting.

Editor: Chip Intelligence – Linzi (Comprehensive reports from Fast Technology, Weifeng Network)

——————————-

Related Articles:

Moore’s Law Fails: What Other Methods Do We Have to Dramatically Increase Computing Performance?

Samsung Negotiates Graphics Chip Technology Licensing with AMD and NVIDIA, Aiming to Develop GPU Independently

AMD Unleashes 32 Cores, Intel Responds with 28 Cores!

AMD Plans to Raise $1.02 Billion: New Architecture Zen Aiming to Strike Intel

Samsung Note 7 Recall Truth: Was It Really for Honor?

GlobalFoundries Launches 12nm FD-SOI Process: Performance Comparable to 10nm FinFET, Lower Cost

Intel’s Seventh-Generation Core Processor Kaby Lake Detailed: 14nm+ Process, Nearly 20% Performance Increase

Intel Opens Foundry for ARM Chips: TSMC and Samsung Face Great Pressure

NVIDIA’s Next-Generation Tegra Chip to be Released Next Month: Codenamed Parker, Upgrading to 16nm Pascal GPU

———————————–

For industry exchanges and collaborations, please add WeChat: xintiyan001

Submit articles to: [email protected]

Chip Intelligence Official Group: 221807116