Half a year later, real-time ray tracing will be standard on Android flagships

In addition to the iPhone equipped with self-developed A-series chips, most of the SoCs of Android manufacturers come from Qualcomm, MediaTek and Samsung.

▲ Picture from: wccftech

However, Qualcomm, MediaTek and Samsung are not deeply involved in SoC's independent design of CPU and GPU architecture, and prefer to focus on Arm's public version architecture.

This brings a problem. When the Arm public version is poorly designed , it is easy to cause a chain reaction, the flagship chip does not perform well, and finally the Android flagship product also consumes a large area of ​​heat and power.

▲ Picture from: "Game of Thrones"

If the process yield of the chip foundry is not high, it will be even worse.

In recent years, Android high-end SoCs have also fallen into such a vicious circle, and the Android flagships of various factories have also been dragged into the water in recent years.

Originally considered a relatively niche SoC knowledge, it has gradually become a purchasing awareness of ordinary consumers.

However, compared with the process technology that is more difficult to catch up with, after Arm's public version architecture is finalized, it can gradually reverse the disadvantage through optimization.

▲ The first generation Armv9 CPU core picture comes from: Arm

Last year, Arm introduced the Armv9 instruction set in the public version architecture, bringing three new public version CPU cores, Cortex-X2, Cortex-A 710 and Cortex-A510.

In the end, in the Snapdragon 8 Gen1 and Dimensity 9000, this new architecture has a certain improvement compared to the previous generation, but there is still a lot of room for optimization .

At the Total Compute Solutions 2022 conference a few days ago, Arm announced the latest generation of the public version of the architecture, including the new Cortex-A three CPU cores and a new GPU series.

The core of the new generation of CPU architecture is almost revolving around optimization and enhancement, while the GPU is really about innovation.

CPU focuses on improving energy efficiency

The advantage of Arm chips is the energy efficiency ratio, but in recent years, the emergence of super-large core Cortex-X has also begun to make a lot of progress in the absolute performance of Arm chips on the mobile side.

The emergence of a new generation of Cortex-X3 is still for peak performance. Under the same 3.3GHz clock frequency as Cortex-X2, X3 has a 25% performance improvement.

If the X3 runs at 3.6GHz, there is a 34% improvement over some mainstream notebook processors.

Obviously, the overclocking potential and higher performance (and power consumption) of the Cortex-X3 super core is for Arm's PC, not for the next generation of mobile phone flagship SoCs.

The big core Cortex-A715 is more like an improvement on the previous A710 in terms of naming, focusing on efficiency.

After the core structure is optimized, the Cortex-A715 brings a 5% performance improvement and a 20% reduction in power consumption.

From the curve point of view, no matter what state it is in, the A715 has a better energy efficiency ratio than the A710. To this end, Arm also said that under the same frequency and cache, the A715 already has performance comparable to the X1.

In addition, on the big core, Arm completely canceled support for 32-bit App. As a result, the A715's instruction decoder is 4x smaller than its predecessor, making the A715 core more efficient in area, power, and execution.

The name of the middle core is still Cortex-A510, but Arm has still optimized it and called it a brand new A510. (It feels more like a redo of the A510, after all, the previous game had a lot of problems.)

The new A510 core, with improved timing and optimized efficiency, reduces power consumption by 5%. Most importantly, the A510 has re-added support for 32-bit apps.

To this end, this new A510 will also be used in other fields such as the Internet of Things, and similarly, not only for smartphones.

In addition to the optimized iterations of the three own Cortex cores, Arm's DSU (DynamIQ Shared Unit) is also updated to support 12 cores and 16MB of L3 cache.

As a result, future Arm processors may have a richer core combination. In addition to the common 1+3+4 three-plex architecture, it can also be a combination of 1+4+4, 2+2+4, etc., and even in some In pursuit of higher-energy chips, the ultimate combination of 8+4+0 can also be used.

In general, Arm's new generation of Cortex cores are mostly optimizing the structure of the previous generation, trying to provide a better energy efficiency ratio, and correcting some unreasonable designs of the previous generation.

In addition, Arm's core design is not limited to smartphones, but has begun to expand the coverage of each core. The ultra-large core X3 focuses on the development of the PC field, while the mid-core A510 has stronger compatibility for the Internet of Things and other fields.

GPU ray tracing will become standard

In the past, Arm's public GPU was weaker than the CPU core, more like a basic configuration.

Arm's public version of Mali GPU is constantly iterated. In the previous generation of flagship chips, the public version of Mali's performance is close to Qualcomm's self-developed Adreno.

With the release of a new generation of CPU cores, Arm has also updated a new generation of public GPUs. Unlike the previous routine upgrades, this time Arm brings a new Immortalis flagship series.

▲ Immortan Joe picture from: "Mad Max 4"

The biggest difference between Immortalis series and Mali in function is the support for ray tracing, and in terms of core, Immortalis-G715 has at least 10 cores, while Mali-G715 can be equipped with 7~9 cores.

The difference in the number of cores and the support for ray tracing on the hardware make the Immortalis-G715 more useful. Like the Cortex-X3, it is more like making a foreshadowing for the Arm desktop chip in advance.

Andy Craigen, director of product management at Arm, claims that the Immortalis-G715 uses only 4% of the shader core area, which brings a 3x performance boost through hardware acceleration.

In fact, software-based ray tracing is available on the previous-generation Mali-G710 GPU from Arm. In addition, Samsung's Exynos 2200 is also working with AMD to introduce support for ray tracing in hardware.

Whether it is Arm or chip manufacturers, ray tracing on the mobile terminal is being deployed in advance, but at present, most of the external demos are a few demos, and game developers have not made clear actions.

▲ New GPU ray tracing demo demo (movies are compressed), pay attention to the changes in light and shadow Image from: Arm

For mobile 3D games, a wide range of hardware light is supported, somewhat like the darkness before dawn.

However, in an interview with TheVerge, Arm's Paul Williamson hinted that when the terminal equipped with the Immortalis-G715 GPU goes on the market next year, there will be corresponding products to provide an interesting experience.

▲ Picture from: Android Authority

The regularly updated Mali-G715, like the Cortex-A715, focuses on improving energy efficiency. Compared with the G710, it has a 15% energy efficiency optimization.

In addition, Arm has also updated a GPU that is positioned slightly lower than the Mali-G715, the Mali-G615. The difference is that it has fewer cores (less than 6) and is geared towards mid-range Arm processors.

To say that the Mali-G715 and G615 are regular iterations, the Immortalis-G715 that supports light tracking on the hardware is a big upgrade, and has better computing performance for 3D images.

The Immortalis series of GPUs are similar to Cortex-X super cores, more like Arm prepared for desktop processors, and it seems that they are ready to start working on Arm PCs.

Chip makers are starting to target the Arm PC field

In addition to Qualcomm, Samsung Exynos and MediaTek will directly use Arm's public CPU and GPU architecture, but the core frequency and process are different.

▲ The addition of AMD did not save the Exynos 2200

After Samsung's previous generation Exynos 2200 began to cooperate with AMD to launch the Xclipse GPU, currently only MediaTek is developing the Dimensity series according to the Arm public version architecture in the mainstream SoC market.

Similar to Samsung Exynos finding AMD to cooperate, Qualcomm directly acquired NUVIA, a company with its own design of Arm chip architecture, the purpose is to start its own Arm architecture chip.

Qualcomm has also prepared a chip code-named "Hamoa" internally, and the initial target is the Apple M chip. Its CPU will introduce a self-designed IP architecture core, rather than the Arm public version architecture.

Many analysts also said that Qualcomm also plans to introduce the CPU core of its own design IP into the Snapdragon chip, of course, the premise is that the performance is good enough.

▲ Apple chips have been ahead of Qualcomm, MediaTek, Samsung

The reason why it does not switch directly like the Apple A chip is actually because the market share of the Snapdragon chip is too large and the risk is too high.

Today, Arm's public version architecture is more like a foundation, getting closer and closer to giving manufacturers a reference.

▲ Apple chips are synonymous with "differentiation"

Capable chip manufacturers are also gradually starting to design their own core architecture, in order to have stronger differentiation like Apple's A and M chips, so as to obtain more orders.

The Arm public version architecture has also begun to focus on the creation of the Arm ecosystem. The peak performance of the ultra-large-core Cortex-X and Immortalis series of GPUs is improved for the layout of the Arm PC processor, and the redesign of the Cortex-A510 has better performance. Backward compatibility.

The smartphone market has transformed into a stock market. Whether it is a chip manufacturer such as Qualcomm or a core IP provider such as Arm, the huge potential Arm PC field may be their next growth point.

#Welcome to pay attention to the official WeChat account of Aifaner: Aifaner (WeChat: ifanr), more exciting content will be brought to you as soon as possible.

Love Faner | Original link · View comments · Sina Weibo