Arm introduces second-generation Armv9 processors, plus a ray-tracing GPU

After debuting the brand-new Armv9 architecture last year, Arm has just announced the second generation of Armv9 CPUs, spearheaded by the new Cortex-X3 CPU flagship, alongside the Cortex-A715 and a revised Cortex-A510. Additionally, the company introduced its first-ever flagship GPU, the Immortalis-G715, complete with support for hardware-level ray tracing. These new components also power Arm’s new Total Compute Solutions.

All of these will be found in products launching in 2023, so it’ll be awhile before you see these benefits in a consumer product.

Arm Cortex-X3 and Cortex-A715

First off, the new CPUs based on the second generation of Armv9 come with some big performance improvements, which always tends to be the case with Arm. Starting with the Cortex-X3, which is part of the Cortex-X series of custom CPUs, where can chip in to customize the final partners design of the chip. The Cortex-X3 promises a 25% increase in performance over the Cortex-X2 in the latest Android smartphones. The new chip is also more efficient, achieving the same level of performance while using less power, though Arm didn’t dive into the details of that.

Graph depicting the Arm Cortex X3 performance and power consumption compared to the Cortex-X2

Image credit: Android Authority

It also delivers up to 34% higher single-threaded performance than the latest Windows laptops. As noted by Android Authority, that’s compared to the Intel Core i7-1260P, which is a power-hungry 28W processor.

Additionally, the DynamIQ Shared Unit (DSU-110) platform has also been upgraded, and it can now be scaled up to 12 cores and 16MB of L3 cache. This enables more scalability, so Arm processors can be adapted for everything from phones to more high-end laptops.

The Arm Cortex-A715 comes with its fair share of upgrades, but it’s mostly focused on efficiency. Arm touts up to 20% more energy efficiency compared to the previous Cortex-A710 CPU, which should result in better battery life for devices using it. On the performance side, Arm is promising a 5% increase compared to the last generation, and it can actually match the performance of the Cortex-X1 CPU launched in 2020.

Graph showing the performance and consumption of the Cortex-A715 CPU alongside the Cprtex-X3 and Cortex-A510

Finally, the revised Cortex-A510 doesn’t come with any performance changes, but it does manage to save 5% more power than the previous iteration, delivering even better efficiency to prolong battery life. Additionally, this CPU core no longer has any 32-bit support, which was part of Arm’s roadmap to transition fully to 64-bit.

Arm Immortalis-G715, Mali-G715, and Mali-G615: Ray tracing comes to mobile

On the GPU side, Arm wasn’t content with upgrading the existing Mali GPUs. This time, we get a brand-new flagship GPU, the Immortalis-G715. This is the first Arm GPU to support hardware-accelerated ray tracing, which makes it actually feasible to use on mobile. However, the company did introduce two new Mali GPUs, the Mali-G715 and Mali-G615.

If you don’t know what ray tracing is, it’s a more realistic way to generate lighting and shadows in games. This is because the GPU actually calculates the individual light paths in a given scene, simulating the way light propagates in real life. You can see it in action below.

While energy consumption and area are usually big concerns when it comes to enabling ray tracing, Arm says the feature only uses 4% of the shader core area in the Immortalis-A715, yet it delivers a 300% performance increase compared to software-based ray tracing solutions, such as what’s found on the Mali-G710 GPU from last year.

Other improvements are similar across the entire range of new GPUs. Arm touts at 15% performance on the architecture level alone compared to the previous generation. The new GPUs also all support Variable Rate Shading, which can deliver performance improvements by adjusting the shader quality in different parts of the scene, so rendering power is focused on the most important parts of said scene. For example, it allows the GPU to render graphics on a finer granular level in parts of the scene with more action, where you’re focused on what’s changing, but use less detailed rendering on static background elements.

Arm is also touting improvements to the Execution Engine and the fused multiply-add (FMA), resulting in 2x improvements in FMA power, but with only a 27% aread increase. ML performance has also doubled thanks to the addition of the Matrix Multiply instruction. Finally, improvements to the Command Stream Frontend make it faster than the previus generation, with triple the peak triangle output. Arm has also optimized level of detail (LOD) lookups in the Texture Mapper resulting in twice the troughput, and the inclusion of the Arm Fixed Rate Compression technology that debuted in the Mali-G510 helps save bandwidth.

While most of the features are supported across the three GPUs, there are naturally differences in power. The Immortalis-G715 has 10 or more cores, while the Mali-G715 has 7 to 9 cores, and the Mali-G615 has 6 cores or fewer.

Arm Total Compute Solutions

As we mentioned at the top, all of these technologies come together to create the second generation of Arm Total Computer Solutions, or TCS. These are unified solutions that include all of the computing components required to power a device, and they’re available in various tiers to target different types of devices.

For example, there are premium solutions, which may feature a Cortex-X3 cores along with Cortex-A715 and Cortex-A510, plus an Immortalis-G715 GPU, and those are geared towards flagship phones, laptops, and desktops. Then there are the performance solutions, focused on Cortex-A715 and Cortex-A510 cores, combined with the Mali-G715 and Mali-G615, meant for mid-range smartphones, Chromebooks, and TVs. Finally, the efficiency solutions focus on Cortex-A510 GPUs and low-power Mali-G310 GPUs, and they’re meant for low-end phones, smartwatches, AR glasses, and so on.

Different Arm TCS 2022 solutions for different market segments

As an example, a premium solution can deliver up to 28% more gaming performance compared to the first-generation Arm TCS, in addition to reducing DRAM traffic by up to 23% and power consumption by up to 16%. A solution using the new CPU cores and a Mali-G715 GPU can also deliver big machine learning performance improvements, particularly for camera and video features.

The new Arm TCS22 solutions also come with security improvements across three layers: In-process security, secure firmware upgrade, and trusted execution environment (TEE). In-process security now comes with asymmetric Memory Tagging Extension (MTE), courtesy of the new Armv9 CPUs, plus enhanced PAN to protect against JITed code attacks. The TEE has also been strengthened with control flow integrity and memory integrity protection.

All of this will be coming to Arm-based devices in 2023, and it’s looking to be a big year, with the debut of hardware-based ray tracing on mobile in addition to the usual efficiency slow of performance and improvements across the board.

Source: Arm

Leave a Comment