ARM releases Coretx-X3 /A715/A510 CPU: up to 12 cores fully moving towards 64 bits

ARM releases Coretx-X3 /A715/A510 CPU: up to 12 cores fully moving towards 64 bits.

After launching the ARMv9 instruction set for the next ten years in March last year, ARM launched the first-generation products based on ARMv9 in May, including CPUs such as Cortex-X2 and A710.

On June 28, ARM launched Cortex-X3, A715 and A510 Refresh. and other new generation ARMv9 products.

This generation of Cortex processors is oriented towards platforms such as smartphones, tablets and notebooks, so there are more things released.

There is also a new DSU-110 large and small core scheduling architecture, which can achieve up to 8+4+0 total 12 cores.

First of all, let’s look at the overall part of the CPU. Cortex-X3 and A715 are upgraded versions of Cortex-X2 and A710 respectively. Not only the peak performance is increased by 25%, but the hybrid performance and energy efficiency are also increased by 20%.

It is worth mentioning that Cortex-X3 and A715 have already abandoned the aarch32 instruction set, that is, abandoning 32 bits and turning to 64-bit architecture.

What about the original 32 bits? ARM also launched the A510 refresh, which can also be called the A510 v2 version, which not only improves performance and energy efficiency, but also supports up to 12-core architecture, and is optionally compatible with the aarch32 instruction set, but it is limited to specific options.

In general, in the architecture of the Cortex-X3, A715 and A510 v2 generation, ARM’s determination to fully switch to 64-bit is certain, although 32-bit is not absolutely eliminated, leaving less and less 32-bit compatible fields.

Let’s take a look at the specific changes of the three CPU architectures of Cortex-X3, A715 and A510 v2.

As a super-large core, Cortex-X3 is the main performance direction. The architecture of this generation has been improved a lot.

The number of instructions per cycle of the decoder has been increased from 5 to 6, the out-of-order execution window has been increased from 288 to 320, and the integer ALU unit has been increased from 4. raised to 6 and so on.

In other respects, the L2 cache capacity of the Cortex-X3 has also been increased from 512KB to 1MB.

The final change is that the performance has greatly increased.

Compared with the previous generation Android flagship, the performance has increased by 25%, and compared with the performance of the notebook computer, the performance has increased by 34% .

The official performance comparison of ARM is a bit messy. The performance of this generation alone has improved by 22% and 28%. After a number of claims, there is an 11% increase in IPC performance.

Next, let’s talk about the A715’s large-core architecture. It does not pursue absolute performance, but pays more attention to energy efficiency.

Under the same performance, it saves 20% energy consumption compared with A710, and under the same power consumption, the performance is improved by 5% compared with A710.

In addition, due to the abandonment of the aarch32 instruction set, the area efficiency of the A715 core has greatly increased, and only 1/4 of the area of the A710 is required.

Although the overall performance has not improved much, the energy efficiency and area efficiency have greatly increased.

Finally, there is the architecture of the A510 small core.

Now the A510 v2 has not changed much compared to the A510 in 2020.

The focus is still on optimizing energy efficiency.

The same performance power consumption is reduced by 5% and the frequency is increased by 5%.

Therefore, the previous 1.8GHz The frequency can now be boosted to around 1.9GHz.

There is also DSU-110, this is a new generation of core interconnect technology, the core collocation is more flexible, in addition to 1+3+4, there are 1+4+4, 2+2+4 and 8+4+0, The last configuration is for high-performance platforms such as notebooks.

In other respects, cores such as Cortex-X3, A715 and A510 v2 are also optimized for more advanced processes, such as Samsung TSMC’s 5nm and 4nm processes.

ARM also provides developers with a convenient development platform and tool VFP, which can better ground simulation test, etc.

Finally, CPU architectures such as Cortex-X3, A715, and A510 v2 will be launched as soon as the end of this year.

If nothing else, the Snapdragon 8 G2 will be replaced by a new architecture at the end of the year.