Arm Launches the Cortex-X4, A720 and A520, Immortalis-G715 GPU

[erek]

Pretty cool

"The new Immortalis-G720 is Arm's most performant and efficient GPU ever, as we continue to push the boundaries of visual computing. It delivers 15 percent performance and efficiency improvements over the previous generation, as well as a 40 percent uplift in system-level efficiency, leading to higher quality graphics for more immersive visual experiences.

Alongside the Immortalis-G720, we add to our world-class portfolio of GPUs with the new Arm Mali-G720 and Mali-G620. Through these new Mali GPUs, we are committed to bringing premium graphics features to a wider market of consumer devices quicker.

CPU performance leadership for intelligent AI
As part of TCS23, we are demonstrating our ongoing commitment to CPU performance leadership. We are announcing a new Armv9 Cortex CPU compute cluster, which for the third consecutive year delivers double-digit performance gains alongside significant efficiency improvements.

A vital part of this high-performance cluster is the new Arm Cortex-X4, our fourth-generation Cortex-X core, which pushes the limit of performance on flagship smartphones. It is the fastest CPU that we have ever built, bringing 15 percent more performance compared to the Cortex-X3. Meanwhile, the new power efficient microarchitecture consumes 40 percent less power than Cortex-X3 on the same process. These performance and efficiency gains bring the on-device experiences, like UI responsiveness and application launch time, to the next level and enable next-gen AI and ML-based applications."

Source: https://www.techpowerup.com/309290/arm-launches-the-cortex-x4-a720-and-a520-immortalis-g715-gpu

 

[1_rick]

Down in the comments someone mentioned how Apple's low-power cores are so much better than the standard Arm ones and wondered if Apple is somehow keeping Arm from making major improvements.

Edit: whoops, the comments of the Anandtech article on the same subject.

 

[Lakados]

Down in the comments someone mentioned how Apple's low-power cores are so much better than the standard Arm ones and wondered if Apple is somehow keeping Arm from making major improvements.

Edit: whoops, the comments of the Anandtech article on the same subject.

Nah Apple can’t do that to ARM any more than AMD can do that to Intel. Apple just spent an absurd amount of money to optimize designs and layouts. Apples designs are also optimized to a degree where they are completely non functional off of a TSMC process (AMD is in the same boat there) but ARM proper needs theirs to be generic so Intel, Samsung, Global, or TSMC could make it. The lack of specificity in their designs doesn’t let them reach the same heights as Apple can because Apple can run much tighter tolerances and is designed to push what ever TSMC node they are on to its fullest.

 

[1_rick]

Nah Apple can’t do that to ARM any more than AMD can do that to Intel. Apple just spent an absurd amount of money to optimize designs and layouts. Apples designs are also optimized to a degree where they are completely non functional off of a TSMC process (AMD is in the same boat there) but ARM proper needs theirs to be generic so Intel, Samsung, Global, or TSMC could make it. The lack of specificity in their designs doesn’t let them reach the same heights as Apple can because Apple can run much tighter tolerances and is designed to push what ever TSMC node they are on to its fullest.

I re-read the comments and someone replied, saying that Apple's little cores are much larger than Arm's and says that's a lot of it, FWIW.

 

[Lakados]

I re-read the comments and someone replied, saying that Apple's little cores are much larger than Arm's and says that's a lot of it, FWIW.

In terms of physical size, it isn't much bigger, but TSMC has some very specific layering techniques that lets Apple get much more cache and far higher densities. So technically it is "larger".