The “Mongoose” Inside Samsung Galaxy S7 Gives It Thinking Powers

Share on twitter
Tweet
Share on whatsapp
WhatsApp
Share on facebook
Share
samsung m1 mangoose

samsung m1 mangooseShort Bytes: Samsung’s flagship Galaxy S7 packs an octa-core processor which includes four home-baked cores made by Samsung, dubbed as M1 or Mongoose. These cores have enabled better branch prediction rate to the mobile platform by introducing neural network-based branch predictors.

Galaxy S7 is Samsung’s flagship for the year 2016. The Exynos 8890-packed smartphone has something more than just the processing power. Its Galaxy’s thinking ability.

The big.LITTLE octa-core SoC that powers the Galaxy S7 is designed using the 14nm FinFET process. Four of the eight cores are Cortex-A53 running at 1.6GHz and the rest are M1 cores running at 2.3GHz which Samsung calls as “custom core”. These custom cores are based on the Mongoose or M1 architecture which is an in-home Samsung product designed over a timespan of three years.

What’s so special about the Mongoose?

The custom core is the thing that gives thinking abilities to the S7’s CPU. Its branch predictor is based on a neural network. The prime job of the branch predictor is to make an assumption of what part of the code (branch) would execute next. Adding a neural network to the branch predictor would allow it to make a better idea of what code an app would be going for.

This helps in maintaining the flow of the instruction pipeline by putting up the instructions in advance. It saves you the effort of cleaning the pipeline if some different instruction is to be executed than the predicted one (in the case of the normal branch predictor).

The neural network said in the case of the Mongoose is an ANN (Artificial Neural Network). These are designed to resemble the functioning of the biological neural network of the brain. They are created using a number of parallelly arranged processors separated in tiers. The networks thus created are adaptive but require feeding of data initially. For example, to detect a dog in an image, the ANN should be fed with various images of dogs and also the images which don’t have a dog, say, a cat. Their prediction abilities become better over time.

Putting neural networks inside processors is not something out of the blue. Long-time silicon veterans like Intel and AMD have been putting similar and better things in their chips. Recently launched AMD Zen does have an improved branch predictor with a hashed perceptron predictor which was confirmed by Mika Clark, designer of Zen architecture. “Maybe I should have called it a neural net,” said Mark.

Samsung presented their new microarchitecture M1 a.k.a. Mongoose on Monday at the Hot Chips conference in Cupertino, California.

— via The Register

If you have something to add, tell us in the comments below.

Also Read: Briana — Use This Free Software And Give Artificial Intelligence Powers To Your PC

Aditya Tiwari

Aditya Tiwari

Aditya likes to cover topics related to Microsoft, Windows 10, Apple Watch, and interesting gadgets. But when he is not working, you can find him binge-watching random videos on YouTube (after he has wasted an hour on Netflix trying to find a good show). Reach out at [email protected]

New on Fossbytes

Scroll to Top