AMD continues to study hybrid CPU architectures
AMD filed a patent for hybrid processors featuring heterogeneous general-purpose cores as far back as 2017, but the company does not seem to have plans to bring such CPUs to the market before operating systems for PCs can use them properly. Joe Macri, Vice President and Product Chief Technology Officer at AMD, sees how different types of cores could bring advantages to PCs, but not before OSes get an appropriate scheduler and memory allocator (via PC Gamer).
Arm’s Big.Little heterogeneous multi-core CPU technology, which uses high-performance ‘big’ cores for demanding tasks and low-power ‘little’ cores for background tasks, has proven to be quite useful on Google Android-based smartphones. Eventually something similar was introduced by Apple for its iOS-powered devices. Intel unveiled its first codenamed Lakefield hybrid processor for ultra-light and innovative form factors earlier this year and plans to bring its hybrid concept to higher-performance PCs in 2021 with Alder Lake CPUs.
AMD says that it has been studying hybrid CPU architectures for a while, but does not have any immediate plans it could share.
“We’ve been studying Big.Little,” said Joe Macri, during an engineering roundtable, reports PC Gamer. “It’s been 15+ years, so this is not a new concept in any way shape or form. We continue to study it, we continue to look at it. We’re not going to talk about whether we’ll do it or not, but I’m going to talk about some of the challenges of it and around what you really want to do with it. Is the goal power efficiency? Is the goal more performance? Is the goal just marketing, ‘I want more core count’, regardless of what it does for the other two variables? […] We’re not going to do it just to have a bigger number.”
Building a CPU with different types of general-purpose cores and a system architecture around it are just two parts of a multi-part equation. Operating systems need to learn how to properly schedule tasks and allocate memory for different types of cores in a bid to enhance user experience noticeably. Furthermore, actual programs have to seamlessly switch from one type of cores to another.
“Over time I think there will be a point when we are going to need ‘little,’ and it will be a point in time when the OS has the right attributes, the right capabilities in its scheduler and memory allocator, we’ll have the right memory subsystem,” said Macri. ‘We’ll be able to give you not just a little bit better experience, but a much better experience. If we can’t get that experience change to be noticeable, why do it?”
AMD knows a lot about the importance of OS schedulers. Back in the day the company’s FX-series processors featuring the Bulldozer microarchitecture could not hit performance numbers AMD anticipated because of Windows’s scheduler, so AMD had to work with the software giant to get everything right. Even an optimized scheduler did not help AMD’s Bulldozer to beat competitors, but it did improve performance and the lesson has certainly been learnt.
“Just driving up the core count with little isn’t going to be that useful until software comes along,” said Macri. “It’s not a simple problem, and I think there’s been at least one company who has got it right. I wouldn’t say all the companies have got it right that have started doing it. What we’re really fighting, for Big.Little with AMD, is that Mike [Papermaster’s] team is doing such a great job with Big, it makes it hard to win with Little.”
When it comes to low power vs. high performance cores, there is always a dilemma about efficiency. Is it more efficient to use low-power cores along with memory and storage for a relatively long time or just use high-performance cores (as well as RAM and storage) for a short period of time and then idle everything? AMD currently has very competitive high-performance microarchitectures that can scale and be tailored both for low-power and high-end applications, so it is perfectly logical for the company to use what it has today. Anyhow, AMD officially does not have any low-power ‘little’ cores in its roadmap just yet.