Whats the purpose of E-Cores in a CPU

[devc1]

I understand that the efficient cores use less energy, but this seems useless to me. Because we we (the OS) can just reduce cpu clock and power consumption dynamically using ACPI ?

[iansjack]

Because you want low-power and high-power processes to run at the same time.

[devc1]

But I think we can manage the power consumption and clock rates of individual cores at the same time ?

[sounds]

I'll reference an article from digitaltrends.com even though I think the way it bags on AMD is both inaccurate and shortsighted:

a hybrid CPU combines performant (P) cores and efficient (E) cores onto a single processor. This design — known as big.LITTLE — was pioneered by chip designer ARM, and you can find it in nearly all mobile devices available today. Apple brought that design to laptops and desktops, and now Intel is following suit.

...

Intel sees the writing on the walls. The company hasn’t been shy about pointing out Apple as its true competitor in the future, not AMD. Meanwhile, AMD continues to stick with architectures that focus on fast cores and a lot of them instead of focusing on a hybrid approach.

- https://www.digitaltrends.com/computing ... lder-lake/

You're correct that even for a performance core, lowering the clock and voltage gets you a disproportionate win for power draw. As the big CPU manufacturers get more competitive, they're looking even deeper for an edge, especially for mobile devices where performance is relegated to second priority under low power usage.

When a performance core is designed, it's given extra logic to run multiple instructions per clock. The IPC score achieved by a modern P-core when it has optimized code gets up to about 5 for x86 and about 7 for Apple arm. But here's the key - that extra physical hardware can't all be shut off when the core is running at low clocks and low voltage.

So the big CPU manufacturers have started adding little cores (E-cores) that are physically unable to chew through instructions at the same pace. But the E-cores have every trick in the book when it comes to efficiency:

1. Smaller instruction decoder
2. Fewer instruction-level parallelism "ports"
3. Fewer entries in the reorder buffer
4. Fewer cache prefetchers
5. Smaller cache
6. Transistor layouts are optimized for low power, at the expense of being slightly larger on-chip
7. Transistor chemistry and geometry are optimized for low power, at the expense of not being able to run at the highest clock speeds
8. Intel even left out the most expensive instructions (such as AVX-512) which caused some problems because a process could be running on a P-core, where it checked first and then decided to enable the AVX-512 code path. But then it got rescheduled by the OS on an E-core, where it immediately crashed.

If you're thinking a hybrid cpu with P-cores and E-cores is a big pain, you're not wrong! It seems we're going to be stuck with them though.

(Oh, and AMD has been really smart to adopt a wait-and-see approach here. AMD's chiplet innovations are paying off in spades while Intel wades through the morass of coming up with a decent hybrid design.)

[Octocontrabass]

Intel's E-cores aren't very power-efficient compared to downclocked P-cores. They're really optimized for cost: Intel can fit four E-cores in the space of one P-core, so you can get more threads to chew on parallel tasks without sacrificing peak single-thread performance. They're banking on most consumer tasks needing either a small number of very fast threads or lots of not-so-fast threads.

Future E-cores might be better in the power efficiency department, of course.

[Schol-R-LEA]

One thing to keep in mind is that most of the tasks which people use their computers for - web surfing, simple spreadsheets, word processing, various sort of background tasks such as music or updates - have minimal processing overhead by the standards of modern processors. Since these are often running simultaneously, it makes more sense to focus on providing additional processors with a smaller die footprint, rather than fewer full-power cores in the same die space. Or at least that's the theory.

It makes a lot of sense with mobile processors, where you rarely need the high-performance ('big') cores, but are constantly running low-demand processes which the 'little' cores excel at. It's less obvious how useful it is on a desktop or full-sized laptop system, but again, for the majority of users - especially business users - it still mostly holds true. Unless you are gaming, editing audio and video, image rendering, performing major number-crunching, compiling code, etc., the majority of the computer's time is spent either idle, or performing background tasks which as a rule can be done with the efficiency cores with no loss in performance. Your typical PC user will rarely need the performance cores at all.

There is a reason why the majority of the tech-related blogs, webzines, and YouTube channels focus on gaming - it is the most common place where significant demands are placed on modern computers. Video games have always been at the forefront of the industry's interests simply because they are the one thing most people can relate to which puts the system through its paces. The fact that most people never play games of that type, or only do so infrequently when they do, is irrelevant - game performance is a flex for the manufacturers.

[JoeBox9]

Aren't they cheaper to put in there?