Hacker News Reader

Comment by drcode | Hacker News Reader

It seems to me M1-style performance should be easy to achieve for other large companies with access to decent foundries... one would think Samsung could just use a similar ARM core, then study the design of the M1 to come up with a competitive cache design. (but I'm no expert, happy to have someone correct my view with better information)

View on HN

If that were the case Apple couldn’t have maintained a huge performance lead in smartphone CPUs for half a decade.

Excellent point. Also Apple has a unique advantage of hardware + OS all built under their control and carefully optimized.

Regular chip companies can't influence their customers as well as Apple's hardware team influences the OS departments.

They had a lead because they were optimizing for a different target than Qualcomm. Apple was more concerned about single core performance while Qualcomm was focusing on lower cost and efficiency.

1. There was no similar ARM core until recently.

2. Samsung/Qualcomm/etc have lower prices and lower margins than Apple so they can't afford to make chips that are as big as Apple Silicon. For example, the M1 has four big cores but competitors use one big core and three medium.

2. To be fair, A14 is just 2 big cores compared to other smartphone SoCs' 4(3+1) big cores. (but Apple's core design is richer than others)

If it was that easy Exynos wouldn’t already lag the A14 so badly.

It should gain maybe 15% in a smaller process, but has a lot farther to go.

I was curious about the actual performance difference and I don't know what the big name benchmarks are, but a quick Google search showed the centurion benchmark [1] which puts the A14 just 5% ahead of the newest Exynos and Snapdragons. That's not too big of a lead? Last years benchmark against the A13 showed a much bigger gap tho.

[1] https://www.techcenturion.com/smartphone-processors-ranking

The industry standard benchmark for completely different CPUs has been SPEC, going back to the days when every UNIX workstation vendor also had it's own custom CPU and instruction set.

Exynos and Snapdragons aren't even close.

https://www.anandtech.com/show/16252/mac-mini-apple-m1-teste...

The benchmark uses Snapdragon 865, not the newer generation. The Snapdragons also run at mobile phone power budgets.

Fine, here you go: https://www.anandtech.com/show/16463/snapdragon-888-vs-exyno...

That's single threaded performance though. Snapdragons have more cores at the expense of single core performance.

The Snapdragon 888 only has one X1 core.

1x Kryo 680 (ARM Cortex X1-based) Prime core @ 2.84GHz, 1x 1MB L2 cache

3x Kryo 680 (ARM Cortex A78-based) Performance cores @ 2.4GHz, 3x 512KB L2 cache

4x Kryo 680 (ARM Cortex A55-based) Efficiency cores @ 1.8GHz, 4x 128KB L2 cache

Vs 2 big 3.1 GHz and 4 small 1.8 GHz on the A14.

Comparing single core performance is kinda cherry-picking a benchmark where the A14 is supposed to be ahead.

The A55 is an in-order, 2 wide, low performance core from 2017 that was an embarrassing choice to put in a modern SOC a couple of years ago.

Compare it to Apple's "little" Icestorm cores, which are at about the performance level of everyone else's big cores a couple of years ago.

>The Icestorm design is actually a quite major leap for Apple as it sees the introduction of a third integer ALU pipeline, and a full second FP/SIMD pipeline, vastly increasing the execution capabilities of this core. At this point it would be wrong to call it a “small” core anymore as it now essentially matches the big core designs from Arm from a few years ago, being similar in complexity as an A75.

https://www.anandtech.com/show/16192/the-iphone-12-review/2

Reply on news.ycombinator.com