Dolphin, the GameCube and Wii emulator - Forums

A new article has been published on the Dolphin blog - Temptation of the Apple: Dolphin on macOS M1

Feel free to discuss the article in this thread.

I have recently bought a M1 MacBook Pro and of course I had to try Dolphin on it so I was glad I found the PR and Skylersaleh's kindly provided builds Been keeping up with the ARM builds before they were official and it really runs great with Vulkan (MoltenVK).

So anyway, congrats and thanks to all involved for your great work!

I'm glad to see the great strides Dolphin is making on these new M1 CPUs. Gives me another reason to upgrade my aging Hackintosh system. Probably will wait for the M2 or M3 for even better performance though

I've said this before, but when Acorn designed the first ARM CPU, they weren't aiming for power efficiency as the main goal - it was supposed to compete with the processors in other home computers at the time. The power efficiency was somewhat of a shock - the first time an ARM system booted, they'd not yet connected up the power connectors to the CPU, and instead it was capable of running just off the data lines. That meant it was a really interesting architecture for low-power devices, and that's where the focus has been. It's great that ARM's finally being allowed to stretch its legs again and show what it can do with a more generous power budget.

That said, the 9900K used in the article is hardly the most power-efficient x86-64 processor, so it would have been nice to have a comparison with something contemporary with the M1 aimed at the same market segment instead of something prioritising getting bigger numbers than AMD no matter the cost.

@AnyOldName3

The idea of using the 9900K was to show how the M1 faired against the absolute best case scenario, performance wise. It makes sense to put that in the tests. They are essentially showing that it's only a little bit slower than the absolute fastest x86 processor you can have (that the testers had access to) despite it's efficiency. I thought it got the point across well. They are also limited to what they MayImilae/JMC47 have access to. They aren't going to go out a buy a whole new system just for this test.

I was just saying that it would have been nice to have the extra data, that's all. I'm well aware that real life adds limitations.

The closest like-for-like we had available was my 2018 Intel MacBook Pro. It's not a great comparison as it is a few years old and a higher class machine that hasn't had its tier replaced with M1/2/1X yet but, it's what we have! It's still amusing that this $3000 machine was utterly curbstomped by a $1000 MacBook Air M1. (and probably without the myriad of visual glitches that have plagued this laptop from day one. uuughhh Intel) If you want more like for like testing of intel versus M1, you'll have to go to a proper M1 review.

And yea, the 9900k is basically a how does it fair against a solid gaming machine kind of test, and its a hella fighter! That and based on the data we found, M1 would only need to go up to ~4ghz to beat the 9900k. So whenever the M1X/M2 comes out... nice to have this data point already established, no?

A very interesting read to be sure. I'm not a huge CPU expert (embedded SW Engineer by trade), though it is interesting how we see the pendulum swinging back toward RISC. To be fair though, I wonder if part of the problem is Intel just not making huge developmental strides since the Nehalem/Sandy Bridge days.

Regardless, I see the effect in my job. For almost 15 years, we used primarily PPC chips in our embedded designs (especially the PPC 700 series. But now? We're using ARM for pretty much everything (except for some rad-hard work where there aren't a lot of better options yet).

Also nice to see crossplay is mostly working; I suspect as it gets used you'll find a lot of little things that no one really knew/cared about, but the fact it works at all is a credit to Dolphins design.

As always, keep up the good work. And "really" looking forward to the next dev diary.

ARM CPUs, and especially the ones made by Apple, have been seeing amazing improvements year-over-year to shorten the gap in per-core performance compared to x86 CPUs. It definitely doesn't help that Intel was stuck floundering with minimal year-over-year improvements on 14nm since 2014-15, giving AMD time to catch up in per-core performance with Zen.

Intel did what companies with effective monopolies do, i.e. nothing meaningful for several years. Maybe if AMD had been capable of competing in the period where they weren't, we'd have had x86 processors maintain the gap with ARM, or maybe CISC would have hit a wall by now and we'd live in a RISC-V-powered utopia, or maybe things would be largely the same as they are now, but with somewhat faster desktops available.