QuaternionsRock

That’s actually not what I was referring to, although the unified memory architecture is certainly more power efficient for mixed-intensive workloads. The cost of transferring to/from dedicated GPU memory is (unsurprisingly) quite large.

Here is a great article on the topic. Basically, x86 spends a comparatively enormous amount of energy ensuring that its strong memory guarantees are not violated, even in cases where such violations would not affect program behavior. As it turns out, the majority of modern multithreaded programs only occasionally rely on these guarantees, and including special (expensive) instructions to provide these guarantees when necessary is still beneficial for performance/efficiency in the long run.

For additional context, the special sauce behind Apple’s Rosetta 2 is that the M family of SoCs actually implement an x86 memory model mode that is selectively enabled when executing dynamically translated multithreaded x86 programs.

It is always quite amusing to see a billion dollar corporation beaten in its own game :)

More information/context, if you’re curious:

Rosetta 2 in particular isn’t full emulation because the API is the same for both architectures - it is only dynamic ISA translation. I expect that Prism will be slightly closer to full emulation; there is simply no way Microsoft will reimplement all of the legacy Windows APIs on ARM.

WINE is a great example of something that is also not a full emulator, but for the opposite reason: it does not perform any ISA translation or hardware emulation, but rather only syscall (API) translation.

Oh yeah, clearly I did not read the article well. Still, it doesn’t mean what you think it does.

First, Yuzu is more of an alternative API implementation than an emulator in this setup. The stock Switch OS and API implementation have been entirely replaced with Linux and the Yuzu implementation of the API. Given recent performance uplifts in the Linux kernel, I’m not surprised that Linux+Yuzu beats the first-party implementation.

Second, the use of the word “emulation” in the above thread is really a misnomer: Rosetta 2, Prism and the like all perform what is called dynamic ISA translation. Yuzu need not perform ISA translation when running on ARM hardware.

So if any developer wants to support modern devices they have to port to that new hardware.

See, you say that, but it doesn’t seem like Rosetta 2 going anywhere any time soon, which means developers aren’t pressured their software to ARM.

This article fails to mention the single biggest differentiator between x86 and ARM: their memory models. Considering the sheer amount of everyday software that is going multithreaded, this is a huge issue, and the reason why ARM drastically outperforms x86 running software like modern web browsers.

Apple controls the whole ecosystem on Macs.

In what sense? The vast majority of macOS software is downloaded/installed from the internet, just like Windows.

I don’t see it working because the Windows APIs are a dozen self-oxidizing dumpster fires scattered into the wind, but that’s a different story.

Yuzu can exhibit superior performance because the Switch is rocking the Tegra X1 from 2015. Yuzu absolutely cannot beat the Switch with contemporary hardware and/or comparable power consumption.

And are you done or not? Make up your mind.

I’m sorry?

Out of curiosity - do you think your opinion will change once on-device (i.e., power efficient) AI becomes the norm?

The capabilities and utility of contemporary LLMs are wildly overstated by many, but the claim that they are completely useless is dubious imo. Nothing they generate can be treated as fact (and shame on those who suggest you do), but I can say with certainty that it has made my life as an indie programmer much easier, and I know I’m not alone in that.

Sports are much older than the concept of a “ruling class”.

The argument that they are used as a distraction in capitalist societies does not detract from how central they are to the human condition. Hell, games aren’t even specific to humans

Hah, cool fantasy bro. GPT-9’s first output was

As an AI, I cannot predict whether humans can solve climate change. Is there anything else you would like help with?

Alignmentmaxxed