Hacker News Reader

Comment by Someone1234 | Hacker News Reader

I feel like Steam Deck is very successful; the next big milestone, which may be a doozy, is migrating off of x86-64 to ARM64. I understand why Steam Deck 1.0 is x86-64 since something akin to Rosetta or WOW64 is likely as much work again as it was to get the initial Steam Deck off the ground.

However, battery life for a mobile console cannot be ignored. The Steam Deck can last as little as under 2 hrs depending on workload. A switch to ARM64 (e.g. ARMv9.0) would likely over double the battery life at the bottom and top end.

As a side benefit, because high-end ARM64 SoCs are so common in phones, Chromebooks, tablets, steaming boxes, et al., they may be able to bring the prices down, further increasing their competitiveness.

PS - No doubt I'll get a bunch of replies with "But use this One Simple Trick™ to get longer battery life" which, while true, also applies to ARM64.

View on HN

Why do you think ARM64 would improve battery life so much? The Steam Deck can draw up to 20W, most of that due to the GPU. Apple's M1 laptops have a higher TDP, larger battery, and a similar 2-4 hour battery life on the most intense games.

If you want more battery life, Steam Deck has a bunch of controls to throttle CPU/GPU performance, including limiting the framerate.

This will only happen if Microsoft also pushes game developers towards ARM64.

You're talking about Rosetta but Rosetta is a transition period. If the Steam Deck is the only device to run on ARM64, and game developers keep providing only x86 builds, the drawbacks of ARM will outweight the benefits.

The issue is the trade-off with performance though, that'd make it even more of a niche device.

And almost all games are x86 only so either you need to do that opcode level conversion too (on top of the Windows / DirectX APIs), or limit yourself to only ARM games (realistically mostly FOSS games). Apple has done it well for M1/2 to be fair, but it's far from easy.

So you could get much better battery life but it'd be very niche.

> A switch to ARM64 (e.g. ARMv9.0) would likely over double the battery life at the bottom and top end.

Only if the games are compiled for ARM64, and even then the performance improvements over x86 are not guaranteed compared to just switching to a better process node. A big attraction of the Steam Deck is the massive back catalog of already existing games on Steam, all of which are compiled for x86. If your ARM CPU is spending its time emulating x86, you're not going to get much of an improvement in battery life.

A Steam Deck v2, still running on an AMD CPU, but on TSMC 5 or 3nm, is more likely than one with an ARM64 processor.

We've seen with the M range of SoCs that even with the emulation overhead, you can gain battery life even accounting for that. Plus the OS itself would be natively compiled, so that's always an efficiency gain.

But how much of that is because they were ARM vs being the first desktop-class CPUs on a 5nm process?

Because ARM64 is significantly more R&D dollars spent on it than x86-64. That's a trend that is going to continue.

So the question is do you keep clinging to a slowly dying platform that is falling further and further behind, or do you go to the commodity platform that is going from strength to strength?

The Steam Deck will wind up on ARM64 eventually, it is a matter of "when?" not "if?"

They should just skip the battery altogether imo..

Reply on news.ycombinator.com