It’s already very good, and improving practically every week.
Right now, you can play The Witcher 3 at 720p, medium settings, at a stable 30 fps on Snapdragon 8 Gen 2 handhelds. Or the same settings in Cyberpunk 2077 on Snapdragon 8 Elite handhelds.
Which is basically what it runs like on a Steam Deck, except you get better battery life, a smaller and lighter device, and the device itself is anywhere between 2 to 5 times cheaper.
Now, keep in mind 8 Elite aren’t even the best supported SoCs (8 Gen 2 is among them though), and we’re still at least a couple years away from any new Valve handheld.
And Valve has just released an ARM device with an ARM build of SteamOS.
At this point I’d say it almost doesn’t make any sense to build a Steam Deck successor on x86.
I’m not very concerned about performance. I’d be pleased with a lighter, quiter, less expensive and more battery-efficient device that reaches the same performance (something ARM promises) but I expect a very large chunk of the available PC library (Steam and non-Steam) to run on it, through the emulation layer, and through the OS compatibility layer.
Cyberpunk is almost 6 years old now and Witcher 3 released over 11 years ago (recent W3 remaster and today’s CP 2.5 update notwithstanding). I really hope those improvements keep pace and the Deck 2 can do better than medium settings 30 FPS on these.
Steam Deck 2 will do native 4k 120 fps on Witcher 3 with RayTracing enabled and max settings. I am Nostradamus and this will happen in the next 2 years.
The Steam Deck 2 will have 2 CPUs, one ARM chip and one x86 for backwards compatibility, in case there are problems running specific games on ARM. I am Nostradamus and this will happen in the next 2 years.
x86 is a standardized, essentially open platform. ARM isn’t. I like that the Steam Deck is just a portable PC that you can install any OS you want on, and I hope the Steam Deck 2 is the same.
Yeah, x86 is so open only Intel and AMD can legally design and manufacture chips on that architecture. Aside from a Chinese company that slowly reverse engineers it (currently it’s on Intel 8th Gen level).
As opposed to ARM, which anyone can design, modify, or just use existing designs for a comparably small fee. And anyone can manufacture ARM chips, too, given they have the necessary resources. In practice it most likely means hiring a company to manufacture chips.
In the entire history of electronics there was never any technology that was immediately standardized. It always takes time. If we wanted to stick to already existing standards, we’d still be using things like the i386 or older.
The main problem with ARM is lack of a UEFI equivalent, which is currently becoming available (ARM SystemReady).
It’s up to individual vendors whether or not to participate. So far adoption is pretty much ubiquitous in the server space, but not nearly as spread in consumer devices. I’d say this is because a lot of companies aren’t interested in giving people the ability to change software on their phones or tablets (with proposed or already present legislation all around the world to help limit this as well).
An example of a laptop certified for ARM SystemReady is the ThinkPad X13s (Snapdragon 8cx gen 3). All you need to do to install a custom OS is enable Linux boot in the bios, and boot up any installation media. Drivers are in the kernel; no need for a custom iso or anything.
It only can on the high end. There are no x86 CPUs that’d allow you to play lightweight games while the entire system is pulling just 3 watts.
Not to mention the difference in heat generation. Even at comparable power use, the Steam Deck is loud and getting hot, while something like an AYN Thor is silent and barely warm to the touch.
Which is why every single mobile and low power device in existence is based on either ARM or RISC-V.
Another problem is that x86 is needlessly expensive, because Intel and AMD have a duopoly. You either buy at their insane prices, or you don’t buy at all.
It’s an outdated ISA that is still only used, because the two giants abuse their market power. Even if x86 could match ARM 1-to-1, there’d be no reason to stick with x86, just for the price difference alone.
If it can, then I’d like to see it happen. Even Intel’s Celerons use more power than average ARM SOCs, while offering performance straight out of the 1990s.
CPUs from the N100 - Core 3 N355 lineup are more reasonable, but still fall behind by a large margin, and throttle easily (more on the N355 side). Whereas AMD is still struggling to produce anything that isn’t a mobile heater.
What x86 chip offers both performance and power-efficiency comparable to, say, a Snapdragon 8 Gen 2?
Isa plays little role in power efficiency or heat.
This is completely incorrect, both in theory, and in practice, as we can clearly see. x86, as a CISC ISA, has been specifically designed without any power-efficiency considerations.
As opposed to ARM and RISC-V, which pretty much exist specifically to be power efficient, as is very common for RISC ISAs.
This is, quite literally, what distinguishes these two ISAs, and what made both of them happen in the first place.
If arm could match 1:1 it still wouldnt be viable. It’d need to be at least 10% more preformant for the same spec range.
Again, I’d like to see some reasoning or examples behind this.
Why would ARM need to offer better performance to be viable? It’s already incomparably cheaper, more open, more power-efficient, and requires less cooling at comparable performance and power drain levels.
It is already superior in every category. What you’re asking for is not only unreasonable, but also pointless and unnecessary.
Here’s hoping this is completely wrong, and any potential Steam Deck 2 won’t be on x86…
Is emulation so good already that compatibility isn’t a concern anymore?
Hmm, kind of?
It’s already very good, and improving practically every week.
Right now, you can play The Witcher 3 at 720p, medium settings, at a stable 30 fps on Snapdragon 8 Gen 2 handhelds. Or the same settings in Cyberpunk 2077 on Snapdragon 8 Elite handhelds.
Which is basically what it runs like on a Steam Deck, except you get better battery life, a smaller and lighter device, and the device itself is anywhere between 2 to 5 times cheaper.
Now, keep in mind 8 Elite aren’t even the best supported SoCs (8 Gen 2 is among them though), and we’re still at least a couple years away from any new Valve handheld.
And Valve has just released an ARM device with an ARM build of SteamOS.
At this point I’d say it almost doesn’t make any sense to build a Steam Deck successor on x86.
I’m not very concerned about performance. I’d be pleased with a lighter, quiter, less expensive and more battery-efficient device that reaches the same performance (something ARM promises) but I expect a very large chunk of the available PC library (Steam and non-Steam) to run on it, through the emulation layer, and through the OS compatibility layer.
Cyberpunk is almost 6 years old now and Witcher 3 released over 11 years ago (recent W3 remaster and today’s CP 2.5 update notwithstanding). I really hope those improvements keep pace and the Deck 2 can do better than medium settings 30 FPS on these.
Steam Deck 2 will do native 4k 120 fps on Witcher 3 with RayTracing enabled and max settings. I am Nostradamus and this will happen in the next 2 years.
Speaking from my experiences, no, not yet.
On which device?
Ayn Thor, Steamframe, and a few Android phones.
The Steam Deck 2 will have 2 CPUs, one ARM chip and one x86 for backwards compatibility, in case there are problems running specific games on ARM. I am Nostradamus and this will happen in the next 2 years.
hell yeah
x86 is a standardized, essentially open platform. ARM isn’t. I like that the Steam Deck is just a portable PC that you can install any OS you want on, and I hope the Steam Deck 2 is the same.
Yeah, x86 is so open only Intel and AMD can legally design and manufacture chips on that architecture. Aside from a Chinese company that slowly reverse engineers it (currently it’s on Intel 8th Gen level).
As opposed to ARM, which anyone can design, modify, or just use existing designs for a comparably small fee. And anyone can manufacture ARM chips, too, given they have the necessary resources. In practice it most likely means hiring a company to manufacture chips.
In the entire history of electronics there was never any technology that was immediately standardized. It always takes time. If we wanted to stick to already existing standards, we’d still be using things like the i386 or older.
The main problem with ARM is lack of a UEFI equivalent, which is currently becoming available (ARM SystemReady).
It’s up to individual vendors whether or not to participate. So far adoption is pretty much ubiquitous in the server space, but not nearly as spread in consumer devices. I’d say this is because a lot of companies aren’t interested in giving people the ability to change software on their phones or tablets (with proposed or already present legislation all around the world to help limit this as well).
An example of a laptop certified for ARM SystemReady is the ThinkPad X13s (Snapdragon 8cx gen 3). All you need to do to install a custom OS is enable Linux boot in the bios, and boot up any installation media. Drivers are in the kernel; no need for a custom iso or anything.
Being a standardized, essentially open platform means you don’t have to built it in silicone anymore. Frame runs x86 great on ARM.
Why not x86? It can basically match arm for perf per watt and its perf per dollar is better.
It only can on the high end. There are no x86 CPUs that’d allow you to play lightweight games while the entire system is pulling just 3 watts.
Not to mention the difference in heat generation. Even at comparable power use, the Steam Deck is loud and getting hot, while something like an AYN Thor is silent and barely warm to the touch.
Which is why every single mobile and low power device in existence is based on either ARM or RISC-V.
Another problem is that x86 is needlessly expensive, because Intel and AMD have a duopoly. You either buy at their insane prices, or you don’t buy at all.
It’s an outdated ISA that is still only used, because the two giants abuse their market power. Even if x86 could match ARM 1-to-1, there’d be no reason to stick with x86, just for the price difference alone.
It can match on all ends. Isa plays little role in power efficiency or heat. X86 would have no overhead on games.
If arm could match 1:1 it still wouldnt be viable. It’d need to be at least 10% more preformant for the same spec range.
If it can, then I’d like to see it happen. Even Intel’s Celerons use more power than average ARM SOCs, while offering performance straight out of the 1990s.
CPUs from the N100 - Core 3 N355 lineup are more reasonable, but still fall behind by a large margin, and throttle easily (more on the N355 side). Whereas AMD is still struggling to produce anything that isn’t a mobile heater.
What x86 chip offers both performance and power-efficiency comparable to, say, a Snapdragon 8 Gen 2?
This is completely incorrect, both in theory, and in practice, as we can clearly see. x86, as a CISC ISA, has been specifically designed without any power-efficiency considerations.
As opposed to ARM and RISC-V, which pretty much exist specifically to be power efficient, as is very common for RISC ISAs.
This is, quite literally, what distinguishes these two ISAs, and what made both of them happen in the first place.
Again, I’d like to see some reasoning or examples behind this.
Why would ARM need to offer better performance to be viable? It’s already incomparably cheaper, more open, more power-efficient, and requires less cooling at comparable performance and power drain levels.
It is already superior in every category. What you’re asking for is not only unreasonable, but also pointless and unnecessary.
Yeah… Didn’t a Valve engineer say they’re looking to make it ARM?
Nothing I’ve seen suggest moving away from arm?