According to YouTuber Dave2D who managed to get their hands on the Lenovo Legion Go S, the SteamOS version destroys Windows in performance and battery life.
That is a bit surprising, because I have used a Legion Go (non S) with both Windows and Bazzite and performance seemed pretty comparable across both. I certainly didn’t notice double the battery life at any point. Maybe I just didn’t bench the same set of games, this seems very specifically to yield best results on CDPR games. Or maybe it’s because these benches are just for the Z2 and not the Z1 Extreme version, and this is very specific to that chip.
It could also be the memory management/config is different on the SteamOS side and some games are getting different amounts of VRAM across OSs? How do these stack up to Bazzite on the same hardware? Is there an advantage to brand name SteamOS?
I want to see more benchmarks from more people with more configs. Everybody in the tech industry is busy fawning over overengineered fans over in Computex and this actually interesting release isn’t getting the right amount of coverage.
Seeing this with Bazzite, Garuda, “vanilla” distros like Mint, Arch, Manjaro, etc. would be really interesting, I agree.
My amateur guess from the outside would be, that SteamOS is perhaps stripped down in a way, that “normal” Linux background stuff only gets booted up when switching into Desktop mode, which would explain the massive improvements for battery life. But that is a guess, cannot be sure about that at all.
On the Deck itself there are APU customizations at play, but not here. I don’t know that the underlying OSs are fundamentally different. I know one is arch and the other is Fedora, but they’re both immutable distros and should mostly be running the same things when launching in game mode.
For battery life it could just be a configuration difference in how the benchmarks were run on both OSs, or even down to the manufacturer software. Benchmarking hardware is hard, what can I say. I can say Dave2D isn’t great at it, which I suppose is not the point of his channel, but I certainly wish some of the more technical channels weren’t distracted right now, because there’s an interesting three-way comp to be had here and some digging into interesting things.
I’d like to see someone else reproduce the original test results. Seen the same set of data posted multiple times and honestly the numbers look like major outlier and either different settings or something wrong on wibdows.
That may be legitimate if the Windows settings are the factory settings. That’s why I was pointing at memory management, because if you have a 32 GB device and you’re assigning 3GB to VRam while the SteamOS version does something different things may get funky results in some games, especially running at higher resolutions and so on.
So it’s entirely possible that the out-of-the-box setup of these machines on Windows and SteamOS are legitimately that different but that a better Windows config would mitigate it, which is still bad for something sold with a preinstalled Windows image, for the record. Or maybe the overhead of Win11 is just that big, I don’t know. Would certainly love to see someone look into it.
I can tell you that bumping the default VRAM allocation on Windows handhelds has taken some AAA games from unplayable to quite solid in my experimentation, but I’m not gonna sit here swapping OSs and games back and forth for benchmarks. At least not for free.
That is a bit surprising, because I have used a Legion Go (non S) with both Windows and Bazzite and performance seemed pretty comparable across both. I certainly didn’t notice double the battery life at any point. Maybe I just didn’t bench the same set of games, this seems very specifically to yield best results on CDPR games. Or maybe it’s because these benches are just for the Z2 and not the Z1 Extreme version, and this is very specific to that chip.
It could also be the memory management/config is different on the SteamOS side and some games are getting different amounts of VRAM across OSs? How do these stack up to Bazzite on the same hardware? Is there an advantage to brand name SteamOS?
I want to see more benchmarks from more people with more configs. Everybody in the tech industry is busy fawning over overengineered fans over in Computex and this actually interesting release isn’t getting the right amount of coverage.
Seeing this with Bazzite, Garuda, “vanilla” distros like Mint, Arch, Manjaro, etc. would be really interesting, I agree.
My amateur guess from the outside would be, that SteamOS is perhaps stripped down in a way, that “normal” Linux background stuff only gets booted up when switching into Desktop mode, which would explain the massive improvements for battery life. But that is a guess, cannot be sure about that at all.
On the Deck itself there are APU customizations at play, but not here. I don’t know that the underlying OSs are fundamentally different. I know one is arch and the other is Fedora, but they’re both immutable distros and should mostly be running the same things when launching in game mode.
For battery life it could just be a configuration difference in how the benchmarks were run on both OSs, or even down to the manufacturer software. Benchmarking hardware is hard, what can I say. I can say Dave2D isn’t great at it, which I suppose is not the point of his channel, but I certainly wish some of the more technical channels weren’t distracted right now, because there’s an interesting three-way comp to be had here and some digging into interesting things.
I’d like to see someone else reproduce the original test results. Seen the same set of data posted multiple times and honestly the numbers look like major outlier and either different settings or something wrong on wibdows.
That may be legitimate if the Windows settings are the factory settings. That’s why I was pointing at memory management, because if you have a 32 GB device and you’re assigning 3GB to VRam while the SteamOS version does something different things may get funky results in some games, especially running at higher resolutions and so on.
So it’s entirely possible that the out-of-the-box setup of these machines on Windows and SteamOS are legitimately that different but that a better Windows config would mitigate it, which is still bad for something sold with a preinstalled Windows image, for the record. Or maybe the overhead of Win11 is just that big, I don’t know. Would certainly love to see someone look into it.
I can tell you that bumping the default VRAM allocation on Windows handhelds has taken some AAA games from unplayable to quite solid in my experimentation, but I’m not gonna sit here swapping OSs and games back and forth for benchmarks. At least not for free.