Hacker Newsnew | past | comments | ask | show | jobs | submit | more buyucu's commentslogin

Good. The Western world needs to get over its China-fetish.


Strix Halo from AMD appears to be a much more consumer-friendly alternative than DGX Spark.


It was never fun to travel to the US, and the border agents have always been rude, unprofessional and arbitrary. But it seems like it has gotten much worse, and there is no way I'm traveling to the US these days.


Republicans were always the party of big government.


I play a lot of older games from 90s and 2000s. They certainly work better on Linux+Wine+DXVK than on Windows.


It can get funny sometimes. There a game from 1999 that I had troubles running in Wine on Linux all the way back in 2005, and wrote a patch for it:

https://aow.heavengames.com/cgi-bin/forums/display.cgi?actio...

And then the same error message started showing up on Win10. Thankfully the same patch still works...


Not to mention for the different games that need different environments/configurations/libraries it's much easier to manage a bunch of profiles than a bunch of windows installations/VMs.


pretty much everything runs faster on Linux than on Windows.


I use Arch pretty much exclusively these days on my desktop, but this isn't quite true. Most of the time, Proton has a small performance impact (around 5% lower FPS), but some games tend to suffer more. For example Helldivers 2 runs around 10fps lower, which is pretty significant since I only got around 60-70 FPS on Windows at 4K (using a 3090).

Still, Proton is an amazing tool and these days it just works so well. The only games that don't work are those that are intentionally broken by invasive kernel-level anticheats. I won't be buying Battlefield 6, too bad for EA, there are now thousands of other games to play on Linux.


The general performance loss with the DX12 -> Vulkan translation on Linux especially with Nvidia hardware recently had the cause identified and will hopefully get solved in the near future. It has to do with descriptors and how Nvidia handles it is the general gist of it. A new Vulkan extension will be developed that more closely resembles how DX12 does things as I understand it, and then Nvidia and others can use that to hopefully solve this once and for all.

Here[1] is the full presentation and the slides[2] from it.

[1] https://video.tuwien.ac.at/events/xdc/v/OlwauRVEIGa

[2] https://indico.freedesktop.org/event/10/contributions/402/at...


Does this mean the performance gap of Proton for games using DX12 will close? What are the concrete implications of this?


It means that at the very least the Nvidia specific performance loss of 10-20% will disappear. That is already not an issue with AMD cards as I understand it.

If it'll further increase performance beyond that remains to be seen. I suspect there will always be some amount of overhead, although at least with earlier versions of DirectX it is quite minimal already.


You’re comparing a game running on a compatibility layer, running on Linux to a game running directly on Windows. Not quite what the parent comment was stating.


There are few native Linux games, so performance on Proton is the essential benchmark. Besides, even when Linux native versions are available, the Windows version on Proton often offers a better experience (fewer obvious bugs).


I think grosswait is talking about the dx12 -> Vulkan translation layer. Running Windows games on Proton that use Vulkan get more comparable performance than Windows games that use dx12.


Proton is essentially the only stable ABI that Linux game devs can target


I didn't do any tests myself but based on what I read you still pay a hefty performance penalty when using e.g., a 5090 on Linux.


That's expected. Always use AMD for Linux gaming.

There is some WIP to address it for Nvidia, but it requires new Vulkan features.

See: https://indico.freedesktop.org/event/10/contributions/402/at...


If you use a Microsoft product for anything security sensitive, you only have yourself to blame when it inevitably goes wrong.


I don't understand why people have a hard time understanding 'garbage in, garbage out'. If you train your model on junk, then you will have a junk model.


I suspect this is “the computer is always right”.

A lot of people think computers have better answers than people.

AI is just another type of computer. It knows a lot of things and sounds confident. Why wouldn’t it be right?


aws and azure are massively overpriced. there is no reason to use them.


I want one! Seriously, this should be a thing with more scents.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: