Linux Gaming Performance Doesn't Appear Affected By The x86 PTI Work

[dc740]

I already wrote about network in the article on Wednesday.

I know I'm not a Premium member (at least for now), but could I suggest to perform tests on the Loading times? I mean... we all know the patch affects I/O. so... Pre/post loading times would be more helpful. Like: do games take more time to load using an HDD? (I have an SSD for the OS, but HDD for the games). What about startup times of the entire system?

These are the day to day basic tests that everyone should be looking at.

We all know that framerate was not affected that much, but I'm not always playing. There are other tasks you do daily, like constantly closing and opening apps, loading levels while gaming, starting up the system, waking up from hibernation, ¡¡¡¡compilation!!! (for those of us that code for a living). Was the linux kernel compilation time affected?

Thank you, Michael

[Michael]

Seconded! This should require more syscalls and TLB flushes than normal gaming after the game has loaded and is just running in an automated loop.

I already wrote about network in the article on Wednesday.

[Adarion]

It would be interesting to see what kind of difference there is in network performance (which would certainly impact multi-player games), since sockets require _tons_ of syscalls.

Seconded! This should require more syscalls and TLB flushes than normal gaming after the game has loaded and is just running in an automated loop.

[foppe]

Michael, you may want to test this with a nvidia card as well, since I'm hearing they're more reliant on syscalls than AMD cards, which avoid that since gcn1.1. (Though since Steve at HUB didn't find much, this may be irrelevant.)

[gururise]

Definitely need some tests while Gaming + Multitasking, as many people stream while gaming and this may affect performance severely. Another test that would be very interesting would be framerate while using SteamLink. I know myself and many other Linux gamers use Steam in-home streaming or SteamLink to game.

[newwen]

It would be interesting to see what kind of difference there is in network performance (which would certainly impact multi-player games), since sockets require _tons_ of syscalls.

Networking and also file system access. I guess that it might impact game loading times a little.

[Niarbeht]

Networking tests are already in the works for publishing shortly.

I suspect this'll vary by game. I'd suspect games with optimized and efficient netcode probably won't be impacted significantly. Games with a high tick-rate, though, or lots of big packets, those might have bigger issues.

[linuxgeex]

Michael You should be the first major review site to benchmark the impact on Virtualization tech. I suspect that the impact will be drastic for KVM / VirtualBox / Xen / Jailhouse / VMWare because the impact will be squared, or worse. ie if it's a 20% impact on the host and a 20% impact on the guest kernel too then the overall impact will be >40% on the app. PostgreSQL is going to get hammered in a KVM guest...

[kaels]

How does it affect gaming + CPU-heavy multitasking? Can we expect worse framerates (or periodic framerate drops) while streaming, recording, background rendering, running multiple game clients, hosting a game server in a VM, etc.?

How about background tasks that are relatively lightweight, but still need to update in real time: voice chat, incoming streams, torrents, etc.?

[Michael]

It would be interesting to see what kind of difference there is in network performance (which would certainly impact multi-player games), since sockets require _tons_ of syscalls.

Networking tests are already in the works for publishing shortly.