Atomic Heart on Xe-LP shows lower performance than on weaker device with Windows
System information
System:
Host: oxp Kernel: 6.5.9-1-cachyos arch: x86_64 bits: 64 compiler: gcc
v: 13.2.1 Desktop: KDE Plasma v: 5.27.9 tk: Qt v: 5.15.11 wm: kwin_wayland
dm: greetd Distro: Arch Linux
CPU:
Info: quad core model: 11th Gen Intel Core i7-1165G7 bits: 64 type: MT MCP
arch: Tiger Lake rev: 1 cache: L1: 320 KiB L2: 5 MiB L3: 12 MiB
Speed (MHz): avg: 466 high: 931 min/max: 400/4700 cores: 1: 400 2: 400
3: 400 4: 400 5: 400 6: 400 7: 931 8: 400 bogomips: 44851
Flags: avx avx2 ht lm nx pae sse sse2 sse3 sse4_1 sse4_2 ssse3 vmx
Graphics:
Device-1: Intel TigerLake-LP GT2 [Iris Xe Graphics] driver: i915 v: kernel
arch: Gen-12.1 ports: active: DP-2 off: eDP-1 empty: DP-1,DP-3,DP-4
bus-ID: 00:02.0 chip-ID: 8086:9a49
Device-2: Logitech BRIO Ultra HD Webcam driver: snd-usb-audio,uvcvideo
type: USB rev: 2.1 speed: 480 Mb/s lanes: 1 bus-ID: 3-5.3:6
chip-ID: 046d:085e
Display: wayland server: Xwayland v: 23.2.1 compositor: kwin_wayland
driver: X: loaded: modesetting alternate: fbdev,intel,vesa dri: iris
gpu: i915 display-ID: 0
Monitor-1: DP-2 model: Samsung LS34A650U res: 3440x1440 dpi: 109
diag: 869mm (34.2")
Monitor-2: eDP-1 model: TV080WUM-NL0 res: 1600x2560 dpi: 369
diag: 202mm (8")
API: EGL v: 1.5 platforms: device: 0 drv: iris device: 1 drv: swrast gbm:
drv: iris surfaceless: drv: iris wayland: drv: iris x11: drv: iris
API: OpenGL v: 4.6 compat-v: 4.5 vendor: intel v: N/A glx-v: 1.4
direct-render: yes renderer: Mesa Intel Xe Graphics (TGL GT2)
device-ID: 8086:9a49 display-ID: :0.0
Currently running mesa-git 24.0.0_devel.179524. ffa458ee-1
If applicable
- Wine/Proton version: Proton Experimental (as of 2023-10-28) and Proton 8.0-4
Describe the issue
While checking the somewhat lackluster performance of Atomic Heart on my device, I came across this video on YouTube: https://www.youtube.com/watch?v=DAiwXMwveAM
I am aware that Proton and the translations incur some performance penalty, however I am comparing the latest mesa-git (as of 2023-10-28) against a device in a video of 2023-02-23.
Additionally, the device is running an 1145g7 (Intel Xe with 80 Execution Units) with 3200MHz RAM vs my 1165g7 (Intel Xe with 96 Execution Units) running 4266 MHz RAM.
My device is running with an increased power limit and is reporting (via turbostat -q -s GFXMHz,GFXWatt,PkgWatt,CorWatt,RAMWatt,PkgTmp
) a consistent 1300 MHz on the GPU (= maximum):
[...]
PkgTmp GFXMHz PkgWatt CorWatt GFXWatt RAMWatt
79 1300 31.25 13.23 13.30 0.00
79 1300 31.25 13.23 13.30 0.00
PkgTmp GFXMHz PkgWatt CorWatt GFXWatt RAMWatt
83 1300 32.57 13.43 13.78 0.00
83 1300 32.57 13.43 13.78 0.00
PkgTmp GFXMHz PkgWatt CorWatt GFXWatt RAMWatt
77 1300 29.48 11.75 13.08 0.00
77 1300 29.48 11.75 13.08 0.00
PkgTmp GFXMHz PkgWatt CorWatt GFXWatt RAMWatt
72 1300 30.12 12.08 13.25 0.00
72 1300 30.12 12.08 13.25 0.00
PkgTmp GFXMHz PkgWatt CorWatt GFXWatt RAMWatt
74 1300 29.06 11.43 13.09 0.00
74 1300 29.06 11.43 13.09 0.00
[...]
Based on this, I think that there might be some issue at play here, which is why I want to raise the case.
Some FPS ranges that I am seeing (via Steam FPS Overlay, using settings identical to the ones in the video) during the introduction sequence:
Scene during initial boat ride | My device | Reference |
---|---|---|
while the robot tendrils are outside of the hand | 21 - 22 | 23 - 25 |
when passing the other boat | 20 - 21 | 21 - 25 |
when passing the group of people on the right | 23 - 25 | 24 - 28 |
when the robot offers soda | 21 - 23 | 24 - 28 |
when robot on bike is stuck in the tree | 23 - 26 | 31 - 37 |
when passing the balloons on the left | 26 - 28 | 28 - 33 |
Any extra information would be greatly appreciated
This behavior has been present with mesa-git 23.3.0_devel.178568. 9c2212f9-1 as well.