Overwatch
New Benchmark
aggregate test of different settings
All test where done with 540 FPS limit unless specified. QUICKPLAY: Esperança. WL means Wayland in this case. Meant to post this under the OW2 section instead of OW1
Distro: Arch Linux
Kernel: 6.7.10-273-tkg-pds-llvm
GPU Driver:
GPU: AMD Radeon RX 6900 XT (RADV NAVI21)
CPU: 12th Gen Intel Core i9-12900KF
RAM: 16 GB
fsync vs esync vs off
wayland,lutris-ge-33,dxvk async,1080p setting all low,texture low,AA off
Distro: EndeavourOS
Kernel: 6.0.7-272-tkg-pds
GPU Driver: 4.6 Mesa 22.2.2
GPU: Radeon RX 6600/6600 XT/6600M
CPU: AMD Ryzen 5 3600 6-Core Processor
RAM: 8 GB
Distro: EndeavourOS
Kernel: 5.15.74-271-tkg-pds
GPU Driver: 4.6 Mesa 22.2.1
GPU: Radeon RX 6700/6700 XT/6750 XT / 6800M
CPU: AMD Ryzen 5 5600X 6-Core Processor
RAM: 16 GB
Distro: Arch Linux
Kernel: 5.19.3-1-cachyos-bmq
GPU Driver: 4.6 Mesa 22.1.7
GPU: Radeon RX 6900 XT
CPU: 12th Gen Intel Core i9-12900KF
RAM: 16 GB
Test on Hanamura with Bots Attack and Defense
I both times played mercy (Idk why it didn't get this information but I used a 3070) by the caped version, i've capt the fps ingame to 210
Distro: Ubuntu 22.04.1 LTS
Kernel: 5.15.0-46-generic
GPU Driver: 4.6.0 NVIDIA 510.85.02
GPU:
CPU: 11th Gen Intel Core i7-11700F @ 2.50GHz
RAM: 32 GB
Triple Buffering VS TearFree
custom game code: HEZHJ Thanks Xypher didn't know this code existed
Distro: Arch Linux
Kernel: 5.19.1-2-cachyos-pds
GPU Driver: 4.6 Mesa 22.1.6
GPU: Radeon RX 6900 XT
CPU: 12th Gen Intel Core i9-12900KF
RAM: 32 GB
Xephyr Fullscreen Comparison
Can't Confirm this based of charts but I fell like Xephyr increases input latency. FPS cap 370 not sure how it getting 97 avg of 444
Distro: Arch Linux
Kernel: 5.19.1-2-cachyos-pds
GPU Driver: 4.6 Mesa 22.1.6
GPU: Radeon RX 6900 XT
CPU: 12th Gen Intel Core i9-12900KF
RAM: 32 GB
Distro: Artix Linux
Kernel: 5.18.14-265-tkg-pds-llvm
GPU Driver: 4.6 Mesa 22.1.3
GPU: Radeon R9 290/390
CPU: AMD Ryzen 5 5600X 6-Core Processor
RAM: 16 GB
Wine Version Lurtis Default / Lutris-GE-Proton / Proton tkg 6.8
custom game code: HEZHJ
Distro: Artix Linux
Kernel: 5.18.12-264-tkg-pds
GPU Driver: 4.6 Mesa 22.1.3
GPU: Radeon R9 290/390
CPU: AMD Ryzen 5 5600X 6-Core Processor
RAM: 16 GB
Distro: Ubuntu 20.04.3 LTS
Kernel: 5.11.0-34-generic
GPU Driver: 4.6.0 NVIDIA 470.63.01
GPU:
CPU: AMD FX-4170 Quad-Core Processor
RAM: 8 GB
Distro: Ubuntu 20.04.3 LTS
Kernel: 5.11.0-34-generic
GPU Driver: 4.6.0 NVIDIA 470.63.01
GPU:
CPU: AMD FX-4170 Quad-Core Processor
RAM: 8 GB
Distro: Pop!_OS 21.04
Kernel: 5.11.0-7633-generic
GPU Driver: 4.6.0$ NVIDIA$ 470.57.02
GPU: GeForce GTX 1050 Ti
CPU: AMD Ryzen 7 2700X Eight-Core Processor
RAM: 16 GB
Distro: Arch Linux
Kernel: 5.13.15-xanmod1
GPU Driver: 4.6 Mesa 21.2.1
GPU: Radeon RX 470/480/570/570X/580/580X/590
CPU: AMD Ryzen 7 5800X 8-Core Processor
RAM: 16 GB
NGGC testing highlight
Testing NGGC with a highlight, trying to get more consistent results
Distro: Arch Linux
Kernel: 5.13.1-178-tkg-cacule
GPU Driver: 4.6 Mesa 21.2.0-devel
GPU: Radeon RX 6700/6700 XT / 6800M
CPU: AMD Ryzen 5 1600 Six-Core Processor
RAM: 16 GB
NGGC testing
Testing RADV's NGGC running from Gibraltar cart starting point to its final point
Distro: Arch Linux
Kernel: 5.13.1-178-tkg-cacule
GPU Driver: 4.6 Mesa 21.2.0-devel
GPU: Radeon RX 6700/6700 XT / 6800M
CPU: AMD Ryzen 5 1600 Six-Core Processor
RAM: 16 GB
Distro: Pop!_OS 21.04
Kernel: 5.11.0-7620-generic
GPU Driver: 4.6.0$ NVIDIA$ 465.31
GPU: GeForce GTX 750 Ti
CPU: Intel Core i5-3570 CPU @ 3.40GHz
RAM: 8 GB
Disable HT For Better Frametimes (For Higher Core CPU'S)
HT @4.7 Ghz and HT Off @4.8 Ghz These test contradict my previous test (in terms of 1% lows being higher with HT Off) seems like Lutri's WINE Runner Version 6.4 benefits from Having CSMT on and HT off unlike Lutri's WINE Runner Verions 6.0 This test has a lower fps avg compared to the custom gamemode running constantly at 400, Due to the map being Workshop Chamber maybe if FPS was higher HT might have a beneficial effect on performance. (These benchmarks were performed on Rialto in a competitive game around the first bridge/spawn area of attack and only lasted around two minutes as it makes benchmarking faster and a slightly more accurate IMO as switching sides in replay causes fps to vary drastically and cap at 60 FPS for a bit. My previous test where I was gettting 400 FPS avg was done on Workshop Chamber which is the smallest map in the game and is not indicative or realistic gaming performance.
Distro: Garuda Linux
Kernel: 5.11.14-zen1-1-zen
GPU Driver: 4.6.0 NVIDIA 465.24.02
GPU: GeForce RTX 2080 Ti Rev. A
CPU: Intel Core i7-8700K CPU @ 3.70GHz
RAM: 32 GB
Frametime Comparison Between Different Settings
These Results are outdated and should not be used as they do not accurately represent gaming performance since the release of Lutri's WINE Runner Version 6.4 (I still have a use for this comparison but will might delete it in the future)
Distro: Garuda Linux
Kernel: 5.11.8-zen1-1-zen
GPU Driver: 4.6.0 NVIDIA 460.67
GPU: GeForce RTX 2080 Ti Rev. A
CPU: Intel Core i7-8700K CPU @ 3.70GHz
RAM: 32 GB
HT ON / HT OFF Overwatch Kill Boss Custom Gamemode
CPU is clocked @4.8 Ghz Game Settings: Fullscreen Borderless Vsync Off Triple Buffering Off FPS Limiter 400 Texture Quality Set To High Used replay and spectated the same person (Boss) for the entire match. I used that gamemode due the the excess abilities being used causing a lot of particle effects. Workshop chamber map was used for this gamemode being the smallest map in the game fps is through the roof and is not indicative of the performance one gets during a real match of quick play or competitive.
Distro: Garuda Linux
Kernel: 5.11.7-zen1-1-zen
GPU Driver: 4.6.0 NVIDIA 460.56
GPU: GeForce RTX 2080 Ti Rev. A
CPU: Intel Core i7-8700K CPU @ 3.70GHz
RAM: 32 GB
Distro: Arch Linux
Kernel: 5.11.5-136-tkg-pds
GPU Driver: 4.6.0$ NVIDIA$ 455.50.10
GPU: GeForce GTX 1060 6GB
CPU: Intel Core i3-4370 CPU @ 3.80GHz
RAM: 12 GB
Distro: Arch Linux
Kernel: 5.11.5-136-tkg-pds
GPU Driver: 4.6.0$ NVIDIA$ 455.50.10
GPU: GeForce GTX 1060 6GB
CPU: Intel Core i3-4370 CPU @ 3.80GHz
RAM: 12 GB
fsync (futex2) vs fastsync
Benchmark of the performance of fsync vs fastsync in Overwatch.
Distro: Arch Linux
Kernel: 5.11.5-136-tkg-pds
GPU Driver: 4.6.0$ NVIDIA$ 455.50.10
GPU: GeForce GTX 1060 6GB
CPU: Intel Core i3-4370 CPU @ 3.80GHz
RAM: 12 GB
cpuset trick
Game settings: 1920x1080, 50% render scale, Low graphics settings (to get CPU bound scenario), fps uncapped to 400fps. Windowed Fullscreen, LXDE, X11. ------------------------------------ DXVK: Compiled from git, commit 0eec9584 compile flags: cpp_args and c_args = -march=native -O3 -pipe -flto -floop-strip-mine -fno-semantic-interposition -fipa-pta -fdevirtualize-at-ltrans c_link_args = -flto -static static-libgcc ------------------------------------ Wine-tkg: Fsync enabled env vars: STAGING_SHARED_MEMORY=1 __GL_DXVK_OPTIMIZATIONS=1 _protonify="true"_CROSS_FLAGS _GCC_FLAGS and _CROSS_FLAGS="-O3 -pipe -floop-strip-mine -fno-semantic-interposition -fipa-pta -fdevirtualize-at-ltrans" ------------------------------------- Kernel: linux-tkg command-line: random.trust_cpu=on nowatchdog mitigations=off amdgpu.ppfeaturemask=0xffffffff noreplace-smp clearcpuid=514 compile settings: _sched_yield_type="0" _rr_interval="default" _tickless="1" _fsync="true" _compileroptlevel="2" _processor_opt="native" _smt_nice="true" _timer_freq="500" ----------------------------------------------- Naming convention: - obs: obs recording happens with software x264 encoder, medium preset, FHD 60fps, High profile, Tune: animation - cpuset-smt: cpuset is used by separating vcores by smt, for each physical core, one smt goes to one cpuset and the other goes to the other cpuset - cpuset-ccx: cpuset is used to separate the ccxs, one ccx for overwatch, the rest for everything else (including obs recording if enabled) When there's a number 1 or 2, the computer has been restarted and the simulation redone. It seems that a 5fps variation is within margin of error. --------------------------------------- RAM: 4 dimms, single rank, 3400MHz, CR 1, CL 15, RCDRD 19, RCDWR 15, RP 16, RAS 34
Distro: Arch Linux
Kernel: 5.10.8-112-tkg-upds
GPU Driver: 4.6 Mesa 20.3.3
GPU: Radeon RX 5600 OEM/5600 XT / 5700/5700 XT
CPU: AMD Ryzen 7 3700X 8-Core Processor
RAM: 32 GB
5950X, 6800XT, 64 GBs of RAM 1440p Ultra
Colourfulness, FakeHDR, Vibrance, cas at 4K Ultra
Distro: Garuda Linux
Kernel: 5.10.4-107-tkg-bmq
GPU Driver: 4.6 Mesa 21.0.0-devel
GPU: AMD/ATI
CPU: AMD Ryzen 9 3900X 12-Core Processor
RAM: 32 GB
Xorg vs XWayland
Performance comparison between the game running on X (OpenBox) and on Wayland (Sway) under XWayland.
Distro: Arch Linux
Kernel: 5.4.71-ck1-1-lts
GPU Driver: 4.6 Mesa 20.2.1
GPU: Radeon RX 460/560D / Pro 450/455/460/555/555X/560/560X
CPU: Intel Pentium CPU G4620 @ 3.70GHz
RAM: 16 GB
Distro: Manjaro Linux
Kernel: 5.9.0-1-MANJARO
GPU Driver: 4.6 Mesa 20.2.1
GPU:
CPU: Intel Core i5-8250U CPU @ 1.60GHz
RAM: 7 GB
CPU schedulers benchmark: CFS vs CFS-tweaked vs Cachy vs MuQSS vs PDS
5.4.70 kernel was used for Cachy scheduler. 5.4.71 kernel was used for other schedulers. Cachy version is 5.4-r7-fix1.
Distro: Arch Linux
Kernel: 5.4.70-1-lts
GPU Driver: 4.6 Mesa 20.2.0
GPU: Radeon RX 460/560D / Pro 450/455/460/555/555X/560/560X
CPU: Intel Pentium CPU G4620 @ 3.70GHz
RAM: 16 GB
Distro: Ubuntu 20.04.1 LTS
Kernel: 5.8.10-tkg-upds
GPU Driver: 4.6$ Mesa$ 20.3.0-devel$ 2020-09-18$
GPU: Radeon RX 470/480/570/570X/580/580X/590
CPU: AMD FX-8320E Eight-Core Processor
RAM: 16 GB
Distro: Arch Linux
Kernel: 5.8.7-arch1-1
GPU Driver: 4.6 Mesa 20.1.7
GPU: Radeon RX 470/480/570/570X/580/580X/590
CPU: AMD Ryzen 5 1600 Six-Core Processor
RAM: 16 GB
Distro: Arch Linux
Kernel: 5.4.63-1-lts
GPU Driver: 4.6 Mesa 20.1.7
GPU: Radeon RX 460/560D / Pro 450/455/460/555/555X/560/560X
CPU: Intel Pentium CPU G4620 @ 3.70GHz
RAM: 16 GB
Distro: Arch Linux
Kernel: 5.4.61-1-lts
GPU Driver: 4.6 Mesa 20.1.6
GPU: Radeon RX 460/560D / Pro 450/455/460/555/555X/560/560X
CPU: Intel Pentium CPU G4620 @ 3.70GHz
RAM: 16 GB
Distro: Arch Linux
Kernel: 5.4.61-1-lts
GPU Driver: 4.6 Mesa 20.1.6
GPU: Radeon RX 460/560D / Pro 450/455/460/555/555X/560/560X
CPU: Intel Pentium CPU G4620 @ 3.70GHz
RAM: 16 GB
Distro: Arch Linux
Kernel: 5.7.12-zen1-1-zen
GPU Driver: 4.6 Mesa 20.1.5
GPU: Radeon RX 470/480/570/570X/580/580X/590
CPU: AMD Ryzen 7 1700X Eight-Core Processor
RAM: 16 GB
250 cap RADV/ACO vs AMDVLK vs AMDGPU-PRO
250 cap, most settings low or off, AMD vulkan loaders
Distro: Arch Linux
Kernel: 5.8.0
GPU Driver: 4.6 Mesa 20.1.5
GPU: Radeon RX 5600 OEM/5600 XT / 5700/5700 XT
CPU: AMD Ryzen 7 3800X 8-Core Processor
RAM: 16 GB
Overwatch 250cap
Overwatch, Lutris wine 5.7.6, 250fps cap (240hz display based), Ubuntu 20.04 (Mesa 20.2, AMDGPU kernel driver). First point gibralter.
Distro: Ubuntu 20.04 LTS
Kernel: 5.8.0-050800rc6-generic
GPU Driver: 4.6$ Mesa$ 20.2.0-devel$ 2020-07-22$
GPU:
CPU: AMD Ryzen 7 3800X 8-Core Processor
RAM: 16 GB
Distro: Pop!_OS 20.04 LTS
Kernel: 5.4.0-7634-generic
GPU Driver: 4.6.0$ NVIDIA$ 440.82
GPU: GeForce GTX 1080
CPU: AMD Ryzen 5 2600 Six-Core Processor
RAM: 24 GB
Distro: Arch Linux
Kernel: 5.7.2-zen1-1-zen
GPU Driver: 4.6 Mesa 20.1.1
GPU: Radeon RX 470/480/570/570X/580/580X/590
CPU: AMD Ryzen 7 1700X Eight-Core Processor
RAM: 16 GB
Distro: Gentoo/Linux
Kernel: 5.5.8-gentoo
GPU Driver: 4.6.0 NVIDIA 440.59
GPU: GeForce GTX 1660 Ti
CPU: AMD Ryzen 7 2700X Eight-Core Processor
RAM: 32 GB
Overwatch - 1660TI - Ryzen 7 2700X - Gentoo Linux
Stock Clocks for GPU and CPU Settings mixed between low and Ultra for ~144 fps (slightly lower)
Distro: Gentoo/Linux
Kernel: 5.5.8-gentoo
GPU Driver: 4.6.0 NVIDIA 440.59
GPU: GeForce GTX 1660 Ti
CPU: AMD Ryzen 7 2700X Eight-Core Processor
RAM: 32 GB
Overwatch Ultra quality on Manjaro Linux with Kernel 5.6-rc-tkg-pds with an RX 5600 XT/Ryzen 2600X
Overwatch multiplayer on Ultra graphics settings on Manjaro Linux, running kernel 5.6-rc-tkg-pds, with ACO, DXVK, wine version 5.2-tkg-protonified. Graphics card is an RX 5600 XT with a Ryzen 5 2600X
Distro: Manjaro Linux
Kernel: 5.6.0-rc2-1-tkg-pds
GPU Driver: 4.5 Mesa 19.3.4
GPU: Radeon RX 5700 / 5700 XT
CPU: AMD Ryzen 5 2600X Six-Core Processor
RAM: 16 GB