Avalanche110 Posted March 6, 2021 Posted March 6, 2021 Hello Everyone! I recently decided to bite the bullet and get a new Graphics Card. I had absolutely no Idea what to expect, I had read many forum posts that indicated a negligible improvement. Which is NOT GOOD for such a massive investment. While waiting for the new card to arrive I decided that I wanted to quantify exactly what improvement I had gotten for the investment. This was not an easy undertaking. So, Here goes....... Intel 8700k – delided -overclocked to 5.0GHz @ 1.35v Asus 1080ti 11Gb Vram – 400MHz overclock on Vram – 200Mhz OC on the core Coolmaster Captain 360 AIO Asus Z370e Motherboard 32Gb of G.Skill 3200mz Ram Samsung 970Pro m.2 Corsair RM1000i Power Supply Pimax 5k+ Pitool version – V1.0.1.268 Firmware – V2.1.255.270 Fov – Normal Refresh rate – 90Hz Render Quality – 1 Steam Vr Settings: Resolution per eye: 500% - 5728 x 4708 Re-projection: (off) MotionSmoothing: (off) Nvidia control Panel settings: Prefer maximum performance (on) Virtual reality pre-rendered frames (3) Windows 10 home: Version 20H2 Hyperthreading: Disabled Process Lasso: UNINSTALLED Windows Page file: 32Gb Game DVR: (Disabled) Windows Game mode: (Disabled) Full Screen optimization: (Disabled) Nvidia Display driver: 461.72 Spectre / Meltdown Protection: Disabled Vr Shader Mod (clear Water): Installed Mask Size: 0.675 Frustrum Culling (on) DCS Version: 2.5.6.61527
Avalanche110 Posted March 6, 2021 Author Posted March 6, 2021 Test Scenario 1: I loaded a very large Dynamic Mission on the Syria map onto a dedicated server Running on my machine ( Helps with FPS ) This mission is extremely taxing, it has thousands of lines of lua script running along with hundreds of units. As this is a big year for helicopters in DCS, I started with a Gazelle taking off from a FOB in Karsi. Flight Plan would take me through a mountain valley, over the plains toward a Target near Aleppo. Mission start time was just before Dawn, This is normally EXTREMELY taxing on the GPU and its not uncommon for my FPS to drop down to Single digits during Day break. This scenario was chosen as it reflected "real world" performance. as opposed to a basic free flight mission. It also what I spend the majority of my time doing, so It was relevant to me. Test Methodology. Using FPS VR to measure FPS and frame times. Flight was recorded using OBS. Upon finishing the test Flight, I reviewed the recording and polled the results at 5 minute intervals. I will post the mission on Youtube so all can evaluate for themselves. 1080ti 1 minute = 38FPS – GPU FRAMETIME = 26ms CPU FRAMETIME = 11.9ms GPU usage = 98% 5 minutes = 31FPS – GPU FRAMETIME = 32ms CPU FRAMETIME = 16.4ms GPU usage = 98% 10 minutes = 31FPS – GPU FRAMETIME = 33.3ms CPU FRAMETIME = 14.8ms GPU usage = 98% 15 minutes = 22FPS – GPU FRAMETIME = 47.5ms CPU FRAMETIME = 15.5ms GPU usage = 98% 20 minutes = 46FPS – GPU FRAMETIME = 21.5ms CPU FRAMETIME = 15.5ms GPU usage = 98% 25 minutes = 47FPS – GPU FRAMETIME = 20.9ms CPU FRAMETIME = 15.6ms GPU usage = 98% 30 minutes = 41FPS – GPU FRAMETIME = 23.8ms CPU FRAMETIME = 14.8ms GPU usage = 98% 35 minutes = 47FPS – GPU FRAMETIME = 20.3ms CPU FRAMETIME = 14.0ms GPU usage = 98% 40 minutes = 42FPS – GPU FRAMETIME = 30.8ms CPU FRAMETIME = 21.2ms GPU usage = 98% 45 minutes = 41FPS – GPU FRAMETIME = 25.7ms CPU FRAMETIME = 14.9ms GPU usage = 98% 50 minutes = 45FPS – GPU FRAMETIME = 22.7ms CPU FRAMETIME = 16.2ms GPU usage = 98% DCS SETTINGS.bmp
speed-of-heat Posted March 6, 2021 Posted March 6, 2021 my initial analysis based upon the data you have posted is that your 1080ti which the card you say you have in your spec, is slightly (hilariously) overloaded by the 500% steamvr settings... and at a guess so would a 3090... turning on the pd to 0.5 won’t help when you add 2xaa in game...you are just robbing Peter to pay Paul. SYSTEM SPECS: Hardware AMD 9800X3D, 64Gb RAM, 4090 FE, Virpil T50CM3 Throttle, WinWIng Orion 2 & F-16EX + MFG Crosswinds V2, Varjo Aero SOFTWARE: Microsoft Windows 11, VoiceAttack & VAICOM PRO YOUTUBE CHANNEL: @speed-of-heat
Avalanche110 Posted March 6, 2021 Author Posted March 6, 2021 New GPU: Asus Rog Strix 3090 OC edition. (insert aubergine emoji) Unboxing a GPU worth several orders of magnitude more than my car was fun. Set it up with little to no hassle. The only variable that changed between tests was instead of using MSI afterburner for the 1080ti, I switched to Asus GPU tweak II. Conducted the same mission in the Gazelle. 3090: 1 minute = 61FPS – GPU FRAMETIME = 15.0ms CPU FRAMETIME = 15.9ms – GPU = 87% (improvement = 23FPS) + 60.52% 5 minute = 53FPS – GPU FRAMETIME = 15.9.2ms CPU FRAMETIME = 18.1ms – GPU = 81% (improvement = 19FPS) + 61.29% 10 minute = 49FPS – GPU FRAMETIME = 20.0ms CPU FRAMETIME = 29.8ms – GPU = 90% (improvement = 18FPS) + 58.06% 15 minute = 50FPS – GPU FRAMETIME = 20.9ms CPU FRAMETIME = 25.9ms – GPU = 91% (improvement = 28FPS) + 127.27% 20 minute = 48FPS – GPU FRAMETIME = 20.1ms CPU FRAMETIME = 20.1ms – GPU = 90% (improvement = 2FPS) + 4.34% 25 minute = 59FPS – GPU FRAMETIME = 13.4ms CPU FRAMETIME = 16.0ms – GPU = 70% (improvement = 12FPS) + 25.5% 30 minute = 56FPS – GPU FRAMETIME = 15.0ms CPU FRAMETIME = 17.2ms – GPU = 62% (improvement = 15FPS) + 36.58% 35 minute = 50FPS – GPU FRAMETIME = 15.3ms CPU FRAMETIME = 19.1ms – GPU = 60% (improvement = 3FPS) = 6.38% 40 minute = 55FPS – GPU FRAMETIME = 16.1.4ms CPU FRAMETIME = 17.6ms– GPU = 82% (improvement = 13FPS) +30.95% 45 minute = 51FPS – GPU FRAMETIME = 15.5ms CPU FRAMETIME = 18.3ms – GPU = 71% (improvement = 10FPS) + 24.39% 50 minute = 42FPS – GPU FRAMETIME = 19.3ms CPU FRAMETIME = 23.1ms – GPU = 72% (improvement = -3FPS) - 6.66% Average percentage increase = 38.96% increase in performance (when GPU is the bottle neck ) Average FPS increase = 10.9FPS (26.07%) 1
Avalanche110 Posted March 6, 2021 Author Posted March 6, 2021 (edited) 10 minutes ago, speed-of-heat said: my initial analysis based upon the data you have posted is that your 1080ti which the card you say you have in your spec, is slightly (hilariously) overloaded by the 500% steamvr settings... and at a guess so would a 3090... turning on the pd to 0.5 won’t help when you add 2xaa in game...you are just robbing Peter to pay Paul. I had only been using this setting for 1 week, It was based on a post from Lucas S. Walter on Youtube. He did it as a joke, and found improved performance. I tried it and was also surprised to find a noticeable reduction in stuttering. Like most of us. I have spent days agonising over optimisation and VR performance. More comparisons to Come........ Edited March 6, 2021 by Avalanche110 1
speed-of-heat Posted March 6, 2021 Posted March 6, 2021 (edited) Sorry that’s funny. Turn down to 200% or 150% Use a pd of 1.0 In practice if you are pushing the practical resolution much above 3000x 3000 per eye you are going to have a problem... my settings are in my sig, don’t look at percentages as much as total resolution per eye. Edited March 6, 2021 by speed-of-heat SYSTEM SPECS: Hardware AMD 9800X3D, 64Gb RAM, 4090 FE, Virpil T50CM3 Throttle, WinWIng Orion 2 & F-16EX + MFG Crosswinds V2, Varjo Aero SOFTWARE: Microsoft Windows 11, VoiceAttack & VAICOM PRO YOUTUBE CHANNEL: @speed-of-heat
speed-of-heat Posted March 6, 2021 Posted March 6, 2021 And turn off AA SYSTEM SPECS: Hardware AMD 9800X3D, 64Gb RAM, 4090 FE, Virpil T50CM3 Throttle, WinWIng Orion 2 & F-16EX + MFG Crosswinds V2, Varjo Aero SOFTWARE: Microsoft Windows 11, VoiceAttack & VAICOM PRO YOUTUBE CHANNEL: @speed-of-heat
Avalanche110 Posted March 6, 2021 Author Posted March 6, 2021 Test Scenario 2: Loaded the same mission to a dedicated Server. This time chose a F/A-18C taking off from Ramat David for a A2A Cap mission. Test mission was cut shorter than I would have liked as I ate a R2-7ER in the second mission. 1080ti 0 minute = 31FPS – GPU FRAMETIME = 30.6ms CPU FRAMETIME = 20.3ms 5 minute = 34FPS – GPU FRAMETIME = 29.3ms CPU FRAMETIME = 20.7ms 10 minute = 11FPS – GPU FRAMETIME = 87.3ms CPU FRAMETIME = 18.4ms 15 minute = 35FPS – GPU FRAMETIME = 18.0ms CPU FRAMETIME = 11.4ms 20 minute = 46FPS – GPU FRAMETIME = 16.7ms CPU FRAMETIME = 10.2ms 25 minute = 55FPS – GPU FRAMETIME = 18.2ms CPU FRAMETIME = 11.1ms 30 minute = 58FPS – GPU FRAMETIME = 16.8ms CPU FRAMETIME = 10.9ms 35 minute = 51FPS – GPU FRAMETIME = 19.6ms CPU FRAMETIME = 13.2ms 3090 0 minute = 46FPS – GPU FRAMETIME = 18.8ms CPU FRAMETIME = 25.0ms – GPU = 77% (improvement = 15FPS) + 48.38% 5 minute = 39FPS – GPU FRAMETIME = 20.0ms CPU FRAMETIME = 26.1ms – GPU = 74% (improvement = 5FPS) + 14.70% 10 minute = 22FPS – GPU FRAMETIME = 47.7ms CPU FRAMETIME = 46.8ms – GPU = 92% (improvement = 11FPS) + 100.00% 15 minute = 64FPS – GPU FRAMETIME = 11.3ms CPU FRAMETIME = 14.5ms – GPU = 68% (improvement = 29FPS) + 82.85% 20 minute = 69FPS – GPU FRAMETIME = 10.8ms CPU FRAMETIME = 14.6ms – GPU = 64% (improvement = 23FPS) + 50.00% 25 minute = 61FPS – GPU FRAMETIME = 13.5ms CPU FRAMETIME = 15.4ms – GPU = 64% (improvement = 6FPS) + 10.90% 30 minute = 64FPS – GPU FRAMETIME = 18.2ms CPU FRAMETIME = 16.2ms – GPU = 67% (improvement = 6FPS) + 10.3% 35 minute = 73FPS – GPU FRAMETIME = 12.8ms CPU FRAMETIME = 12.9ms – GPU = 66% (improvement = 22FPS) = 43.13% Average percentage increase = 51.08% increase in performance (when GPU is the bottle neck) Average FPS increase = 14.6FPS 31.33%
Glide Posted March 6, 2021 Posted March 6, 2021 Nice tests. Any idea why the FPS seem to improve after 15 minutes? Are there many units getting destroyed in the mission or a huge amounts of ordinance flying about??
nrosko Posted March 6, 2021 Posted March 6, 2021 Is your CPU bottlenecking at all? I'm guessing not. 500% in steam vr then 0.5 in game, I think this will break my system but will give it a pop. Win 10 64//4.5g i7 Kaby Lake//gtx Titan x pascal//16gb 3200ram//Asus Maximux Hero IX//Oculus Rift//
Avalanche110 Posted March 6, 2021 Author Posted March 6, 2021 1 hour ago, glide said: Nice tests. Any idea why the FPS seem to improve after 15 minutes? Are there many units getting destroyed in the mission or a huge amounts of ordinance flying about?? In the First test, Its the time of day. Just as the sun rises DCS tries to compute a ridiculous amount of shadows. It puts a huge strain on Both the CPU and GPU. In the Second test, the time of day is factor for the first 15min. once airborne FPS improves significantly. My 1080ti never had a problem at altitude, but taxiing around runways was becoming a problem. 2
Avalanche110 Posted March 6, 2021 Author Posted March 6, 2021 40 minutes ago, nrosko said: Is your CPU bottlenecking at all? I'm guessing not. 500% in steam vr then 0.5 in game, I think this will break my system but will give it a pop. Lucas s/ Wolta posted this about a week ago. I tried it and was surprised by the results. However, what Speed of heat mentioned above does make sense also. Their may be something to it, their may not. I have just tried another test with DCS pixel density at 1.0 and SteamVR Resolution at 150% See results Below: VR performance.bmp 2
Avalanche110 Posted March 6, 2021 Author Posted March 6, 2021 4 hours ago, speed-of-heat said: Sorry that’s funny. Turn down to 200% or 150% Use a pd of 1.0 In practice if you are pushing the practical resolution much above 3000x 3000 per eye you are going to have a problem... my settings are in my sig, don’t look at percentages as much as total resolution per eye. What your saying does make sense. You posted a link to a "PDSS calculator" in another thread. I'm not sure i am using it correctly as I don't know what value to add For SS PD =0.5 - Resolution 5728 x 4708 - SS= 1 ( 67418 Pixels ) vs PD =1.0 - Resolution 3176 x 2612 - SS= 1 ( 82957 Pixels ) Anyway, I took your advice and dropped Resolution to 150% and increased DCS P.D to 1.0, I got a 7.67% decrease in average FPS. 3090: Pixel density 0.5 – Resolution per eye – 500% 1 minute = 61FPS – GPU FRAMETIME = 15.0ms CPU FRAMETIME = 15.9ms – GPU = 87% 5 minute = 53FPS – GPU FRAMETIME = 15.9.2ms CPU FRAMETIME = 18.1ms – GPU = 81% 10 minute = 49FPS – GPU FRAMETIME = 20.0ms CPU FRAMETIME = 29.8ms – GPU = 90% 15 minute = 50FPS – GPU FRAMETIME = 20.9ms CPU FRAMETIME = 25.9ms – GPU = 91% 20 minute = 48FPS – GPU FRAMETIME = 20.1ms CPU FRAMETIME = 20.1ms – GPU = 90% 25 minute = 59FPS – GPU FRAMETIME = 13.4ms CPU FRAMETIME = 16.0ms – GPU = 70% 30 minute = 56FPS – GPU FRAMETIME = 15.0ms CPU FRAMETIME = 17.2ms – GPU = 62% 35 minute = 50FPS – GPU FRAMETIME = 15.3ms CPU FRAMETIME = 19.1ms – GPU = 60% 40 minute = 55FPS – GPU FRAMETIME = 16.1.4ms CPU FRAMETIME = 17.6ms– GPU = 82% 45 minute = 51FPS – GPU FRAMETIME = 15.5ms CPU FRAMETIME = 18.3ms – GPU = 71% 3090 Pixel density 1.0 – Resolution per eye – 150% 1 minute = 61FPS – GPU FRAMETIME = 14.6ms CPU FRAMETIME = 15.2ms – GPU = 91% (improvement = 0FPS) + 0% 5 minute = 50FPS – GPU FRAMETIME = 16.0ms CPU FRAMETIME = 18.8ms – GPU = 82% (improvement = -3FPS) – 5.6% 10 minute = 44FPS – GPU FRAMETIME = 21.3ms CPU FRAMETIME = 25.3ms – GPU = 94% (improvement = -5FPS) - 10.20% 15 minute = 46FPS – GPU FRAMETIME = 23.2ms CPU FRAMETIME = 25.4ms – GPU = 88% (improvement = -4FPS) - 8.00% 20 minute = 50FPS – GPU FRAMETIME = 18.8ms CPU FRAMETIME = 18.0ms – GPU = 82% (improvement = +2FPS) + 4.16% 25 minute = 48FPS – GPU FRAMETIME = 18.1ms CPU FRAMETIME = 19.2ms – GPU = 72% (improvement = -6FPS) - 10.90% 30 minute = 49FPS – GPU FRAMETIME = 16.0ms CPU FRAMETIME = 19.2ms – GPU = 67% (improvement = -11FPS) - 19.64% 35 minute = 48FPS – GPU FRAMETIME = 18.6ms CPU FRAMETIME = 20.1ms – GPU = 71% (improvement = -2FPS) = -4.00% 40 minute = 43FPS – GPU FRAMETIME = 18.2ms CPU FRAMETIME = 22.1ms – GPU = 64% (improvement = -12FPS) – 21.81% 45 minute = 52FPS – GPU FRAMETIME = 17.8ms CPU FRAMETIME = 18.3ms – GPU = 71% (improvement = +1FPS) + 1.96% Average percentage decrease = -7.40% decrease in performance Average FPS Decrease = -4.1FPS -7.67%
Glide Posted March 6, 2021 Posted March 6, 2021 1 hour ago, Avalanche110 said: Average percentage decrease = -7.40% decrease in performance Average FPS Decrease = -4.1FPS -7.67% Performance looks much more consistent this time around.
speed-of-heat Posted March 6, 2021 Posted March 6, 2021 2 hours ago, Avalanche110 said: What your saying does make sense. You posted a link to a "PDSS calculator" in another thread. I'm not sure i am using it correctly as I don't know what value to add For SS PD =0.5 - Resolution 5728 x 4708 - SS= 1 ( 67418 Pixels ) vs PD =1.0 - Resolution 3176 x 2612 - SS= 1 ( 82957 Pixels ) Anyway, I took your advice and dropped Resolution to 150% and increased DCS P.D to 1.0, I got a 7.67% decrease in average FPS. 3090: Pixel density 0.5 – Resolution per eye – 500% 1 minute = 61FPS – GPU FRAMETIME = 15.0ms CPU FRAMETIME = 15.9ms – GPU = 87% 5 minute = 53FPS – GPU FRAMETIME = 15.9.2ms CPU FRAMETIME = 18.1ms – GPU = 81% 10 minute = 49FPS – GPU FRAMETIME = 20.0ms CPU FRAMETIME = 29.8ms – GPU = 90% 15 minute = 50FPS – GPU FRAMETIME = 20.9ms CPU FRAMETIME = 25.9ms – GPU = 91% 20 minute = 48FPS – GPU FRAMETIME = 20.1ms CPU FRAMETIME = 20.1ms – GPU = 90% 25 minute = 59FPS – GPU FRAMETIME = 13.4ms CPU FRAMETIME = 16.0ms – GPU = 70% 30 minute = 56FPS – GPU FRAMETIME = 15.0ms CPU FRAMETIME = 17.2ms – GPU = 62% 35 minute = 50FPS – GPU FRAMETIME = 15.3ms CPU FRAMETIME = 19.1ms – GPU = 60% 40 minute = 55FPS – GPU FRAMETIME = 16.1.4ms CPU FRAMETIME = 17.6ms– GPU = 82% 45 minute = 51FPS – GPU FRAMETIME = 15.5ms CPU FRAMETIME = 18.3ms – GPU = 71% 3090 Pixel density 1.0 – Resolution per eye – 150% 1 minute = 61FPS – GPU FRAMETIME = 14.6ms CPU FRAMETIME = 15.2ms – GPU = 91% (improvement = 0FPS) + 0% 5 minute = 50FPS – GPU FRAMETIME = 16.0ms CPU FRAMETIME = 18.8ms – GPU = 82% (improvement = -3FPS) – 5.6% 10 minute = 44FPS – GPU FRAMETIME = 21.3ms CPU FRAMETIME = 25.3ms – GPU = 94% (improvement = -5FPS) - 10.20% 15 minute = 46FPS – GPU FRAMETIME = 23.2ms CPU FRAMETIME = 25.4ms – GPU = 88% (improvement = -4FPS) - 8.00% 20 minute = 50FPS – GPU FRAMETIME = 18.8ms CPU FRAMETIME = 18.0ms – GPU = 82% (improvement = +2FPS) + 4.16% 25 minute = 48FPS – GPU FRAMETIME = 18.1ms CPU FRAMETIME = 19.2ms – GPU = 72% (improvement = -6FPS) - 10.90% 30 minute = 49FPS – GPU FRAMETIME = 16.0ms CPU FRAMETIME = 19.2ms – GPU = 67% (improvement = -11FPS) - 19.64% 35 minute = 48FPS – GPU FRAMETIME = 18.6ms CPU FRAMETIME = 20.1ms – GPU = 71% (improvement = -2FPS) = -4.00% 40 minute = 43FPS – GPU FRAMETIME = 18.2ms CPU FRAMETIME = 22.1ms – GPU = 64% (improvement = -12FPS) – 21.81% 45 minute = 52FPS – GPU FRAMETIME = 17.8ms CPU FRAMETIME = 18.3ms – GPU = 71% (improvement = +1FPS) + 1.96% Average percentage decrease = -7.40% decrease in performance Average FPS Decrease = -4.1FPS -7.67% thats facsinating did you also turn of AA? SYSTEM SPECS: Hardware AMD 9800X3D, 64Gb RAM, 4090 FE, Virpil T50CM3 Throttle, WinWIng Orion 2 & F-16EX + MFG Crosswinds V2, Varjo Aero SOFTWARE: Microsoft Windows 11, VoiceAttack & VAICOM PRO YOUTUBE CHANNEL: @speed-of-heat
speed-of-heat Posted March 6, 2021 Posted March 6, 2021 out of inter 2 hours ago, Avalanche110 said: What your saying does make sense. You posted a link to a "PDSS calculator" in another thread. I'm not sure i am using it correctly as I don't know what value to add For SS PD =0.5 - Resolution 5728 x 4708 - SS= 1 ( 67418 Pixels ) vs PD =1.0 - Resolution 3176 x 2612 - SS= 1 ( 82957 Pixels ) your PD of 0.5 you are driving 6,212,206 pixels and your pd of 1 you are still driving more pixels at 6,866,512 try dropping your steam SS to 125% SYSTEM SPECS: Hardware AMD 9800X3D, 64Gb RAM, 4090 FE, Virpil T50CM3 Throttle, WinWIng Orion 2 & F-16EX + MFG Crosswinds V2, Varjo Aero SOFTWARE: Microsoft Windows 11, VoiceAttack & VAICOM PRO YOUTUBE CHANNEL: @speed-of-heat
Avalanche110 Posted March 7, 2021 Author Posted March 7, 2021 10 hours ago, glide said: Performance looks much more consistent this time around. Just to clarify, the inconsistency in frame rates for the first 15 minutes is largely caused by the time of day. this is where the DCS graphics engine is the issue, as opposed to resolution or Antialiasing. I chose to use this as the test because the 1080ti could not cope with the work load and it became unflyable. The 3090 had a 100% improvement, doubling FPS in this scenario. Making something that was unflyable a minor inconvenience. I Re-Ran the the test using the same settings, Just changed the time of day from 07:30 to 08:30 and got a 20% increase in FPS for the first 20min. 3090: Pixel density 0.5 – Resolution per eye – 500% Time of day = 07:30 1 minute = 61FPS – GPU FRAMETIME = 15.0ms CPU FRAMETIME = 15.9ms – GPU = 87% 5 minute = 53FPS – GPU FRAMETIME = 15.9.2ms CPU FRAMETIME = 18.1ms – GPU = 81% 10 minute = 49FPS – GPU FRAMETIME = 20.0ms CPU FRAMETIME = 29.8ms – GPU = 90% 15 minute = 50FPS – GPU FRAMETIME = 20.9ms CPU FRAMETIME = 25.9ms – GPU = 91% 20 minute = 48FPS – GPU FRAMETIME = 20.1ms CPU FRAMETIME = 20.1ms – GPU = 90% 3090 Pixel density 0.5 – Resolution per eye – 500% Time of day = 08:30 1 minute = 63FPS – GPU FRAMETIME = 14.4ms CPU FRAMETIME = 16.4ms – GPU = 83% (improvement = +2FPS) + 3.27% 5 minute = 52FPS – GPU FRAMETIME = 16.5ms CPU FRAMETIME = 16.9ms – GPU = 75% (improvement = -1FPS) – 1.8% 10 minute = 55FPS – GPU FRAMETIME = 17.1ms CPU FRAMETIME = 16.9ms – GPU = 89% (improvement = +5FPS) = +12.24% 15 minute = 49FPS – GPU FRAMETIME = 18.3ms CPU FRAMETIME = 18.6ms – GPU = 70% (improvement = - 1FPS) - 2.00% 20 minute = 52FPS – GPU FRAMETIME = 15.0ms CPU FRAMETIME = 17.6ms – GPU = 66% (improvement = +4FPS) + 8.33% Average percentage increase = +20.04% You can see that the GPU utilization was significantly lower in the second test. 1
Glide Posted March 7, 2021 Posted March 7, 2021 You get fewer peaks and valleys with the new settings. First settings 22-73 fps. Second set 43 - 61 fps.
Avalanche110 Posted March 7, 2021 Author Posted March 7, 2021 18 hours ago, speed-of-heat said: out of inter your PD of 0.5 you are driving 6,212,206 pixels and your pd of 1 you are still driving more pixels at 6,866,512 try dropping your steam SS to 125% I spent the day playing with Various settings to dial in the best results. Using the PDSS calculator, I worked out that a P.D of 1.0 and Steam Resolution of 125% is the Same number of Pixels as P.D 0.5 - Steam Resolution at 496% Tested both and found no difference either way. Disabled all antialiasing, really didn't notice a significant boost in GPU or CPU frame times, So opted to keep it. I Disabled "fixed Foveated Rendering" in PiTool - this was increasing CPU frame times significantly, While Decreasing GPU load. It would seem that OBS was bottle necking my CPU, so I stopped using it. I also opted to disable Tacview. I conducted the test mission in the gazelle once again. All DCS settings remain unchanged. DCS P.D: 0.5 SteamVR Resolution per eye: (496%) DCS Dedicated server - Syria Mission Start time: 08:30 1 minute = 68FPS – GPU FRAMETIME = 16.7ms CPU FRAMETIME = 10.0ms – GPU = 93% (improvement = +5FPS) + 7.93% 5 minute = 62FPS – GPU FRAMETIME = 15.5ms CPU FRAMETIME = 13.4ms – GPU = 91% (improvement = +9FPS) + 16.98% 10 minute = 66FPS – GPU FRAMETIME = 13.8ms CPU FRAMETIME = 13.5ms – GPU = 93% (improvement = +11FPS) + 20.00% 15 minute = 64FPS – GPU FRAMETIME = 16.0ms CPU FRAMETIME = 13.1ms – GPU = 88% (improvement = +15FPS) + 30.61% 20 minute = 70FPS – GPU FRAMETIME = 14.0ms CPU FRAMETIME = 11.1ms – GPU = 85% (improvement = +18FPS) + 34.61% 25 minute = 74FPS – GPU FRAMETIME = 13.5ms CPU FRAMETIME = 10.2ms – GPU = 89% (improvement = +15FPS) + 25.42% 30 minute = 70FPS – GPU FRAMETIME = 15.3ms CPU FRAMETIME = 10.4ms – GPU = 91% (improvement = +14FPS) + 25.00% 35 minute = 68FPS – GPU FRAMETIME = 16.8ms CPU FRAMETIME = 10.5ms – GPU = 94% (improvement = +13FPS) + 24.00% 40 minute = 68FPS – GPU FRAMETIME = 15.6ms CPU FRAMETIME = 11.2ms – GPU = 91% (improvement = +13FPS) + 23.00% Conclusion: Upgrading from a 1080ti to a 3090 has increased performance by an average of 28.7% ( in an apples for apples test scenario ) As the test above indicates, The 3090 is capable of an average 67.7 FPS, flying a helicopter at treetop level In Syria with MSAA and antialiasing turned on. Ultimately, For every 1% increase that a 3090 provides, It costs $92.39 (USD) 3 1
imacken Posted March 7, 2021 Posted March 7, 2021 7 hours ago, Avalanche110 said: I spent the day playing with Various settings to dial in the best results. Using the PDSS calculator, I worked out that a P.D of 1.0 and Steam Resolution of 125% is the Same number of Pixels as P.D 0.5 - Steam Resolution at 496% That is correct 0.5*0.5*5=1.25 or 125%. I still don’t understand why you mess around with PD and SS. Leave PD at 1.0 and test with different SS values. 1 Intel i7 12700K · MSI Gaming X Trio RTX 4090 · ASUS ROG STRIX Z690-A Wi-Fi · MSI 32" MPG321UR QD · Samsung 970 500Gb M.2 NVMe · 2 x Samsung 850 Evo 1Tb · 2Tb HDD · 32Gb Corsair Vengance 3000MHz DDR4 · Windows 11 · Thrustmaster TPR Pedals · Tobii Eye Tracker 5 · Thrustmaster F/A-18 Hornet Grip · Virpil MongoosT-50CM3 Base · Virpil Throttle MT-50 CM3 · Virpil Alpha Prime Grip · Virpil Control Panel 2 · Thrustmaster F-16 MFDs · HTC Vive Pro 2 · Total Controls Multifunction Button Box
Avalanche110 Posted March 9, 2021 Author Posted March 9, 2021 16 hours ago, imacken said: That is correct 0.5*0.5*5=1.25 or 125%. I still don’t understand why you mess around with PD and SS. Leave PD at 1.0 and test with different SS values. I Left it with a P.D of 0.5 for consistency when trying out various other settings. The settings I had used previously was P.D at 1 and resolution to 1.25 in Pitool. I did notice an improvement in stuttering when I switched to using SteamVR to set Resolution. Pitool is a notoriously buggy piece of software. Despite having disabled Tacview in game, It was still running on the server. The server logs indicated that :GetWorldObjects was using 7% CPU. I disabled Tacview and turned mirrors off in DCS settings and now I have an average FPS in the mid 70's in the same scenario as above. I think I should fire up the 1080ti again for another side by side comparison.
stacula666 Posted March 9, 2021 Posted March 9, 2021 7 hours ago, Avalanche110 said: I Left it with a P.D of 0.5 for consistency when trying out various other settings. The settings I had used previously was P.D at 1 and resolution to 1.25 in Pitool. I did notice an improvement in stuttering when I switched to using SteamVR to set Resolution. Pitool is a notoriously buggy piece of software. Despite having disabled Tacview in game, It was still running on the server. The server logs indicated that :GetWorldObjects was using 7% CPU. I disabled Tacview and turned mirrors off in DCS settings and now I have an average FPS in the mid 70's in the same scenario as above. I think I should fire up the 1080ti again for another side by side comparison. Keep this thread going 1
parxuk Posted March 9, 2021 Posted March 9, 2021 11 hours ago, Avalanche110 said: I think I should fire up the 1080ti again for another side by side comparison. As a 1080ti owner who would have to sell a kidney to afford a 3090, I would appreciate knowing just how worthwhile the difference was if you are going to compare the two. 1 Intel i7 8700K @ 5Ghz, Aorus Z370 Gaming5, 32GB Corsair Vengeance RGB Pro 3000MHz, Nvidia 3080Ti FE, Samsung 970 NVME, Samsung 512GB EVO SSD x 2, Thrustmaster Warthog HOTAS, Saitek Rudder Pedals, HP Reverb G2, Windows 11 x64, DCS Modules = FC3, A10c Warthog, F14 Tomcat, F18 Hornet, F16 Viper, Persian Gulf, Syria.
dburne Posted March 9, 2021 Posted March 9, 2021 (edited) I have been running DCS for several weeks now with an EVGA 3090 FTW3 Ultra. I think DCS likes all of that Vram of the GPU. I have been very pleased with the performance in DCS. I run very high gfx settings but I do run Motion Smoothing on and forced at 45 fps in my Reverb G2 at 100% Steam VR resolution (3172x3100 per eye). I find I don't need any MSAA at this setting and to me it looks very good. I do though have an i9 9900k I am running at 5.2 GHz on all 8 cores. No trouble maintaining the 45 fps in all the scenarios I have flown in - single player campaigns flying with the Hornet. Edited March 9, 2021 by dburne 2 Don B EVGA Z390 Dark MB | i9 9900k CPU @ 5.1 GHz | Gigabyte 4090 OC | 64 GB Corsair Vengeance 3200 MHz CL16 | Corsair H150i Pro Cooler |Virpil CM3 Stick w/ Alpha Prime Grip 200mm ext| Virpil CM3 Throttle | VPC Rotor TCS Base w/ Alpha-L Grip| Point Control V2|Varjo Aero|
Avalanche110 Posted March 11, 2021 Author Posted March 11, 2021 (edited) Having removed some pretty significant CPU bottle necks. I decided it would only be fair to do another comparison between the 1080ti and 3090. Test Scenario: Same mission, loaded onto a dedicated server running on the same PC I use for VR. Flying a Gazelle at treetop level from a Fob in Karsi, through mountain valleys to a target near Allepo I decided to be kinder to the 1080ti this time around and set the time of day to 8:30 as opposed to 7:30 (dawn) in the original test. Test Methodology: FPSVR Data logging utility to record GPU & CPU frame times. This method is far less taxing on the CPU than recording the flight in OBS. Removing the CPU bottle neck really helped the 1080 achieve a smooth and consistent frame rate, where as it previously jumped around a lot more. I was instantly reminded of why I decided I needed a new graphics card, average GPU frame times during cold start were in the mid 20's and It was barely making 40FPS. The GPU utilization was pinned at 98% for the entire flight, It did manage to sneak into the 50fps range once I had reached the open planes behind Karsi. The lowest GPU frame time recorded was 17.95ms. Lowest FPS recorded was 25FPS, highest was 53FPS. AVERAGE GPU FRAME TIME = 22.62ms AVERAGE CPU FRAME TIME = 10.20ms AVERAGE FPS = 43.7 Swapped out the 1080ti and reinstalled the 3090. Repeated the test using the exact same settings, drivers, resolution etc. Was blown away at the improvement (again) average GPU frame times approximately 10ms faster. Surprisingly, I did notice marginally better CPU frame times. This may have to do with the higher Vram on the 3090, or the drivers are running more efficiently with the newer Graphics card. There were several occasions when FPS was locked at 90 which is the Maximum Hz the Pimax can run at with a Large FOV setting. AVERAGE GPU FRAME TIME = 11.38ms (improvement: 98.76%) AVERAGE CPU FRAME TIME = 9.88ms (improvement: 3.23%) AVERAGE FPS = 76.0 (improvement: 73.91%) I don't think I could possibly make this test more accurate or the evidence any more empirical. if any one has any suggestions, please let me know. 1080ti FPSVR log 3090 FPSVR Log Edited March 11, 2021 by Avalanche110 Added FPSVR logs 1
Recommended Posts