Sn8ke Posted May 6, 2013 Posted May 6, 2013 I did some testing today to test the effectiveness of SLI vs single GPU. My graphics setting are maxed out minus: civilian traffic, low water, cockpit mirrors, tree draw distance minimal. CPU is a core i7 2600K @ 4.6GHZ. Monitor: 1920 X 1080 large lcd full 1080P. Test was performed on a "taxing" campaign mission loaded with AI. Both test the aircraft overflew the same portions of the city at same speed and altitude. Test one Single 580 GTX 1.5gb: Avg: 42.188 - Min: 33 - Max: 62 Test two SLI 580 GTX 1.5gb: Avg: 59.368 - Min: 48 - Max: 63 Asus Prime Gaming Wifi7 // Intel 14900K @5.5GHz // 64Gb DDR5 6000MHz // 3090 RTX // 4TB Samsung NVME M.2
Wayc00lio Posted May 7, 2013 Posted May 7, 2013 Good results. I'd be interesting in hearing if this is a sustained improvement or just due to that one mission. [sIGPIC][/sIGPIC] Asus ROG Rampage Extreme VI; i9 7900X (all 10 cores at 4.5GHz); 32 Gb Corsair Dominator DDR4; EVGA 1080Ti Hybrid; 1Tb Samsung 960 Evo M2; 2Tb Samsung 850 Pro secondary. Oculus Rift; TM Warthog; Saitek Combat Pros.
TZeer Posted May 7, 2013 Posted May 7, 2013 I did some testing today to test the effectiveness of SLI vs single GPU. My graphics setting are maxed out minus: civilian traffic, low water, cockpit mirrors, tree draw distance minimal. CPU is a core i7 2600K @ 4.6GHZ. Monitor: 1920 X 1080 large lcd full 1080P. Test was performed on a "taxing" campaign mission loaded with AI. Both test the aircraft overflew the same portions of the city at same speed and altitude. Test one Single 580 GTX 1.5gb: Avg: 42.188 - Min: 33 - Max: 62 Test two SLI 580 GTX 1.5gb: Avg: 59.368 - Min: 48 - Max: 63 Just a few tips: If you wanna test GPU performance, don't do it in an environment that will be bottlenecked by your CPU. I have run SLI setup with the exact same cards. And they should perform better. When I had them there was a graphical bug related to the heat blur effect on the A-10C. Whenever the effect was active the FPS would drop by a large margin. It didn't matter if you looked at it or not. This graphical effect is only active when you re below ~260 knots. Which is almost all the time. So set up a simple scenario in the mission editor, with just your plane. Test FPS with heat blur effect turned on and off. You can also just have the plane with no weapons, and increase speed until the you go over the limit where the heat blur effect goes away. http://forums.eagle.ru/showthread.php?t=101459 Certain areas on the map has a large amount of 3d models with a high object count. Those object's are tied to the CPU usage. So more objects on a 3d model = higher load on the CPU. This will affect the FPS in a very big way, and in some cases you will not have a increase in FPS in any way due to CPU being already taxed 100%.
pii Posted May 7, 2013 Posted May 7, 2013 (edited) Just a few tips: If you wanna test GPU performance, don't do it in an environment that will be bottlenecked by your CPU. I have run SLI setup with the exact same cards. And they should perform better. When I had them there was a graphical bug related to the heat blur effect on the A-10C. Whenever the effect was active the FPS would drop by a large margin. It didn't matter if you looked at it or not. This graphical effect is only active when you re below ~260 knots. Which is almost all the time. So set up a simple scenario in the mission editor, with just your plane. Test FPS with heat blur effect turned on and off. You can also just have the plane with no weapons, and increase speed until the you go over the limit where the heat blur effect goes away. http://forums.eagle.ru/showthread.php?t=101459 Certain areas on the map has a large amount of 3d models with a high object count. Those object's are tied to the CPU usage. So more objects on a 3d model = higher load on the CPU. This will affect the FPS in a very big way, and in some cases you will not have a increase in FPS in any way due to CPU being already taxed 100%. The cpu is a bottle neck in this sim due to the flight modeling and all the system modeling, think of all the math that is going on. Any half way descent Video card should be able to handle the simple 3d buildings, even hundreds of them that are in the DCS world. I'll bet if you fly the A-10c vs any FC3 plane over the same area the FC3 planes will preform better fps wise. I haven't tried it myself so that would be a good test for you. And as always I could be wrong:dunno: Edited May 7, 2013 by pii
112th_Rossi Posted May 7, 2013 Posted May 7, 2013 In terms of CPU usage, the CPU in DCS should be mostly calculating the flight dynamics, collision detection and general game logic. Yes certain areas of the map are full of buildings and other objects, but most of them are just simple shapes with virtually no modern day post-processing effects. DCS still uses antiquated. antialiasing methods. No FXAA which is much more efficient. If theres any slow down on a decent gaming rig I would put it down to the engine struggling rather than anything else.
TZeer Posted May 7, 2013 Posted May 7, 2013 The cpu is a bottle neck in this sim due to the flight modeling and all the system modeling, think of all the math that is going on. Any half way descent Video card should be able to handle the simple 3d buildings, even hundreds of them that are in the DCS world. I'll bet if you fly the A-10c vs any FC3 plane over the same area the FC3 planes will preform better fps wise. I haven't tried it myself so that would be a good test for you. And as always I could be wrong:dunno: Nope, you would be surprised how much the object count matters to your FPS. See this thread: http://forums.eagle.ru/showthread.php?t=101491&page=6
Recommended Posts