TZeer Posted February 19, 2013 Author Posted February 19, 2013 Did we get a conclusion/answer to why the triangles and object count doubled when the model got put into the mission editor?
luckybob9 Posted February 19, 2013 Posted February 19, 2013 Interesting results from just one file. Object count could be because stock version displays all 3 LOD's at once and thus more objects at a time per model. Mine switch based on distance. 0-500m, 500-2500m, 2500-10000m Did we get a conclusion/answer to why the triangles and object count doubled when the model got put into the mission editor? One might count polygons while the other counts triangles. that gives you the rough factor of 2 which would make sense to the numbers you are seeing Nevada map contributer EDM Modeling tools FAQ: http://forums.eagle.ru/showpost.php?p=1418067&postcount=1 Seo libh a chairde is chanadh liom. Líonaigí'n oíche le greann is le spórt. Seo sláinte na gcarad atá imithe uainn. Mar cheo an tsléibhe uaine, iad imithe go deo
luckybob9 Posted February 19, 2013 Posted February 19, 2013 (edited) Standalone did AFAIK. So theoretically, if all cmd files converted to edm were done for world, it could explain why there is a reduction in performance in world vs standalone. Naturally there are other factors such as difference in coding but I am disregarding those variables in this thought Maybe it is the zfighting that is constantly making the engine try to figure out which mesh should be shown on top. I dont really know how the engine handles any of that. Just thinking out loud Edited February 19, 2013 by luckybob9 Nevada map contributer EDM Modeling tools FAQ: http://forums.eagle.ru/showpost.php?p=1418067&postcount=1 Seo libh a chairde is chanadh liom. Líonaigí'n oíche le greann is le spórt. Seo sláinte na gcarad atá imithe uainn. Mar cheo an tsléibhe uaine, iad imithe go deo
TZeer Posted February 19, 2013 Author Posted February 19, 2013 Did another test now. SLI vs non SLI. No difference at all. But I did notice now that the CPU was working quite hard when the number of objects went up. So the more objects, the harder the CPU works. The sweet spot for me was around 1100-1200 objects. When I passed that limit the CPU was at max load, and the GPU started reducing it's GPU usage. The more objects I introduced to the scene the lower the GPU usage went as did the FPS. So it seems low numbers of objects but higher number of triangles will work much better for models.
SkateZilla Posted February 19, 2013 Posted February 19, 2013 Bottleneck the CPU and it can no longer drive the GPU. another reason Single Threaded Games arent Good. yeah, the sound is on another core, but how much does that really affect the overall thread usage... CitiesXL has the same problem once your population gets above 150K or so, it turns into a slide show, because the CPU gets so bottlenecked, it can no longer process it's threads and drive/deliver the GPU instructions without slowing everything down. Windows 10 Pro, Ryzen 2700X @ 4.6Ghz, 32GB DDR4-3200 GSkill (F4-3200C16D-16GTZR x2), ASRock X470 Taichi Ultimate, XFX RX6800XT Merc 310 (RX-68XTALFD9) 3x ASUS VS248HP + Oculus HMD, Thrustmaster Warthog HOTAS + MFDs
TZeer Posted February 19, 2013 Author Posted February 19, 2013 Interesting results from just one file. Object count could be because stock version displays all 3 LOD's at once and thus more objects at a time per model. Mine switch based on distance. 0-500m, 500-2500m, 2500-10000m One might count polygons while the other counts triangles. that gives you the rough factor of 2 which would make sense to the numbers you are seeing Hmm, seems a bit strange. Cause the tank model is all modeled and textured in the modelviewer. I can even start it's animations. And the object count never goes above 3. Bottleneck the CPU and it can no longer drive the GPU. another reason Single Threaded Games arent Good. yeah, the sound is on another core, but how much does that really affect the overall thread usage... CitiesXL has the same problem once your population gets above 150K or so, it turns into a slide show, because the CPU gets so bottlenecked, it can no longer process it's threads and drive/deliver the GPU instructions without slowing everything down. I'm one of the lucky ones with a overclocked CPU. But I can just imagine the reduced performance some may get due to non optimized models. Take around the harbor of Novo as an example. Over 2000 objects. Over 700 of them from the containers alone.
speckfire Posted February 19, 2013 Posted February 19, 2013 You guys are on to something. It seems the developer is using horsepower to rotate square wheels instead of making them round like you guys found out. Speed is life !!!
ENO Posted February 19, 2013 Posted February 19, 2013 Well put. I like that reference. "ENO" Type in anger and you will make the greatest post you will ever regret. "Sweetest's" Military Aviation Art
SkateZilla Posted February 20, 2013 Posted February 20, 2013 My Overclock to 5.1Ghz helped.... , but I dont like the temps I see... Windows 10 Pro, Ryzen 2700X @ 4.6Ghz, 32GB DDR4-3200 GSkill (F4-3200C16D-16GTZR x2), ASRock X470 Taichi Ultimate, XFX RX6800XT Merc 310 (RX-68XTALFD9) 3x ASUS VS248HP + Oculus HMD, Thrustmaster Warthog HOTAS + MFDs
BRooDJeRo Posted February 20, 2013 Posted February 20, 2013 (edited) I get noticable speedimprovements under load too when i put my AMD X6 3.2 Ghz to just the clockspeed 3.6/3.8 Ghz. But that seems obvious when 2/2.5 cores are mostly in use to full extend. Performance goes incremental when going up to 4.4 Ghz and tweak the memory and busspeeds. The question i have is it end of the line in terms of core-usage or is it intended to behave like this. Edited February 20, 2013 by BRooDJeRo
71st_Mastiff Posted February 20, 2013 Posted February 20, 2013 Sound slike ED needs to fire whom ever they have doing there optimization. for World EDM...:joystick: "any failure you meet, is never a defeat; merely a set up for a greater come back", W Forbes. "Success is not final, failure is not fatal, it is the courage to continue that counts", "He who never changes his mind, never changes anything," Winston Churchill. MSI z690 MPG DDR4 || i9-14900k|| ddr4-64gb PC3200 |zotac RTX 5080|Game max 1300w|Win11| |turtle beach elite pro 5.1|| ViRpiL,T50cm2||MFG Crosswinds|| VT50CM-plus rotor Throttle || Z10 RGB EVGA Keyboard/ G502LogiMouse || PiMax Crystal VR || 32 Asus||
Kuky Posted February 20, 2013 Posted February 20, 2013 That is great stuff TZeer and good approach to test and get results that lead to good conclusion... Lucky bob knows his stuff and I am confident this cmd to edm conversion went wrong and is reason why the performance hit. I just noticed one of the testers added this to ED bug tracking system so it'll get some attention :thumbup: PC specs: Windows 11 Home | Asus TUF Gaming B850-Plus WiFi | AMD Ryzen 7 9800X3D + LC 360 AIO | MSI RTX 5090 LC 360 AIO | 55" Samsung Odyssey Gen 2 | 64GB PC5-48000 DDR5 | 1TB M2 SSD for OS | 2TB M2 SSD for DCS | NZXT C1000 Gold ATX 3.1 1000W | TM Cougar Throttle, Floor Mounted MongoosT-50 Grip on TM Cougar board, MFG Crosswind, Track IR
luckybob9 Posted February 20, 2013 Posted February 20, 2013 Im just guessing its that. I was just thinking now that its totally possible that the CMD format was unoptimized to begin with and when they got moved to the new format, they kept the CMD optimization and never got the EDM optimization. Given how the original EDM format compares to the current EDM, (there have been multiple iterations), I'd say that the developer knows his skills and does a great job Nevada map contributer EDM Modeling tools FAQ: http://forums.eagle.ru/showpost.php?p=1418067&postcount=1 Seo libh a chairde is chanadh liom. Líonaigí'n oíche le greann is le spórt. Seo sláinte na gcarad atá imithe uainn. Mar cheo an tsléibhe uaine, iad imithe go deo
locoidal Posted February 20, 2013 Posted February 20, 2013 (edited) I would like to know if "EDM models", "flood lights", "heat blur" "chimney smoke .. or smoke/dust in general" were fixed .. how many FPS will come back... I think those things are not so complicated to be fixed... Come on ED!!! give us this extra gift inside the next 1.2.3 update :D P.S: Great work Tzeer and luckybob!! Edited February 20, 2013 by locoidal Asus P6T deluxe V2 | I7 950@3.8 | 6 GB DDR3 (@1600 Tri-channel) 2x ATI 5850 (crossfire-->OFF) | Creative X-Fi | SSD OCZ Vertex 3 | OCZ ZX Series 1250W PSU | TrackIR 5 |
TZeer Posted February 20, 2013 Author Posted February 20, 2013 Would also be nice if we could get a definitive answer from ED regarding the doubling of objects and triangles when it comes to modelviewer versus in-game. It's not a big deal for models consisting of 2-5 objects etc. But when you get the A-10C, F-15E, AH-64D etc, they all have ~50 objects in them when checking them out in the modelviewer. And when they come into the game 2x of that. 10 of those units together and you are getting close to 1000 objects. My overclocked CPU @4,6GHz is at it's max when I reach 1100-1200 objects.
Vlerkies Posted February 20, 2013 Posted February 20, 2013 Cracking thread and great work TZeer! :thumbup: Thermaltake View 91, Z390 Gigabyte Aorus Ultra, i9 9900K, Corsair H150i Pro, 32Gb Trident Z 3200, Gigabyte Aorus Extreme 2080ti, Corsair AX1200i, Warthog A-10 Hotas, MFG Crosswind pedals, TiR5 Pro, HP Reverb Pro
SkateZilla Posted February 20, 2013 Posted February 20, 2013 Engine is showing signs of a CPU bottleneck, I guess.. Windows 10 Pro, Ryzen 2700X @ 4.6Ghz, 32GB DDR4-3200 GSkill (F4-3200C16D-16GTZR x2), ASRock X470 Taichi Ultimate, XFX RX6800XT Merc 310 (RX-68XTALFD9) 3x ASUS VS248HP + Oculus HMD, Thrustmaster Warthog HOTAS + MFDs
159th_Falcon Posted February 20, 2013 Posted February 20, 2013 Hmmm, have similar tests been done yet whit trees? or are those not objects ingame cause there not collidable? [sIGPIC][/sIGPIC] The keeper of all mathematical knowledge and the oracle of flight modeling.:)
TZeer Posted February 20, 2013 Author Posted February 20, 2013 It looks like a lot of individual trees, but they are really large models which consists of 1-3 objects and textures on them. So they are not really that resource hungry.
luckybob9 Posted February 20, 2013 Posted February 20, 2013 (edited) please get me a 100 unit comparison of this: It should be the same but with the transparency fixedtest.rar Edited February 20, 2013 by luckybob9 Nevada map contributer EDM Modeling tools FAQ: http://forums.eagle.ru/showpost.php?p=1418067&postcount=1 Seo libh a chairde is chanadh liom. Líonaigí'n oíche le greann is le spórt. Seo sláinte na gcarad atá imithe uainn. Mar cheo an tsléibhe uaine, iad imithe go deo
TZeer Posted February 21, 2013 Author Posted February 21, 2013 please get me a 100 unit comparison of this: It should be the same but with the transparency fixed I'm currently at work and unable to do any tests. Wont be home until 6th of March.
Witchking Posted February 21, 2013 Posted February 21, 2013 Fantastic work guys. I hope the ED devs are watching this topic and take this into consideration. I think its about time and this is incentive enough to get rid of some of the old flanker 2 era / lockon 1.0 era models/objects. WHISPR | Intel I7 5930K | Nvidia GTX980 4GB GDDR5 | 16GB DDR4 | Intel 730 series 512GB SSD | Thrustmaster WARTHOG | CH Pro Pedals | TrackIR4 pro | |A-10C|BS2 |CA|P-51 MUSTANG|UH-1H HUEY|MI-8 MTV2 |FC3|F5E|M2000C|AJS-37|FW190|BF 109K|Mig21|A-10:SSC,EWC|L-39|NEVADA|
locoidal Posted February 27, 2013 Posted February 27, 2013 Model LODs have been better optimized. Does anyone check if the new models (cranes and so on...) are now been optimized? Asus P6T deluxe V2 | I7 950@3.8 | 6 GB DDR3 (@1600 Tri-channel) 2x ATI 5850 (crossfire-->OFF) | Creative X-Fi | SSD OCZ Vertex 3 | OCZ ZX Series 1250W PSU | TrackIR 5 |
EtherealN Posted February 27, 2013 Posted February 27, 2013 Does anyone check if the new models (cranes and so on...) are now been optimized? I'll take a look at it. If things doesn't check out, I'll make sure it is reported internally. [sIGPIC][/sIGPIC] Daniel "EtherealN" Agorander | Даниэль "эфирныйн" Агорандер Intel i7 2600K @ 4.4GHz, ASUS Sabertooth P67, 8GB Corsair Vengeance @ 1600MHz, ASUS GTX 560Ti DirectCU II 1GB, Samsung 830series 512GB SSD, Corsair AX850w, two BENQ screens and TM HOTAS Warthog DCS: A-10C Warthog FAQ | DCS: P-51D FAQ | Remember to read the Forum Rules | | | Life of a Game Tester
locoidal Posted February 27, 2013 Posted February 27, 2013 I'll take a look at it. If things doesn't check out, I'll make sure it is reported internally. Thanks man ... i think it is not so difficult to get fixed as luckybob already tried to fix some of them :thumbup: Asus P6T deluxe V2 | I7 950@3.8 | 6 GB DDR3 (@1600 Tri-channel) 2x ATI 5850 (crossfire-->OFF) | Creative X-Fi | SSD OCZ Vertex 3 | OCZ ZX Series 1250W PSU | TrackIR 5 |
Recommended Posts