June 6, 20179 yr 10 minutes ago, Nytro said: No, v4 has still no multicore support, neither HT support. The load balance you see come from the OS itsself who handles the CPU load. v4 supports only single Core. And dont use AffinityMask, it will blurr all textures as soon they needs reloaded and autogen gets not generated anymore. That makes absolutely no sense and besides it's incorrect. Even in P3d V2 (and FSX for that matter), various threads were assigned by the app to specific cores.
June 6, 20179 yr As mentioned the trick to even core usage is in Task Manager/Details : Right click on P3D-->Set Affinity-->Check 4 cores are selected--> Click Ok. I always do it. Simulators: Prepar3D v5.4 | X-Plane 12 | DCS World | MSFS 2024 | PC Hardware: Dell U3417W | AMD Ryzen 7 9800 X3D | ASUS TUF 5070 Ti | ASUS TUF B580 Plus Wifi | G.Skill Z5 Neo 64GB 3000Mhz CL30 | Samsung 990 Pro 2TB + 970 EVO Plus 1TB + 860 EVO 2TB + 850 EVO 1TB, Western Digital Black Caviar Black 6TB | Corsair RM1000i | Corsair 280 Titan RX | VRM Fan | Fractal Design Define S2 Gunmetal | Flight Controls: Fulcrum One Yoke | Virpil VPC WarBRD Base | Virpil VPC MongoosT-50CM Grip, Thrustmaster Warthog+F/A-18C Grip | VIER IM POTT Sidestick CPT Side | Thrustmaster TPR Rudder Pedals | Virtual Fly TQ6+Throttle Quadrant | Sismo B737 Max Gear Lever | Monsterteck Desk Mounts | WINWING EfisL+FCU+MCDU | My fleet catalog: Link
June 6, 20179 yr 21 minutes ago, Nytro said: No, v4 has still no multicore support, neither HT support. The load balance you see come from the OS itsself who handles the CPU load. v4 supports only single Core. And dont use AffinityMask, it will blurr all textures as soon they needs reloaded and autogen gets not generated anymore. Of course it uses multi cores for threaded workers which load scenery. Thats reason why you see more blurries etc when use AM wrong way. i7-8700k, RTX2070 Super, custom water cooling, 32GB RAM, 6TB SSD spacePrepar3D v4, X-Plane 11, 40" 4k TVRex, Active Sky, RealAir and A2A pilot, GTN, ORBX mostly everywhere
June 6, 20179 yr 26 minutes ago, Nytro said: No, v4 has still no multicore support, neither HT support. The load balance you see come from the OS itsself who handles the CPU load. v4 supports only single Core. And dont use AffinityMask, it will blurr all textures as soon they needs reloaded and autogen gets not generated anymore. Absolutely not true. [MSI MPG X870E Carbon | 9800X3D (PBO +200Mhz / -20 Offset) | Corsair 64GB DDR5 (Custom Timings) | RTX 4090 Founders Edition (Undervolted) | WD SNX 850X 4TB + 4TB | Antec Flux Pro]
June 6, 20179 yr 35 minutes ago, Jiri Kocman said: WIth TEXTURE_SIZE_EXP = 10, with lower, it goes little bit down 80° !?! Btw, and for the probably 186 time on this forum... I can't find the damn thing "TEXTURE_SIZE_EXP = 10" in neither Prepar3D.cfg neither in Terrain.cfg. Where the hell is he ? Or he's not created in vanilla versions, but you can add it as new entrance ? And how does it help me, exactly ? FreeMeshX Global & P3D Tweak Assistant (PTA) plus a bit of orbx magic
June 6, 20179 yr I've been using Process Lasso to set all other programs that use higher CPU percentage to use all other cores, and run P3D4 on high priority, some mem tweaks in there too. Disclaimer: [email protected] on Asus Maximus X Formula, G.Skill TridentZ RGB 4x8GB 4266/17 XMP, EVGA 2080 ti Kingpin (8400/2160Mhz), Samsung 960 EVO 250GB PCIe M.2 NVMe SSD , 28TB HDD total - 4TB+ photoscenery, Romex Software PrimoCache RAM and SSD cache (must have!), 3x1080p 30" monitors, Samsung Odyssey VR HMD, Pimax 4k & BE HMDs, Samsung Gear VR '17, Homdio v1, Cardboard, custom loop 2x 360x64ML Rads, Thermaltake View 71, VRM watercool, Thermal Grizzly Conductonaut CPU (naked die), Fujipoly / ModRight Ultra Extreme System Builder Thermal Pad on MB VRM. 8x Corsair ML120 (slight positive pressure). 🙂
June 6, 20179 yr 9 minutes ago, gamer19 said: 80° !?! Btw, and for the probably 186 time on this forum... I can't find the damn thing "TEXTURE_SIZE_EXP = 10" in neither Prepar3D.cfg neither in Terrain.cfg. Where the hell is he ? Or he's not created in vanilla versions, but you can add it as new entrance ? And how does it help me, exactly ? TEXTURE_SIZE_EXP is not default in v4 you have to add it into prepar3d.cfg into [TERRAIN] section. Default value is 8, that means that textures 256x256 will be read, with value of 9 textures of 512x512 will be read that, it increase your LOD effectively as better textures are used for distant scenery. Value of 10 uses 1024x1024 textures. Effect in sim - longer loading time, can reduce performance, but no blurries on distant scenery.... better eye candy. 8GB video memory for value 8 is absolutely minimum, 11-12GB cards are mostly must. That is reason that you cant set it in GUI. Default + detailed textures checkbox do same as TSE=9 and it nedds about 4GB VRAM 80° sure even custom liquid cooling, it is OCed 4790k, that is nothing special for that CPU. limit is 99°C i7-8700k, RTX2070 Super, custom water cooling, 32GB RAM, 6TB SSD spacePrepar3D v4, X-Plane 11, 40" 4k TVRex, Active Sky, RealAir and A2A pilot, GTN, ORBX mostly everywhere
June 6, 20179 yr 29 minutes ago, Daedalus said: As mentioned the trick to even core usage is in Task Manager/Details : Right click on P3D-->Set Affinity-->Check 4 cores are selected--> Click Ok. I always do it. I used to do that too for a long long time I was one of the early disciples of it. Since then, I redid tests and there is no performance gain from doing it my end. I also tested core temps since core 0 is working so hard but I saw no lowering of core 0 temps even after doing this. I think SteveW has said in the past, it is an illusion because the usage graph doesn't actually measure true core usage. Here is my conclusion then: The ultimate best way to gauge core usage is to monitor core temperatures. The usage graphs are misleading. That is my understanding.
June 6, 20179 yr Commercial Member 4 hours ago, MikeT707 said: Hi Steve, I have an i7 4790K with HT (hyper threading) ENABLED. In order to give the main process a whole core by enabling only one LP of the first core, as you mention above, what do you recommend? Hi Mike, Same as with eight cores - four LPs one per core, this is the most efficient use of the current CPUs, which can be seen within the AMs I specified for the eight core above. With the eight core (and six core) there's spare cores for the jobscheduler to target with the processes of addon exe apps and other work arising. With only four cores and a sim that wants four cores, those exe apps and other work will have to be done on the cores of the sim. Use a bat file to start addon exe apps on 10,10,00,00 and use an AM=85=01,01,01,01. You can hand over more CPU throughput to addons with an AM=116=01,11,01,00 and putting addons on 00,00,00,11 for which the sim still performs well on only three cores. Adding an AM is actually a necessity for affinity handling apps such as P3D especially when HT is enabled, and even more so as the core counts go up. However a poorly specified AM can reduce the number of LPs the sim starts processes on and so if used incorrectly can make a blurry mess. With the eight core I showed we could use six cores of '01's, this increases scenario loading speed and so can reduce the chance of blurriness in the land textures. We can't do that with only four cores but we can bring the efficiency of HT to increase the loading speed with for example an AM=11,11,01,01=245 where we have enabled six LPs, the first two on the right have a core to themselves '01,01' since they can't split out over more LPs anyway. However the second pair of cores have two LPs enabled per core, this enables the scenery loading to split out and utilise two pairs of LPs loading data, but saturates the HT core during loading - this may not be so bad even if we have addon exe's restricted to those two last cores as with 10,10,00,00 mentioned. Anyway that's a problem for those with only four cores, with six or more, it's easier to keep addon exe interference at bay. Don't be fooled by a little tiny exe app using only 6% CPU time, it can stagger the sim just as any app can. Since most exe addons are also SimConnect apps they create a SimConnect client process on the sim cores loading them up further and adding to networking overhead. These processes work best with two or more LPs as they can be forced to wait too often on a single LP, you can see I specified two LPs for addons, and we can put many addons on the same two LPs.. Remember that fps is generated within the first process from the selection of objects and the frequency of the core, this rate continues irrespective of background throughput. We can't see the difference with CPU setups when the GPU is flat-lined on too much Anti-Aliasing for example, VSync and eye candy should be turned off when comparing CPU setups and AMs. Steve Waite: Engineer at codelegend.com
June 6, 20179 yr Oki thank you then Jiri. I got GTX 770 with tiny 2G I guess I better not touch anything. I got plenty of eye candy right now. I really don't want to ruin my frames, again. Since I've just manage to turn them back to 50+ on the ground, with quite high settings. Hope I won't jinx it again. My i5-3570K CPU @ 3.40GHz is also overclocked (I believe it's on 4G now) but it goes little over 40°. Some Zalman regular fan. Say, this must be some kind of typo right ? Quote 8GB video memory for value 8 is absolutely minimum ?? Since you say: Quote Default + detailed textures checkbox do same as TSE=9 and it nedds about 4GB VRAM It can't be that TSE=8 needs double more. And we all use those TSE=8 values, even guys with weaker cards than mine. FreeMeshX Global & P3D Tweak Assistant (PTA) plus a bit of orbx magic
June 6, 20179 yr Commercial Member 12 minutes ago, glider1 said: I used to do that too for a long long time I was one of the early disciples of it. Since then, I redid tests and there is no performance gain from doing it my end. I also tested core temps since core 0 is working so hard but I saw no lowering of core 0 temps even after doing this. I think SteveW has said in the past, it is an illusion because the usage graph doesn't actually measure true core usage. Here is my conclusion then: The ultimate best way to gauge core usage is to monitor core temperatures. The usage graphs are misleading. That is my understanding. That's correct, If we start P3D without an AM, and then in Task Manager uncheck the first CPU box, those threads of the main process, the UI and the scene rendering on that LP are moved off of it and spread across the remaining cores by the jobscheduler to other LPs, this does not produce a result we desire. Heat = work done, can't say fairer than that. Steve Waite: Engineer at codelegend.com
June 6, 20179 yr Commercial Member 37 minutes ago, denali said: I've been using Process Lasso to set all other programs that use higher CPU percentage to use all other cores, and run P3D4 on high priority, some mem tweaks in there too. Be careful when altering the priority of processes. generally the operating system get's things right as often many processes need to finish in the right order to be efficient. If we set P3D to an overly high priority other process that P3D relies on can be forced to wait unnecessarily, and P3D in turn. Steve Waite: Engineer at codelegend.com
June 6, 20179 yr 3 minutes ago, gamer19 said: Oki thank you then Jiri. I got GTX 770 with tiny 2G I guess I better not touch anything. I got plenty of eye candy right now. I really don't want to ruin my frames, again. Since I've just manage to turn them back to 50+ on the ground, with quite high settings. Hope I won't jinx it again. My i5-3570K CPU @ 3.40GHz is also overclocked (I believe it's on 4G now) but it goes little over 40°. Some Zalman regular fan. Say, this must be some kind of typo right ? ?? Since you say: It can't be that TSE=8 needs double more. And we all use those TSE=8 values, even guys with weaker cards than mine. When I keep my CPU on stock settings it does not go over 70°C at simulator load. But look at TDP it sits about 130W avg but CPU is rated as 88W TDP. If I set that limit, I can sit under 60° but at 4-4.1GHz. Just google for 4790k high tems and you find many who reporst temps close to 80 on default clocks. TEXTURE_SIZE_EXP is set to 8 by default. And sim loads textures 256x256, for me it means that sime eats something between 2 and 3GB VRAM at 4k. With TSE=9 or default and checked checkbox in settings it goes to 8, with TSE=10 I go close to 8GB easily. i7-8700k, RTX2070 Super, custom water cooling, 32GB RAM, 6TB SSD spacePrepar3D v4, X-Plane 11, 40" 4k TVRex, Active Sky, RealAir and A2A pilot, GTN, ORBX mostly everywhere
June 7, 20179 yr Commercial Member 6 hours ago, 777200lrf said: In fsx I could set affinity mask. In p3dv4 if I set affinity mask my scenery all around is very blurry. Am I doing something incorrect? FSX and P3D are similar in many respects but are quite different whereby P3D can have more intense scenes than FSX and so poor loading bandwidth can cause problems such as blurriness. An AM used with FSX may provide a good balance with background performance on FSX but the same setup might not be good enough for P3D scenes. P3D operates in a slightly different way to FSX with less impact on the main process when adding more than four LPs, more than four add to background performace. Steve Waite: Engineer at codelegend.com
June 7, 20179 yr Steve, I just spent about an hour testing out various solutions on my CPU, which is identical to OP's. I saw your AM 5460, which I believe renders the game on the 6 "middle" physical cores, leaving the first and last unused. On one hand, it actually provided an overall FPS boost or rather, an FPS stability. Flying over Copenhagen, alongside FlyTampa's EKCH, there's a lot to render and a lot going on. On the other hand, for the first time, v4 gave me the dreaded blurries. - Running completely default, 8 cores, 16 threads, FPS would be choppy in the 30-38 region in a jet around Copenhagen. - Using Process Lasso to remove the first logical core, leaving the first physical core to deal with the game alone, was a boost in smoothness and around 34-39 around Copenhagen. - AM 5460 would be very close to my constant locked 40 (with Rivatuner), I reckon an extra stutter when it loaded huge chunks of EKCH but I saw quite a lot of blurries. - I then suspected the age old problem of moving rendering off the first core and thread is the reason for the blurries. So I tried AM 65532, which leaves out Core 0 / 1. Again, I saw blurries but also lower FPS due to logical cores being used, again seeing down to 30. - I tried AM 21845, which is basically turning HT off via masks but that was neither smooth nor gave me high framerates. Bit choppy and FPS was sailing from 32 up to 38. - Then I tried completely turning off HT and no AM, which surprisingly seems to be the best solution. No blurries and lowest FPS was 38, almost kept a constant 40. So I think I might run with HT off from here on out. [MSI MPG X870E Carbon | 9800X3D (PBO +200Mhz / -20 Offset) | Corsair 64GB DDR5 (Custom Timings) | RTX 4090 Founders Edition (Undervolted) | WD SNX 850X 4TB + 4TB | Antec Flux Pro]
Archived
This topic is now archived and is closed to further replies.