Jump to content


  • Content Count

  • Joined

  • Last visited

  • Medals

Everything posted by Polymath820

  1. This is just god aweful. Look at FPS plotted over time... really what the hell bohemia... I thought it might be my system I thought it might be something else. Nope it is definitely the RV engine itself. Fix it please. http://www.hardocp.com/article/2013/10/02/arma_iii_video_card_performance_iq_review/1 Before releasing ANYMORE DLC's you really need to fix the problems with the engine currently instead of filling ArmA 3 with bloat-ware that causes more problems! Look at the data. It speaks for itself. http://www.techspot.com/news/54026-arma-3-benchmarked-graphics-and-cpu-performance-tested.html
  2. Polymath820

    Not biting my Tounge anymore

    "High level of competence" Are you joking? You are just going to justify a judgement based on voluminous data? From authors or website Authors. I know for fact websites can be paid to give testimonials and can be paid in perks or money to manipulate data, it frequently happens the term is "appealing to an authority" You appeal to the authority that the information they give you is fact. There is inconsistent data from multiple systems right across the board I've seen I3's performing better on ArmA 3 then an I7. So what is it? Our bottleneck or yours inside of your code? That only shows itself in the event of slightly different micro-architecture execution environments and system components? Oh look setting all settings as low as they can go and setting view distance to maximum hmmm not much core usage difference across the board at all. And GPU usage well that drops to 0 - 14% So tell me again, the game relies on the CPU? then wouldn't ArmA 3 be utilising the multi-cores more? The only thing I have noticed the CPU being eaten by is AI. And just for the record I do not trust anyone including scientists sometimes. Cross-matching and cross-referencing helps. Unless of course a website is endorsed by a scientific community that has stringent quality control. Argue parts Argue people are idiots, but you can't argue you got your recommended specifications wrong and owe people a big apology. I would never have bought ArmA 3 if I knew my system would not handle it. I could have put the money towards something else. I have done extensive performance analysis and my GPU I think is the bottleneck but I'd rather know now and completely know because of the inconsistent data everywhere that if I "do" upgrade my GPU due to my CPU being used little or none, I will get better frame-rates but I don't feel like blowing $400 on a new card if it does nothing.
  3. I would like bohemia to make it so a sector control can behave like a trigger. It already does partially with controlling objectives etc. But lets take it one step further and make it so you can synchronise it to any module including post process. Meaning a "execute if linked components true and go as far as adding an "Conditional module" making it so we can alter the behaviour of the sector control using AND OR NOT NAND etc. In scenario flow.
  4. Polymath820

    Anti-Extasy hack

    Yeah to be a real-hacker you have to know knowledge about electrical engineering and computer science. The formal definition of a hacker is: "A person who modifies a piece of technology to change it for which it was intended" The first account of computer hackers actually was the MIT model railway club where they "hacked" some old phone equipment to switch parts of the track by dialing the number like an ID'ing system. These days hacking is looked in 3 ways: Hacking: Actually know how the code works and understand it fully Cracking: Know how the code works sometimes and uses it for nefarious purposes Script Kiddies: Annoyances or pests. And For the sake of pointing out how funny Script kiddies are: http://gizmodo.com/5624179/microsoft-script-kiddies-and-hackers-accidentally-send-us-their-code-all-the-time
  5. I've been doing some very rudimentary experiments with AI spawning and found a better way to micro-manage missions and prevent large FPS hits. It turns out if you add conditions that after a "group" of AI are dead then wait a certain amount of seconds spawn them and unhide them. When you hide them they disappear from the scene and their simulation is postponed, now what I did was simple enough that conditionallly when AI Group(A) AND AI Group(B) AND AI GROUP© are all dead then spawn units , and wait this way there is always the AI there but not there they are in a "suspended state" this combined with description.ext for wreck and corpse removal almost eliminates all FPS problems with AI as far as I can tell. All I did was created 4 infantry groups 2 Blufor Fire-teams and 2 Opfor Fire-teams 1 Opfor tank and 1 Blufor tank as soon as all the AI were dead in that group and that condition fell silent it would spawn in the next group of AI so instead of spawning all AI at once hiding them "disable simulation / hide from object scene" removes stress on the client from the AI side and the server because after and only after the "groups of spawned" AI are dead they are hidden object simulation disabled, then cleaned up after the wreck timeout in description.ext finishes. So steps to create "simple simulation manager": 1. Create 2 groups of AI 2. Create a Trigger of "When no body is present" execute condition of !(alive ai1) && !(alive ai2) && !(alive ai3) etc. 3. Then execute synchronised module of "hideObjects" under object modifiers 4. Play the mission use cursorTarget setDamage 1; on the objects that are linked to the "alive condition" 5. and as soon as those objects are dead. 6. Hide them and disable simulation frame-rate jump of upto from initial testing 15 - 20 frames See Attached: After Fighting AI: http://i.imgur.com/vBup8A5.jpg (150 kB) While AI Fighting: http://i.imgur.com/F7M0Ai9.jpg (163 kB) Mission Structure: http://i.imgur.com/yni44wN.png (159 kB) The one problem appears to be that the sound loops attached to vehicles are not disabled as well.
  6. An additional piece of information was to make AI spawning more efficient I saw a need to create "AI Processing Pipelines" so for example: When the AI in the scene are completed spawning and after they all die. But this was made more efficient by basing it off the fact that Functions in ArmA 3 are non-scheduled environments and AI spawning is similar to one lump effect so I thought, why not separate the system into pieces: Infantry AI Pipeline (1) Vehicle AI Pipeline (2) Empty Vehicle Pipeline (3) Now after AI have spawned a request is made to the pipelines to line up N number of AI groups or units When N reaches it's limit it maintains the AI Queue at that value which is set to a "safe limit" Then when all the AI are dead in the last spawn cycle the processing queue lets through a few AI groups checks if it's beyond the safety limit and waits. Then once that AI group is dead request the next from the pipelines. A simple Dia diagram has been provided: http://i.imgur.com/Jy3RR5x.png So the best way to describe it would be to create a simple non-scheduled function using pre-compiled .sqf which does the spawning and an execVM which makes calls to 3 other execVM .sqf's Additional Diagrams Provided on the intricate details of the system. Including the pipeline manager and scene manager Other intricate details as stated above would be to use a non-scheduled environment for a re-usable function for spawning units but separate execVM's and scripts for each of the spawning pipelines I know that non-scheduled environments are bad due to them be executed in the fastest possible time but if they are fast and small which don't depend on too many sub-calls their efficiency can be taken full advantage of. http://i.imgur.com/HZNUdqC.png (212 kB)
  7. Polymath820

    Any ideas why this won't work?

    https://community.bistudio.com/wiki/Task TaskState is not expressed as what you may think, it is "TaskDone" so you are using the == comparison operator on a boolean expression that will never fall true. Because you are doing TaskState task8 == "SUCCEEDED" when it should be comparing TaskState task8 == "TaskDone" Because your condition never falls true it never executes because you have 2 true conditions which accounts for a "half truth && is based on boolean algerbra Namely it's similar to multiplication what is 1 * 0 = 0? what is 0 * 1 = 0? 1 * 1 = 1 and 0 * 0 = 0? so in your expression the Boolean break down is All of the AND conditions must be met. The JetEngine is OFF(TRUE) JET IS IN LIST(TRUE) TaskState IS SUCCEEDED(FALSE) 1 * 1 * 0 = 0 therefore no true condition occurs at all. Therefore your code is not executed.
  8. Are you just wanting them to stay in 1 place?
  9. Polymath820

    FX6300 / HD7850 and low FPS

    http://products.amd.com/en-us/GraphicCardDetail.aspx?id=309&f1=&f2=&f3=&f4=&f5=&f6=&f7=&f8=&f9=&f10=&f11=&f12=&f13=&f14=&f15=&f16=&f17=&f18=&f19=&f20=&f21=& If you look on AMDs website they state the video-card supports 128GB/a Memory bandwidth this simply put means how fast the video-card can move data in and out of the memory chips to the inner processing die. That number is best to be larger. For example My current GTX 650 OC is running at 6008Mhz and your video-card at 4008Mhz your video card compensates for the smaller memory clock by having a larger bit width interface. Now there is also another key thing about your video card it has less TMUS(Texture mapping units) simply put it helps the video card process textures better. Faster memory clock means it is capable of dealing with anti-aliasing better as well Now you say the GTX 770 is more expensive it has both higher memory clock and a higher core processor clock nearly 3000Mhz more than the AMD and more TMUS it also has an extra 400 processor cores than the AMD. And your AMD CPU from a technical stand point nothing wrong with it. 12 pipelines is much faster than 14 in Intel CPUs http://videocardz.com/ati/radeon-7000/radeon-7850 http://videocardz.com/nvidia/geforce-700/geforce-gtx-770 You are likely to get benefit in arma 3 by enabling the use of your threads and cores by doing this in steam -exThreads=6 and -cpuCount=6 see if that improves anything I know ArmA 3 has auto detection for quad-cores etc but I don't think hexa-cores are. On a side note people expect huge battles in arma 3 but that being said a large amount of missions are bogged down with unnecessary scripts. Which cause large performance issues find a mission where the mission maker manages the mission correctly such as not revealing AI until the player is close enough an simple Idea is to conceal them as they spawn hiding the fact they just popped in. And when certain groups of AI die reveal the next lot and so on. So the players screen is not overwhelmed and the server is not either. Kind of sad Bohemia doesn't let you invoke the simulation manager in a more dynamic way other than the module.
  10. Polymath820

    FX6300 / HD7850 and low FPS

    That video-card would be great if you were using it as mainly a work-station rendering side of things etc. But AMD: http://www.hwcompare.com/14638/geforce-gtx-770-vs-radeon-hd-7850/ Doesn't have the real-time power you need. Simply put HyperTransport is a bus-communication method that does not use the FSB and share bandwidth with the FSB to communicate with the Memory instead it's a direct communication point to point serial communication to the Bridge (Main bridge) which is where the South side and North-side bridge exist. It's recent revision supports 2.6Ghz. (It potentially might make a difference) due to the smaller L1 cache a HT does 16 bit high speed serial communication full-duplex (both directions) so removing load from transfering large amounts of data to the L1 cache through the FSB may or may not release some stress. This constant comparison about CPU's speed as an explanation for system performance is excessively simple, as CPU's when comparing an intel to AMD is silly they use both different internal architectures and they also use different sized caches. Noticable differences between intel CPU's and AMD are Shared instruction L1 and base instruction L1 16KB Instruction caches * 6 for the guys CPU above. 3 x 64 KB shared instruction caches L1 caches as a general rule a bigger L1 compared with L2 or L3 is better. When you compare intel CPU's to AMD's Intel CPU's use approximately 14 Pipelines per 1 Hz cycle per instruction meaning in 1 cpu cycle an instruction goes through 14 breakdowns, an AMD on the other hand goes through 12 pipelines so in that sense an AMD can do 20% more operations compared to Intel. "Technically". Intel CPU's also use AVX ( Advanced Vector Extensions) and SSE for this you want a CPU that supports the newest versions. There is many details that factor into system performance and Synthetic benchmarks such as "Using an I7-4770k system with a high end video-card" to test FPS is also a type of Synthetic benchmark as the factors involved are both the system environment (components parts) and the software itself. Simply put OC'ing is not as clear cut as you may think in improving performance you need to analyse your system appropriately and upgrade where you see bottlenecks occurring. There is certain details that need to be known to analyse your performance issues: Motherboard type and brand: View Distance (Overall) Preferred View distance: (Object render) Which option causes the largest performance hit: Textures?: Anti-aliasing?: HDR (High Dynamic Range)?: Geometry Count and Vertice?: Shadows?: Ambient Occlusion?(This is basically a mathematical method to approximate the location and ray-scattering of light and will destroy your frames): Dynamic Lights?: Particle Effects?: PIP(Picture in Picture a hidden camera placed on the side of vehicles and in screens)?: Cloud Quality?: Terrain Quality?: Change each of these values independently and monitor CPU / GPU usage of a specific time and note the FPS hit of each option.
  11. Polymath820

    Low CPU utilization & Low FPS

    I mean I am not trying to pretend I know this subject, but I am always willing to know the right information. Always willing to be extremely informed and up to date. But I still do wonder if people are mixing lower end video-cards with high end CPU's. P.S: Maybe Bohemia should eliminate threading where it's not needed. Or look into such a possibility.
  12. Polymath820

    Low CPU utilization & Low FPS

    If you are referring to the finite-state machine scripts. The AI are programmed using pre-calculated decisions. Although, maybe I don't know I've done extensive performance profiling of ArmA 3 using Windows Performance monitor, and configured counters, An interesting thing to note, I am wondering if my system is under-performing due to the GPU usage it is hitting 100% - 80% and my CPU around 40% definitely means my video-card is the bottleneck leveraging Anti-aliasing is also a problem for it it doesn't have enough shader engines or enough overall bandwidth. Something tells me people on here have mixed lower end video-cards with high end CPU's and expect a good result. The memory on the GPU isn't all that important, it's important for texture files etc. but Anything over 1 - 2GB for 1080P is a waste of time. Unless you also have up sampling which still leverages real-time aspects. My video-card is a GTX 650 2GB the memory-bit width is a indicative example of it being a low end video-card the 128bit memory interface. As compared to higher end cards which have 192bit,256bit,384bit. Now if my videocard is taxed to 80 - 100% a video-card upgrade would likely improve the performance of the CPU as the CPU's performance affects the GPU's due to CPU throttling, not bottlenecking where it's trying to pull more data and push more data but the video-cards physical limitations prevent it from doing so and there is no sense in using 100% of the CPU being pushed to it's maximum and over-heating so it throttles to equalise with the GPU. I already understand what FSM's require and how they are done, but couldn't bohemia have used neural-network learning and neural-network AI to teach them the tactics over time? I've done a fair bit of reading on the AI side. https://software.intel.com/en-us/articles/designing-artificial-intelligence-for-games-part-1 https://software.intel.com/en-us/articles/designing-artificial-intelligence-for-games-part-3/ Okay I am starting to think this is bad or arma 3 scripting are doing bad things, scripts are run in "threaded environments" so: why does intel explicitly state that if that is so you shouldn't have 600 - 700 lines of script running... as the task complexity has a negative effect on the threads performance. Because a large amount of sub-systems are relying on other sub-systems. Correct me if I am wrong? But have we all been suckered into believing multi-threading and types of parallisation are the golden bullet to performance problems? And forget that if threading is over-used as with anything it becomes a hindrance than a improvement?
  13. Polymath820

    Low CPU utilization & Low FPS

    It seems like me that was have become desensitised and are now looking for justifications for bohemias inability to optimise a game sufficiently that it can actually run at acceptable frame-rates. I mean there are games that do more work than bohemias game does and they are optimised sufficiently to run on quite a wide range of systems. CryEngine, Unreal-Engine, Unity etc. Why can't ArmA 3 actually have acceptable frame-rate that is maintained? We constantly make excuses for ArmA 3 which is poorly optimised and bohemia has fallen into a behavior rut, of pumping out games which are extremely buggy, unstable, and unoptimised, and the statement that "they do something no one else does". We really need to stop giving excuses to bohemia and get them to fix there crappy engine and game. Because they've been putting it off for nearly 5 years all the way back to operation flashpoint. Stating the game engine is "next gen" nothing about it is even remotely next-gen, bohemia even reduced the graphical fidelity of ArmA 3 dropping to a 16bit HDR rendering compared with ArmA 2's 32bit. And whats more bohemia recently released a patch, that was supposed to add MP fixes for stability. It is amusing irony the patch they released to improve stability made things much worse.
  14. It appear ArmA 3 utilises a lot of disk-reading really big files, ArmA 3 at times is reading 558MB/s now if this is so, it is filling the disk-queue to maximum which it also appear that ArmA 3 has a dip in FPS when doing a lot of these I/O operations, now I am going to induct an idea that if ArmA 3 has an addressable limitation of 32bits then when it reaches 2GB of RAM it is going to shift that into the pagefile, now I am not sure if this works or not but with some tampering of the pagefile "priority system" I was able to get frame-rate increases 2 - 5 frames. I have no "fully tested this" but it appears promising. So it's not CPU ArmA 3 is using all the time it's both CPU and HDD "reading". A likely solution would be to put ArmA 3 specifically on an SSD as it will perform faster than the HDD and slower than RAM but your addressable limitation is what is causing the bottleneck. Which can be over-come by a highspeed disk. It's this "disk queuing" that is causing considerable FPS drops. So not only does ArmA 3 have it's addressable limitation but it requires an SSD to function at full-effectivensss. This is why in the "recommended system specifications it says a "Hybrid(SSD / HDD) / SSD drive" ArmA 3 is basically streaming ENORMOUS files from the HDD at times I am seeing 128KB/s upto 500KB/s+ So as a test someone try putting ArmA 3 on their SSD.
  15. Call Of Duty Ghosts is suffering from performance problems: http://www.reddit.com/r/CODGhosts/comments/1q0zlm/pc_settings_and_tweaks_that_will_greatly_improve/ Battlefield 4 is suffering performance problems: There is literally hundreds of posts... http://www.tomshardware.com/answers/id-1864126/battlefield-performance-issues.html ArmA 3 is suffering performance problems. Correct me if I am wrong, but isn't it a bit suspicious to be seeing this across multiple platforms? And game engines, meaning it implies the code-bottleneck is not in the games them selves? Why are all 3 games suffering effectively the same problems?
  16. Polymath820

    Low CPU utilization & Low FPS

    All of ArmA 3's issues as far as I can see are caused by: http://forums.bistudio.com/showthread.php?176411-ArmA-3-Low-FPS-%28likely-possible-reason%29
  17. Polymath820

    ArmA 3 Low-FPS (likely / possible reason)

    The reason for that is stated above. But ArmA 3 will probably kill the SSD with the amount of reading / writing it does. Could go down the 15K SAS drive route. Yes but the reason ArmA 3 has such poor FPS is because people are using HDD's. It states in the "recommended" requirements that you use an SSD, Hybrid HDD or an SSHD. All you need to do to confirm this is go into windows Adminstrator tools and and find resource manager. Also I recommend windows Sysinternals or ProcessHacker Portable. I/O causes a lot of the FPS drops. ArmA 3 is streaming a lot of data from the HDD the Graphics card appears to be almost completely idle. And CPU around 30 - 40%