Jump to content


  • Content Count

  • Joined

  • Last visited

  • Medals

Posts posted by dnk

  1. ^ makes it sound like you've been banned a few times...

    There is no need to "strike a balance" between this unrealistic recoil and real recoil. We need the real recoil and not this COD or BF like recoil. Arma games have always leaned over towards realism over balance and there's no need to change that.

    There is really no need to create a thread as 500+ people have already voted on and discussed it here: http://feedback.arma3.com/view.php?id=2425

    Fully disagree. Forums are for discussion, feedback tracker is for keeping tabs of issues. This is a discussion, so this forum is the appropriate place for it.
    Regarding recoil,
    Well said. I personally prefer accuracy-based realism over recoil anyway. I love the America's Army system (at least the original - haven't played in a long while) where weapons all had fairly high (by Arma standards) built-in inaccuracy. Even if you put the cursor/sight right on a target, you'd get a like 5-degree Gaussian deviation at best.

    With recoil, in real life it's not as hard to instantly counteract in full-auto to keep the weapon on-target (well, pointed in the right direction, give or take a bit) than it is with a mouse, especially with video lag thrown in. Your muscles can handle a lot of that by themselves, but in the game you have to constantly be using your eyes instead to track where the barrel is pointing and relate that to your desired sight picture, then relate that to delicate hand/wrist motions, and by the time you've started to counteract the most recent recoil movement (if it's random in direction/amplitude), so many milliseconds have passed and it's moved somewhere new, and it does feel very much like "fighting the avatar" rather than "real recoil". The division of time in a game is just so much more grainy than the more fluid non-division in reality - it's always going to be hard to do recoil via video output.

    Which is why I'd just settle for real inaccuracy, and have that increase substantially for X milliseconds after a shot is fired, cumulative, up to some hard upper limit. So, each weapon has a set base inaccuracy (much higher than current, since it's not only weapon-specific MOA but also the shooter's non-super-delta-force-ability/stress), which is adjusted for stance, health and fatigue. Then, each weapon also has an amount of inaccuracy increase per shot. Then each weapon also has a set reduction of inaccuracy per second, as well as a maximum inaccuracy (you're never going to accidentally shoot 90-degrees to your intended target due to recoil).


    MX base inaccuracy = 4-deg (modified by health, fatigue, stance, and skill)

    MX shot increase = 6-deg (modified by health, fatigue, stance, and skill)

    MX decrease/sec = 4-deg 1/(modified by health, fatigue, stance, and skill) [might be adjusted more often than 1 second, but this is the cumulative 1-sec reduction]

    MX max = 14-deg (modified by health, fatigue, stance, and skill)

    So, it takes about 3 seconds to return from full auto to a "perfectly" aimed shot. Any single shot will take 1.5 seconds to fully recover from back to a "perfect" shot. You can adjust the values according to how much you think recoil affects attaining a "perfectly" aimed shot (given battlefield conditions, shooter error, etc), but I think this is clearly the best way to handle it.

    Proposed modifiers:

    "hold breath" (total)x(0.60)

    health (1.0 - 0.01) >> (x1.0 - x1.5)

    fatigue (0.0 - 1.0) >> (x1.0 - x1.5)

    stance (stand, crouch, prone) >> (x1.0, x0.67, x0.33)

    skill (0.01 - 1.0) >> (x2.0 - x1.0)

    So, a crouched man at 0.20 fatigue and 0.92 health with an "aimingskill" of 0.5, who has fired 4 shots in the past 3 seconds on an MX, without holding his breath, would have a Gaussian dispersion with 92% of shots falling within this many degrees of the cursor:

    (base: 4) x (x1.1 x1.04 x0.67 x1.5) + 4x(shot: 6) x (x1.1 x1.04 x0.67 x1.5) - 3x(reduce: 4) x (x1.1 x1.04 x0.67 x1.5)

    = 4.6 + 27.6 - 13.8

    = 18.40 degrees

    = 14 degrees (max)

    That makes way more sense and would best serve both sides of the debate, I think. You could add a camera-only shake/movement for each shot (that doesn't move the cursor, and which will automatically return the camera to be centered on the cursor after X amount of time). That would add to immersion and at least be slightly disorienting vis-a-vis the sight picture, without requiring constant micro-adjustments of the mouse.

  2. Deploying a weapon / proper use of bipods - probably a lot harder with the new animation system and lack of collision on weapons than it was in A2. Also, I've done it in A3 already.

    Low CPU/GPU Utilization - seems like there have been a lot of improvements in performance in the past 6 months. Did you ever consider that this might just not be easily fixed? That it might be something they can't do much about without a total engine rewrite? What do you want BI to say if that's the case, "sorry, our engine ultimately is somewhat flawed?" In what universe do companies do this, and should anyone blame a company for not taking a shotgun to its own foot?

    Firing from vehicles - perhaps this adds ~0 to gameplay but would take 50+ hours to implement

    Feature Request: 3D Optics using Picture-In-Picture engine capability. (Red Orchestra Style) - they came up with a compromise solution. That wasn't enough, only 100% perfect is to be accepted by the community!

    Add ability to climb onto/over objects - in another thread a dev said that he was working on this and it would be released eventually

    Some sort of melee? - adds ~0 to gameplay and would take 50+ hours to implement - it's a modern war simulator where melee is used about 0.1% of the time...

    Realistic Wounding System - seems like this is already partially implemented (leg wounds restrict movement, arm wounds increase sway), and the full implementation (with UI changes) would be unnecessary, excessive, or even put off a lot of the fans (who don't like a UI cluttered with gamey graphs). Ultimately, if you break a bone, you're out of the fight, and for this game that means "dead". Also, if there was serious community desire for this, it could be modded easily enough I'd think (with certain optimizations).

    Bullet-in-chamber accounting - pretty exceptionally minor issue, no? Should be able to be modded easily if you really want it. Perhaps you can just ask the devs to add a script function to allow you to check the number of rounds in a magazine and adjust them through another function (check_magazine_quantity & set_magazine_quantity). Modding it would be easy as 1,2,3 then.

    Grenade throwing is unrealistic (too fast) - the devs disagree with you on realism-gameplay balance perhaps?

    [Feature request] Fast Roping - I've done it in A3 already.

    I would like the others, but again a lot of it comes down to (resources / tasks) and much of this just requires too much of the former while BI has too much of the latter already on its plate. Also, you skipped the part where we've passed the alpha stage and new features are very unlikely. At this point, you're mostly arguing for the next game's features. Name other FPS games that add new features like these post-release that weren't already WIP at release. BI got swamped during the runup to release, they didn't have the resources to take on all these new community ideas, and that's really it. I'm not sure you understand the software development cycle clearly enough...
    [snippies] you're either a hater or a fanboy.

    { insanely disrespectful and full of insult tantrum that has nothing constructive }

    Go find some of those circles in my/others logic and, you know, argue cogently if you disagree. Talking OT shit about random people without being constructive is just derailing flamebaiting. You doing that here?

    If someone is going to come here and make false claims & throw around the words "haters" and "fanboys" I'm going to say something. Because it's bullshit.

    On topic:

    There's plenty of great Mp modes coming. I for one am looking forward to tactical battlefield.

    Yeah, but BI didn't make it, so it doesn't count. Nothing counts unless BI themselves do it. That's the argument. Many of the complaints of BI inaction (on features) have been, are being, or can easily be solved by modders, but BI didn't do it, so they're terrible at pleasing customers... A lot of new (and some old) players not realizing that for this series the community is easily 50% of the development team (especially for scenarios), and complaining because they feel the paid-developers are taking advantage of the modders. Or it's (some) modders complaining because BI won't hire (at the expense of firing a working dev) a direct mod-community representative to keep them up to date on everything the dev team is doing. That or BI just won't confess to failure so these same people can attack them relentlessly for it, as they do for every minor perceived slight and failure up till now.

    The idea that BI has very limited resources and has spent them mostly for this release, which is finished feature-wise, and that the extremely long list of community ideas that came up 6-12 months ago was almost totally unimplementable given their budget/staff/deadlines is not to be considered. I mean, the community asked for the moon, BI clearly couldn't provide it, and now some want to hit them over the head about their inabilities, portraying it as "community betrayal" or whatever.

    There are issues with the game, some players are having big issues, and some features need to be ironed out. I have faith that BI will work in good faith to remedy these situations, but I also recognize reality and the fact that it might take longer than 1 week to fix deeply entrenched issues with the engine or coding.

    The thread is about how "MP is dying fast".
    Ah, never take this stuff at face value. The thread is really about "complain about BI". You can disprove any single point, and in the end it will continue on a new track about "missing features", "poor community interaction", "not pleasing customers", or [insert personal peeve here].

  3. Back on to terrain...

    Why is there always such short groundclutter? In the real world, most natural non-arid habitats have groundclutter well past waist height. A low crouch will make you basically invisible, certainly a prone will make you totally so. Even if only 15% of the groundclutter is X height, after 100m anything under X height is totally concealed. It bugs me to no end how I always feel like I'm running through weed patches instead of actual grasslands, etc.

    Also... New technique for rendering grass at far distance

    It saddens me that after so many retoolings of the game/engine/AI/terrain/etc/etc, this very basic and cringe-worthy terrain masking approach to unit concealment is all we have still.


    It pains me to no end that a game in 1998 had better grass than Arma 3.

  4. On this subject, it would be nice to have a new menu for AI engagement modes. I mean, this is a bit unintuitive, and there seems to be no way to set "white", and also it's not clear to a newcomer what the various modes (stealth-safe) do to this (nothing, but that's not clear - I would think "stealth" would be "green"):

    Combat Modes:

    • "BLUE" (Never fire)
      • When hostile units are detected, they will track them, but will never fire back, even when fired upon.
      • This mode can only be set through the editor or script. No in-game commands to subordinates can set them to combat mode Blue.

      [*] "GREEN" (Hold fire - defend only)

      • When a player orders his units to "Hold fire", the units are set to combat mode Green.

      [*] "WHITE" (Hold fire, engage at will)

      [*] "YELLOW" (Fire at will)

      • If AI spots a target and decides it is in effective range, he will open fire. By default AI is set to YELLOW.
      • If a leader calls Target, the unit with aim without breaking formation.

      [*] "RED" (Fire at will, engage at will)

      • When a leader commands his units to Engage at will, combat mode RED is set. The AI does not keep formation and each member moves individualy. The leader command Disengage will set the units back to fire at will (YELLOW)
      • If Attack and Engage is called, the unit will break formation to find the best place to attack from (combat mode RED).

    Something that's bothered me since the beginning...

    Also, you can't force the AI to fire on someone with the "fire" command if they're out of range (even if that's well within weapon's lethal range). So, a sniper can't fire on a target at 600m no matter what, since he's locked at under 500m.

    Also there's no way to use a "target" mouse command (unless I missed something), only "engage". So, if I have a sniper and want him to target X soldier but stay in place to do so, how? Scroll through the 20 "soldier" entries in the target menu? It's one of those little things that really annoy me.

  5. Something I've noticed with my own build is that once virtual memory gets about 80-85% allocated, the OS/engine stop filling it further. At that point, FPS and GPU memory controller load decouple (they usually track almost 100% the same), though FPS doesn't really get worse.

    Additionally, as the video memory fills up, it has to constantly be partly flushed by the engine to make room for new data. This may have a performance impact, but the real issue might be a lack of bandwidth for the amount of data being loaded into VRAM. You can always try reducing things, particularly object quality and texture quality. That will reduce the size of sudden loads on your system and make memory management easier for the engine/OS.

  6. You must have skipped the underlined part of my post.

    Remember that Steamgraph only keeps track of the people who play the game through steam. 100% of the people who play Arma 3 in MP use Steam to do so. How many % of Arma 2 CO Players start the game through steam? (I believe you can launch a Steam copy of Arma OA without starting steam, am I right?)

    Clearly, most of the new players bought through Steam. Those new players make up a preponderance of OA players because they're specifically DayZ players (that big mid-2012 spike...). I love how this whole argument ignores the undead elephant in the room.

    You can look at swec.se (have I not already linked this in this thread?). It breaks down DayZ/normal players, and it's vastly DayZ players and servers for OA. I believe swec.se includes all online players, not just Steam clients. When you factor this DayZ into the equation, A3 is slightly ahead of OA in MP players(excluding DayZ, since A3 doesn't have it). That's despite not having ACE (for the big private/pub COOP groups), a solid Warfare, more than 1 map (in comparison to 3 maps for OA), and the game very much being a WIP at present. Give it 6 months, and the community will be much stronger, and OA will slowly fall from grace (save DayZ since the standalone will never be released). And that's how it seems to always happen with new BI releases.


  7. Doesn't make sense why the same mission on two different servers has drastically different playability. Here are 2 sessions from today (new main build). Both on the KOTH servers. First server, with 6 players near the round start (X-axis is frame, Y-axis is time between frames) (FPS = 1000/Y-axis) (scale for both graphs are the same):


    Yeah, so that was extremely crappy performance, where there are constant micro and macro stutters. Regularly getting 5FPS or worse spikes. Now, second server, with 25-30 players at the middle of the round (end of round was same):


    Well, that's a solid 22-28FPS (for me, that's decent), not too much stuttering (very playable - I won the round in kill count).

    There weren't many dead vehicles or units in either since it's pure PvP with few vehicles involved. By all means, the second one should've been far worse for that, and for the number of players. My client is the same for both missions, my hardware doesn't change. The pings for both were about the same 250-265 solidly (I also logged these to check). If it's not the server affecting my client, what's causing that horrid play (and that's far from the worst I've seen on this mission - I wasn't kidding about 2min+ "lag")?

    My specs:

    i5 3350P c4 3.1-3.3GHz 32-32-256-6MB cache

    GTS 250 1GB

    2x2GB Crucial 1066Mhz dual channel 7-7-7-16

    Gigabyte B75M

    Win7, up to date

    ~15Mbps line in SE Asia


    1600x900 noAA

    All lowest video


  8. Tested it on the Assault showcase. Can't see much of a difference at all, especially with FPS. New build seems a bit less volatile in a few ways, and the GPU memory was increased more smoothly, but that could just be random. I didn't test more than a couple times for both builds.

    Max CPU/thread use still tracks ~100% with the first core's usage. FPS still tracks ~95% with memory controller load. CPU around 50% +/- 10%, GPU around 100%, VM and physical RAM about 80%. Didn't look like they did much for performance in the SPOTREP.

  9. While there are things to complain about, you're going a bit off the rails, sir.

    I don't recall Arma 2/OA having a huge amount of non-community MP content. Mostly, it was Insurgency, Domi, Evo, Life, Warfare, (later) Wasteland, and DayZ. Currently, in A3, we have Insurgency, Domi, Life, Warfare, Wasteland, and Breaking Point. For every other COOP mission I played, it was community-made by TG or 15thMEU (usually using ACE), who I played with at the time. As has been said, some of these aren't matured yet.

    Not sure what "optimization" means. Everyone throws that word around to mean "I don't like the performance". Seems you triple-counted that one with "stability" and "playability"... I agree these seem to be issues for many in MP, though. Also seems many don't have these issues and are happy to play MP. You might just be one of the unlucky few.

    I think it has been called a "lego box of gaming". You've been around at least 3 years, so why are you suddenly confused about what the Arma series is :D Long-time players can't use the "ignorant noob" card...

    BIS is constantly working on the game. Saying they just "leave it for the community to fix" is just clearly false. This game has an extreme amount of post-release support for an FPS, on the order of MP-only MMOs in terms of that support. It doesn't come out of the box as shiny and defect-free as those, but it also has like 1/50th the budget with the same amount of features, if not more.

    Arma 1 and 2 had the same sorts of initial shortcomings. This is an old song and dance, been going on for quite some time now. It's a game/community that you invest in. It pays out over years. If you want the newest game-du-jour to play for 2 months and ditch, then Arma isn't for you, or you should wait a year or so to buy it on sale when it's nice and patched/modded up. That's probably true for most any game these days anyway.

    Map full of nothing? :D What did Chernarus and Taki have exactly? This map has more than either. What do you want, exactly? Say it, say precisely what would make the map "not full of nothing", given that it's already full of farms, towns, military outposts, and cities. You want BF4 metropolises or something? You want a megacity map? You literally could see exactly what you were getting about a year before release when BI said the exact real-world island the game was based off of. If that was too "empty" for you, why did you buy the game? :D

    Good God, man. Complaining about yourself more like it.

  10. Whatever BI does for their testbed server, I seriously hope they include a script that logs each clients FPS as well. I would assume client-side issues are connected to server-side issues, since there are clearly some servers I have no issues with generally and some I do, with the same basic mission (though perhaps some minor mods or a different version is the reason - or perhaps just a 50ms ping difference). I still think my issues come from ping, and that there may be significant degredation of client-side performance past a certain ping threshold, though I have no concrete way to test this in a controlled environment (so I hope the dev testbed servers can). I should also note that my connection sometimes has microjitters, as I call them in connectivity due to poor Internet infrastructure in my country. This may be getting blown up into these fairly massive lag spikes and frame freezes by the engine/netcode. This may have smaller, harder-to-notice effects for other players.

    For helping find an issue, I suppose it's good to have severe problems others mostly lack, since it might be that these problems exist for everyone, but in far more minimal forms.

    I would also suggest BI, with its testbed servers, create servers in different countries (Cali, UK, Aus, Ger, Rus), then they can compare client performance (assuming players use different servers at times) by their ping, etc, to see if that really is a problem (and it may be causing server-side issues as well - I don't understand how you can isolate one from the other here...).

  11. BUG:

    I got a helicopter support call. I loaded into the chopper, then went to give a new LZ. I ended up clicking several times in the area I wanted since the LZ script takes a while to set the LZ. The spot was on the far west end of the island (I was at base in the east). The chopper flew about halfway, then made a big circle, then flew back to base and hovered over base, then repeated the prior move-circle-return.

  12. Client-side I find it fairly unplayable on most servers. It seems to be pure luck when I can and can't get a reasonable game out of a server/mission, and it's not always player count, though certainly that has an impact. Nothing like 2-minute "lag spikes", intermixed with 500-1500ms lag spikes constantly with just 15 people on PvP. Seems most don't have this issue, so it might be due to my high ping (250-300ms generally).

    Or I play Domi with a 50ms jitter every other second with just 10 guys on. Doesn't happen in SP. Again, might just be me, and might just be due to my high ping, but I don't recall these issues A2/OA at all. By all means, same ping I'd get pretty good FPS with minimal spikes/jitter on any sized server. Played a lot of Warfare then with 30-40 players and all the AI to go with. It would slow, but never with all this stop-and-go nonsense.

    Then there are servers without issue, even with 30 players on sometimes (and other times not). I don't get it.

    ---------- Post added at 10:34 PM ---------- Previous post was at 10:31 PM ----------

    You din't get him right. What he mean is that if towns affect your FPS (your case), then you are NOT CPU bound.
    How do you figure that? Seems towns have an awful lot of geometry per scene.

  13. Thanks for all that. I'll assume your graphics card has 1GB, has fairly recent drivers on it, and your RAM timings are C9-11 thereabouts.

    Have you tried Process Explorer yet? It requires no installation. Just download it, run it, and maybe set the view\update speed: to 5 or 10 seconds for a "big picture" view of the performance. This program can also tell you page faults, working set and private bytes for the program, disk usage, network usage, and gives a core-by-core performance rundown of the CPU. It's not a silver bullet, but it might help spot the exact issue. I'll upload my own results from an hour or more of play in a bit (once I play it :)), for sake of comparison.

    Also would help to have FRAPS running at the same time, making sure to check FPS\Benchmark Settings: Frametimes. This will allow you to quickly create an Excel-based FPS graph you can use to roughly correlate the Process Explorer data with the framerate data. Then run a benchmark once you start the game. Finish it whenever you like, and it will produce a log for that Excel graph.

    For more information, you can get Process Monitor (same website as above). Add 2 filters to it: "path","containts",[Arma 3 directory] AND "path","is","C:\pagefile.sys" (or wherever yours is). This will give you a timeline of disk reads, basically. Then run it when you start the game. Stop it once you finish, and it'll maybe help you to see what's going on with the file reads as you play. This will create a report you can read to relate via timestamps to periods of high disk usage or low FPS from the other 2 monitoring programs.

    For more, more information, use hwinfo. It can log a ton of data (though I'd set the refresh rate at like 1sec+ since it kills FPS in the process).

    Again, this all may lead to nothing, but you might at least better understand the problem.

  14. Opening doors should AT LEAST always be the first command, so I can quick scroll my mouse up and get that instead of something pointless like "reload 556 magazine" (why is this even an action?).

    And removing pointless actions from the scroll menu (like reload current magazine type, inventory, and switch to secondary). We have keys for these things, why also have actions? It pointlessly messes up the action menu.

  15. My complaints were mostly about the AI and playstyles than mission makers. Most of my desired mission changes would not appeal to typical playstyles (no magnified optics, weaker skilled AI, no CAS... the community does not support these things generally), so I don't bother the mission makers with my desires. They've done what they can - given the community wants X, you can't force Y on them.

    I would ask for far more challenging AI that, you know, engages the player instead of waiting to be killed, but this requires a lot of extra scripting, lots of AI units, which cause performance problems. Or something where the player isn't in a permanent position of domination over otherwise helpless (but superhuman) AI. Unfortunately, the big box COOPs tend towards "humans should have totally dominant firepower and unassailable position" OR in cases (insurgency) where humans have to actually be put in harms way to clear things, the lackluster urban AI issues come to the fore. AI just sitting in rooms, lying on the floor, waiting to be found - a long, tedious exercise of Easter egg hunting with M4s, where occasionally you get a curve ball and it gets fun for a minute before the long search resumes. It's not the mission maker's fault - given X resources, you can't create Y.

    I mean, it seems the large-pub COOP community mostly wants to sit on a hill and shoot fish in a barrel with lots of armor/CAS support, and no one minds at all if 1 player hops in an aircraft and levels the AO first. This is what I've seen through countless hours of COOP play among many different servers. I don't get it, that's all. It's dreary to me. I do like PvP, but it has its own problems - namely, balancing teams (KotH = 6v14v14 or such often, leaving newcomers to choose "join the losing side" or "have fun"?).

    And Warfare just hasn't matured yet, as you say. It always had issues with too-hard AI, and for that I lay the blame more at BI's feet for making defaults so insanely high and forcing certain minimum levels (insanely high) for a given difficulty setting (0.7 for Vet is ridiculous).

  16. I agree with the general feeling that there aren't enough mission types, or those that exist are lacking in excitement.

    Regarding COOP, as much as I enjoy it with a good, small squad, it's the most pointless and dull waste of time with more than 15 people on current missions. This is especially true since the current community culture on most of these servers is: snipe/CAS 95% of the AI from a 100% safe distance. Half the time a good CQC mission is ruined because 1 guy in a AH-99 or whatever gets to the AO 10 minutes ahead of the crowd and levels it for a LEET KILL COUNT of 200 while the rest of the server has like 15 per. Of course, since the rest just wanted to shoot fish in a barrel, who cares anyway?

    This could be fixed through improved AI (that aren't terrible at a distance or in urban settings), improved scripting (same), or better mission design (more sniper and AA units). I would bet this snipe-a-thon approach would be greatly diminished if 30% of the AI in a city were snipers (forcing the player to close and use cover/defilade effectively in doing so), and simply eliminating CAS should be done for sure. But then missions would actually be challenging, which many of the large pub players seem to hate (and given some of the respawn systems - 10min time to return to AO - it's understandable I guess).

    Of course, the fact that many missions use too-accurate AI because designers think 0.50-0.70 skill is acceptable makes sniping from a safe distance the only way to play without being instakilled when under 250m away.

    Smaller pubs are more fun for COOP, especially when the players enjoy a challenge, but even there a lot of missions end up being "sit on a hill and shoot into the barrel", which is so extremely tiresome.

    Additionally, I've not felt BECTI to be nearly as fun as it is in A2 right now, so that's also eliminated a very popular and challenging PvP mode from the lineup until it gets into better shape.

    1. Can't tell AI to gear at a specific spot, and can't tell them to use inventory at no spot in particular.
    2. Can't tell AI to switch to their primary (big issue after using inventory is often they have the pistol selected.
    3. AI kind of suck at following me in urban settings (need an urban formation where they just follow my path (as in, not going through adjacent yards/alleys when I'm walking down the street) closely, taking up positions along that path as the formation stops (like on walls, corners, etc - anything within 3m of the path). So, so, so much micro needed to get AI to properly traverse urban areas right now.
    4. AI get stuck a lot still (might be mods/scripts?), and it's hard to unstick them (though possible so far).

  17. You know, it helps to know more details. Like, your system specs, graphical settings, OS, mods used, etc...

    Are you watching your physical and graphical memory usage during play?

    I do with process explorer, and I find that typically the former fills up quickly, while the latter slowly builds up over a half hour or so up to about 85% GPU RAM usage (faster if I move around a lot or have a high VD), at which point it starts cycling up and down. If I stay on "low" texture size, this takes quite a long time to happen, but on "standard" or higher, it fills up really quickly and then will certainly start causing stuttering or slowdowns (I think) as older files are removed and then need to be reread as I move about (causing latency/bandwidth-related slowdowns from the load onto VRAM).

    I would check that as well. As you go across the island, you keep needing more and more textures/models. I'm not sure, but it might start to fill up the virtual address space or physical RAM, and this might cause the engine to start (through dynamic resource management or something?) loading smaller files (lower LODs), even if its regularly clearing out unneeded files from earlier in the session. I think it's a bit simplistic to call this a "memory leak" - more like you just keep adding in more and more models/textures, and the cup fills up eventually.