Noctrach

@IronMike Would it make sense to discuss the AI defensive behaviours with ED in context of the ongoing missile work? An AIM-54Cmk47 launched between 20 and 40 miles has a Pk of near 0 because the AI will always opt for a split-S defence An AIM-120C at identical parameters will have a Pk of roughly 0.5 because about half the time the AI will just crank and fly into the missile. The exceedingly rare occurrence where the AI doesn't immediately split-S and burn away the Phoenix tends to arrive with a little bit more energy than the AMRAAM on long-to-medium shots. Kinetically there's no reason why the Phoenix would perform as much worse as it does... it's all in the AI response.

I'm not sure if this is pure chaff simulation in singleplayer or the AI's ungodly ability to notch SARH missiles, but the AIM-7M has abysmal performance against chaffing AI when launched outside of 4-5 miles. Missiles launched at high mach, medium altitude, at ranges between 7 and 11 miles against MiG-29s and F-5s have a Pk that is very close to 0%. Attached are 24 such launches. Missiles were launched low to high to prevent aircraft radar being notched at any point. Disabling chaff on AI brings these same shots up to a kill probability of nearly 100%. I think a root cause of the issue here is that AI has perfect awareness of the exact angle it needs to fly to notch a SARH (non-emitting) missile. Where it should only have accurate knowledge of the launch platform radar. This negates any BVR tactics like diving down to prevent look-down notching, as the large amount of dropped chaff coupled with a guaranteed split-second missile notch leads to immediate and permanent decoying. AIM7Chaff5.trk AIM7Chaff3.trk AIM7Chaff4.trk AIM7Chaff.trk AIM7Chaff2.trk AIM7Chaff6.trk

Addendum: There's definitely a factor of the object/LOD mix that is together on-screen. E.g. removing either nearby trees or distant objects from the camera FOV will also completely eliminate the framedrops. All shadows AND terrain object shadows turned off Game in fullscreen (script removed)

Not VR, but I'm seeing very curious behaviour on the rendering of overlapping shadows specifically. This can be either: Object Shadows drawn on Terrain Object Shadows (disabling either will fix) Closely packed objects with their own shading (tightly packed groups of trees, even with all shadows turned off) Maybe it can help trace down other performance issues in general. With object shadows enabled the game's render thread exhibits this kind of strange "frameskipping" at regular intervals: Fullscreen mode: So for testing, I added the script suggested by @BIGNEWY here Behold, Borderless Windowed after using alt+enter - The framedrops are gone! (So is my framerate, but its at least stable) Note that the spikes at the end are the screenshot event in both pictures. The stall happens exclusively in rendering. Both pictures are taken with sim paused, hence the CPU frametime of 0.15 ms until moment of screenshot. Note how these spikes also disappear entirely when EITHER Shadows or Terrain Object Shadows are set to OFF Flat Shadows still have the same impact when rendered in Fullscreen Exclusive Large groups of trees may also still cause identical framespike patterns These spikes also disappear if NO entities casting Terrain Object Shadows are present in FOV (e.g. If you look down at an Apache and no trees are in frame --> No dropped frames) These spikes however do happen if BOTH Shadows and Terrain Object Shadows are off, but tightly packed trees are nearby and inside FOV My graphics settings: My system: 5800X3D CPU, 32GB RAM, 6600 XT GPU

Yeah this seems to be the case, given a missile's altitude and speed the AI will have a very strong preference for either tactic.

In terms of Phoenix Pk the way ED's AI evaluates defensive manoeuvres against missiles is also an enormous culprit. A AIM-54 and AIM-120 fired at the same parameters will have the AI do a split-S defence against the AIM-54 90+% of the time, whereas vs the AIM-120 they will almost always do some attempt at cranking and notching. For long-range shots (35+ nmi) AIM-54 almost always arrives at 10 nmi with 0.5-0.8 excess mach compared to the AIM-120, but this doesn't matter in the slightest due to differences in AI manoeuvring.

AMRAAMs have the ASE cue for this, their thrust profile is entirely different and it allows off boresight shots to much greater extent. That said, for a Phoenix, a higher loft than we'd currently see could make sense from a deployment perspective against fighters. I notice manually lofted shots tend to gain/maintain more energy in the descent due to the steeper angle. A steeper descent angle also makes them slightly less vulnerable to a defensive tactic like a split-S, which currently bleeds a ton of energy due to the ED's terminal guidance profile being so aggressive at capturing lead in the vertical (also applies to AIM-120, but this incurs significantly less drag). Downside is that the seeker would be suffering more from clutter. It's hard to say which trade-offs would be favoured irl, but I definitely know which one I'd favour in the sim

Yeah but to my point, and what Karon mentions above as well, this is a particular tournament with a particular rule-set that will see a particular outcome. Mover provides commentary, but he makes it very clear in all his videos that DCS is a simulation/videogame and as such, there's game rules that apply for people to have fun and enjoy themselves. An example of these rules is that all the jets above have fuel for similar time in burner. Fun rule to make fuel management a core component of BFM execution (makes e.g. fuel dumping pre-merge a risk-reward proposition). Good rule to make sure matches don't drag on with someone running around in circles on afterburner. Completely disregards the endurance characteristics and fuel-weight-to-performance of each airframe. The same applies to the pylon discussion, and honestly the state of PvP simming at large tbh (people keep complaining there's no competitive redfor, yet keep slapping AIM-120Cs on anything blue that flies) Constraints make the game

Removing all pylons for the sake of BFM tournaments is something for you can do in the videogame for max performance funsies, but not a configuration these jets were intended to fly in. You will never have that configuration in a combat scenario. If you look at footage from real world BFM sets, you'll see not only pylons, but fuel tanks as well. As someone answered in another thread on this forum: the glove pylons were never designed to be removable (and as a result, never were) and are part of the aerodynamic profile, so they induce effectively zero drag (Similar to AIM-9s on F-16/F-18 wingtips having a drag index of 0) You would not see any positive performance changes even if they were removable. And this is exactly why nobody takes airshow performance as a remotely valid basis for judging combat performance.

Fair but 10% frametime on a 4090 might scale very differently on weaker systems. 2D scales very differently from VR, etc. OP's VR post with the same scene seems to indicate a 15% GPU and near 30% CPU frametime increase with the same hardware. Even as a worst case, that is immense. Similar RAM and VRAM changes as the posted video.

Respectfully, even in this video we're looking at a ~10% average frametime increase, which is pretty significant. The fact it doesn't impact his experience by virtue of having both the fastest GPU and CPU available on the consumer market today does not really discount that. Interesting is also the increase of VRAM usage by 1GB, while simultaneously using a little less system memory.

This helps immensely, not dropping any frames anymore. Thanks

Apologies, I missed your response. But that does mean the 29 nmi cut-off excessively impacts the F-14 in comparison to other jets, since it simply does not have an effective mid-range missile shot. So with that in mind I do understand people are panicking for PvP-type scenarios where jammers are commonplace, as it pretty much removes the F-14 from the BVR roster. Very unfortunate

Just for clarity but from these responses I seem to infer that JAT shots still perform a loft/MBAM on the flyout? Because if that is the case people are really worrying about nothing at all. I think a scary prospect of the 29 nmi hardcoded burn-through is the fact that 20-29 miles in practical terms means "20 miles or less", AIM-54 characteristics being what they are. Traditionally DCS missiles have used pure pursuit guidance on HOJ shots, which would very decisively run a phoenix out of juice in about 15 miles at the altitudes given in the image. If the missile still performs something of a loft outside of burn-through range then there is no reason for anyone to panic and assume we're back to WVR-style engagements. (We'll find out in a couple of hours anyway)

Thanks, a solution would be much appreciated

I think nobody on their testing team has additional resource-intensive processes like OBS running, or they simply do not understand the Windows resource allocation differences between Fullscreen Exclusive and Borderless Windowed. Either way this is a MASSIVELY negative change for those of us that do indeed stream the game for their communities.

Please reconsider this change because it completely broke support for CPU-based streaming. DCS is now effectively running in "Borderless Windowed" which means it does NOT prioritize Windows resources over other processes. As a result, Windows juggles resources between DCS and other processes, resulting in a seriously degraded experience for both the player and viewer.

Once it goes active or the track breaks there's no longer a reason to maintain TWS-A and you're better off resetting to save those precious cycles of AWG-9 processing power. Especially when you start getting to the shorter engagement ranges and TWS persists in tracking some off-heading phantom track. In the back seat I'd just delete this track or switch to P-SRCH depending on the situation, in the front seat the best way to maximize your SA in such a situation is to cycle to RWS.

Hey NineLine, you mentioned this, but the last actual patch note for the L-39 was in December 2021 where a take-off direction was flipped for a single mission. The one before that is December 2020 where they fixed a cockpit oxygen warning. The key issues with this module (reflections, weapon damage) would take no effort to fix, it just hasn't been given even a man-hour of resources in 2 years time. So any news on any fixes in the near future to show us it's not completely abandoned? December gonna be a good month this year for L-39 enjoyers?

I'm pretty sure this has been the case since release no? TWS has always been able to support only a single missile per track. If you want to ripple fire on a single target, use STT instead. Pretty tragic waste of a phoenix though, especially at 50 miles.

Ironically right now there is zero obligation to support the Phoenix. It is currently in its most autonomous state it has ever been hands-down, because the AWG-9 track breaking really kinda means bugger-all if target flies somewhat predictably (like... say... a sliceback or split-S) Hell, you can launch, turn off your radar and the missile will still probably find its way close enough to give the target a bit of a scare. For PvP, e.g. launch-and-leave grinder transitioning into close quarters kill-shots... I think the Tomcat won more than it lost, it's still >7 miles of powered flight on the deck. The AIM-54C is really the least of the F-14s worries in a modern 4th gen arena.

I'm aware. In that sense the missile is in an alright place. My point is more, people are focusing so intensely on a Mach 4+ speed that the missile pretty much provably would not achieve irl, while the real performance gains would exist in loft geometry. With a trade-off as you mentioned in the mid-range shot. The thing is that in the sim right now, the 20-30 nmi "bucket" simply isn't there, as the loft is so smooth/shallow. From the snippets I've read/heard, combined with what would make sense physics-wise, the bucket mostly existed because the missile would loft so aggressively, meaning that below about 30 nmi you are spending as much energy on the transition through the top of the arc as you gained from the initial loft itself. Prior iteration was better in this regard. (y'know, the one where the missile went to space if fired at poor parameters) In its current state the loft at 20-30 nmi does not detrimentally affect the shot in a significant way, and the mentioned range gap doesn't really exist. At 25 miles it's pretty much a slightly obese AMRAAM. In return, this means it performs sub-optimally when launching high-to-low. Right now, you use the Phoenix either high-to-high, or like an old-school AIM-120B. (Including the whole 2019-DCS-launch-and-immediately-turn-cold routine)

I don't know why people keep pulling the aerodynamic performance of the missile into question when it's so close to the NASA diagram. The one thing I do feel can still make massive improvements is the loft/guidance when shooting at targets that are not co-altitude. Example are look-down shots from 45k, 40-45 nmi to targets at 10,000 or below. The missile will barely use its energy to loft, rather choosing to only go up about 8-10,000 feet and then perform a shallow glide to target, all the way cruising through thick air burning off pretty much all its energy on sheer drag. Result is that even such a high-and-fast missile ends up with nowhere near the necessary energy to intercept its target. A higher loft would allow it incur much less drag during both the cruise and terminal phases and come down at a much sharper angle to benefit from gravity. It would be a much more sensible geometry for preserving energy for a heavy missile with a relatively weak, long-burning motor. I'm not saying these shots should actually hit, because evading such a shot at 10,000 is generally a matter of turning cold. The other aspect is that such a sharp descent makes it a bit more vulnerable to ground clutter. (Though on an SPO-15 the geometry makes that a very hard notch to read properly, not that the AI cares ofc...) There's a balance to be struck, but loft tweaks is definitely something I'd hope to see at least for the 54C

Yikes! Pretty sure the afterburner goes through 100 lbs faster than me turning that dial

Just wanted to check, is this intentionally modeled to need a full rotation for each 100 lbs? It makes tweaking it quite the hassle