Biggus

I'm just pointing out that dropping texture quality isn't a fix, it's a mask. I've seen users with 4090s in the past reporting this issue. I'm not seeing any pagefile usage nor more than a few GB of DRAM usage changing upon triggering the issue. I know it's not conclusive at all though. I still suspect it's a faulty shader issue.

My VRAM usage does not change. It remains at 97.6% utilisation both above and below 900ft when set to medium terrain texture quality. One thing that does change is Direct 3D usage. Below 900ft, that ends up pegged to 100%. Above 900ft, it generally hovers between 50 and 70%. If I set aircraft textures to medium, GPU memory usage drops to 90%. However the frame rate still plummets upon descending through 900ft.

Quoting just this part for emphasis, because it's a perfect description. Texture resolution change on other maps might produce a gradual reduction in framerates as you descend generally, but not like this. This is like a switch. I've had so many moments where I've thought I've found a cause, or I've stumbled upon a cure for this issue. I've come back to this forum to report on my findings, only to have the problem reappear. There's a randomness to it. If you load a mission and you don't experience the problem the first time you are under 900ft, you won't get it during the play of that current mission. If you reload the mission, you might or you might not suffer it. If you load into a mission and you've got the sub 900ft frame drop, you might get it during the next mission, you might not. You probably will, though. The best possible culprit I've heard to explain it is that DCS wants to rebuild some splat textures endlessly under some circumstances. That would probably explain part of the randomness and the cause for what is often a quartering of the framerate.

@Czar66 don't be disheartened. It's a fairly nasty issue to diagnose. I've had plenty of moments where I've thought I've isolated a cause, only to have it reappear not long after.

I wouldn't consider the Phantom special radar option a fix for this issue. It's a workaround for a specific module in a specific scenario. I've tested nine modules (F1, F-15E, F-16C, F/A-18C, F-4E, A-10C2, AV-8B, F-14B, M2000C) in unpopulated scenarios at a variety of locations, and every single one of them exhibits performance issues passing through 900ft AGL on the South Atlantic map. There is some randomisation to it - sometimes a module won't suffer in one particular instance, but if you attempt the same mission with the same scenario later, the issue will resurface. Sometimes it improves noticeably after a patch, but then it returns in the next patch cycle even though nothing should have changed at all. VR, flat screen, high settings, low settings, it doesn't matter. The most noticeable individual settings change was to drop forest details factor as far to the left as it will go.

ATC changes are desperately necessary. But they must be part of a communication and control overhaul that includes GCI, AWACS, JTAC and any other class of unit that provides control for tactical aircraft. It would be disappointing to go through flawless startup and departure comms that are immersive and realistic, only to check in with Darkstar or Magic or Overlord and get what we have now. I'd argue that if anything, ATC should be the lowest priority of such an overhaul. It's still completely necessary but in a simulation with the main focus being on combat, effective combat control is several magnitudes of greater importance. But that's just my view and others are more than welcome to disagree.

Today's hotfix has cost me around 10fps on the F1EE take off instant action. I wonder if it's something obscure like shaders being rebuilt badly because of something random?

No and no. I do have priority set to high for DCS.exe in task manager, though. My system is a 5800X, 64gb DDR4 3800mhz, a 3070 and my DCS install is on a pair of Crucial MX500 SSDs (savegames on one, everything else on another). Power plan set to high performance.

No worries, wish it could have identified the actual problem for you. I'm not so sure it's a normal drop though, the "theoretical limit" on the fps graph was saying well over 150fps above 900ft. It's a tolerable drop now for me, but it's still an excessive reduction compared to other maps.

@Holbeach I tried your mission. I'm seeing a performance drop as I pass down through 900ft again, but it's a drop from my 65fps limiter to around 50fps. If I wasn't staring at the graph, I wouldn't have picked up on it. So I guess I'm still seeing degraded performance, but it's not nearly as bad as it used to be on my system. This is with a 5800X, 64gb 3800mhz and a very average 3070.

Fine here, mostly. It does seem to stop working if I eject for any subsequent spawn though.

@Holbeach man that's rough. I've spent a few more hours since those posts and haven't had the 900ft frame rate decimation happen at all. I'll play with my settings and see if I can trigger it.

I've found that after patching, deleting metashaders and fxo and then launching the game, that first session seems to be quite stuttery for me regardless of what I'm doing. I suspect that there's some weirdness happening where it's rebuilding the files, so my update routine usually includes loading a random instant action and then quitting, then starting the game again. You've probably tried this, but I thought I'd throw it out there. I did a little more testing today with both the F1 and the Phantom. Liberal usage of F10 menu, lots of transitioning through that old 900ft boundary. The only real performance slowdowns were in the Phantom, where I was holding 45fps at rooftop height at Mach 0.9 over Rio Grande. I suspect that's more a Phantom performance issue than a South Atlantic performance issue though. I will test the map with some larger missions using more assets over the next week or so, but my confidence with this map is growing.

I think I'm seeing a performance improvement in this patch. Normally with each update, I'll load in to the F1EE instant action mission at Ushuaia, take off and then fly back over the town at low level to evaluate performance using my normal graphics settings. As I've said in a few threads, I have experienced my frame rate being quartered as I pass below 900ft. Today, I did the same instant action mission. Lo and behold, I'm screaming over the rooftops at my frame rate limit of 65fps. I'm hoping my short test tonight is not an anomaly. Good work, devs.

Adding my voice to this thread, I'm extremely disappointed by all of this. It's a very important part of the map and I hope that it is returned in the next update.

Good hardware let down by poor software, a tale as old as time.

I'll investigate the F10 issue. The sub-900ft FPS drop has been an issue for me for a long time now. With my frame limiter being set reasonably low, the forest details fix might just be masking for me.

That's extremely frustrating.

I've had the same happen, and it's not consistent from module to module either. Have you tried turning forest details down to the very minimum on the slider?

Thank you, I appreciate the update.

I noticed that on the Kiowa, before the current patch it would enter data on the right hand screen and after the patch it's entering it on the left screen. Haven't really tried it with any other modules since the patch though.

Yeah, I've capped at 65 as I tend to find head tracker latency issues start to creep into my experience when I go lower than that.

I'm still suffering massive performance loss under 900ft. It seems to be alleviated by reducing forest details to the minimum setting, though.

I'll add a track if it happens again but I haven't seen it reported as yet. I was on a dedicated server with myself and one other player who was designating targets for me in an Apache. I was using the loft delivery mode, and through the first pass Jester gave constant speed call outs, not even a second between each line. It reminded me of some of the AAR Jester callout issues. The first delivery was fine aside from that, but after the second delivery, my frame rate rapidly declined to zero. No crash, but frames measured in minutes rather than seconds.

I suspect that Moza's engineers very well understand the way in which the AB9 is intended to be used, but their marketing department have no idea and therefore cannot support those who are reviewing early samples. That's just my impression though.