Harlikwin

lol, will do I guess, but I don't have much in the way of hard data.

I mean my FT on an empty mission (not this one), are like 7-8ms... Or last time I checked they were, but 12900k, 3090, G2 100%. The aero definately is asking for more pixels tho.

The IR 530 has the same range as the EM version, same motor etc. So, IDK you might know or if our resident Mirage pilot might know. In terms of maneuvering the 530 seems to turn immediately off the rail and be able to pull very high G loads when doing so. Is there any evidence it could actually do this? For example most sidewinders and sparrows of the era could not. The sidewinders from the 60's/70's etc had a ~.5 sec delay before they could maneuver for a variety of reasons. I forget what the delay for the sparrow was but I know it had one.

What is your setup? your CPU frametime seems far more reasonable.

Yup I see exactly this with my 12900k and 3090. For VR, unfortunately the reality is that its all CPU limited at this point, not generally GPU limited. I mean I can crank up stuff for the GPU to do, which is mostly pointless, but until we see improvements on the multicore side for DCS I find the latest greatest cards dubious. Can you comment roughly how much of a boost you got with your prior card? A 3090 of some sort I assume?

The very basic physics are the same, and not hard to calculate or approximate in real time. And FYI pirate does both, it absolutely is used as an imaging FLIR. And yes classified stuff absolutely plays a big role into how it actually is able to generate those tracks with the accuracy it does, and no one is going to tell you how accurately its actually capable of doing it. As for "accuracy" not being modeled in DCS IMO its pretty criminal that it isn't since literally the last 40 years of sensor development have been all about how to reduce "uncertainity" in all of these sensor systems, and using sensor fusion techniques and track files and kallman filters etc. Do you need to model it all down to the last detail? No, you can't and it would be hard. But the fact that DCS has not even attempted to do it outside of the M2k radar is a real shame, because it is doable in broad strokes (as the M2k shows), and if it were done consistently right in the major modules, you'd actually see some very big functional gameplay differences between sensors on a mig21, the mirage F1, and the F18 for example. Which would lead to some serious gameplay changes for all of those modules. But as it stands in DCS a radar from the 1960's is just as good as one from the early 2000's, with literally no drawbacks modeled. And the fact from a radar standpoint that you can fly a mig21 like an F18 is well, not particularly realistic. Simply put, from a sensor standpoint, the defining characteristics of Gen2 fighters was that they had any sort of radar at all, and they were terrible by modern standards, Gen3 greatly improved the functionality of those radars allowing to do stuff like lookdown shootdown, and finally gen4 made those radars good, longer ranged, easy to used, and integrated them with other sensors, and really most us gen4/4.5 fighters have stuff like MSI, its just not modeled much/well in DCS.

You can't really do much with the system "output" without actually knowing at least in some detail how it actually does work. And respectfully no, an IRST seeker works nothing like a radar in the sense you don't get "range" from an IRST without doing some fancy things. So no, unless you know how to model how exactly the IRST is getting ranging info, and how "good" or "bad" that info is (and in many cases its going to be bad) you can't really model it well. I would not call what Razbam did with the HST "good" not even remotely. All it does is mark all "units" in the hud, and generate some "not-so random" noise. It doesn't operate in game at all like an HST, nor does it even remotely replicate what that display should look like. Yes there should be noise, but its very much not random. But they did the best they could with the state the game is in. I had high hopes for the FLIR remodel would let them model some of what those false signals should look like, but thats apparently not how it works. And like all IR systems currently it sees through clouds, that is pretty basic thing to get right about IRST's and IR seekers, and yet like what 2 years later since the clouds were introduced we still have this not so minor problem?

Absolutely correct. And those early systems are currently badly modeled in DCS (IR sees through clouds, no IR clutter etc). And pirate is a quantum leap (literally for the detector ) above those sorts of older sensors, and yes it can basically create and correlate track files as well as ID targets in some case like NCTR and has much better resolution than radar so it can do stuff like "RAID" (well it doesn't need to it is more the point). And honestly beyond, "broad" descriptions, no one is going to disclose much beyond the most general details of those capabilities.

Part of the issue during the last cycle was the fact crypo miners were still buying these cards and willing to pay a premium. As I understand it the entire crypto market has mostly collapsed, so the demand for this should be significantly lower. Hopefully the scalpers loose their asses on this.

I have heard theories that by upping some gfx settings in dcs you can unload the cpu since at some settings it might shift to the gpu. Idk if I believe it, but going from low to high shadows had little impact for me.

Pulse with MTI not doppler.

sir you make me sad, you could have sold it to someone here that would have appreciated it. Also I thought the physical dimensions were pretty close to a 3090?

MSI probably can't/wont be done well, but on the upside at least HB won't have to try to model a modern IRST like PIRATE for the German Trache1/2. The main issue is DCS itself, IRL all sensors have limitations and uncertainties, which for the most part aside from razbam and the M2k radar for example are absolutely not modeled at all in DCS. Like the F18 and 16 radars both should similar problems in similar situations, but ED doesn't model it. IRL that radar contact is basically a "sphere" with a plane somewhere in it, that sphere of uncertainty gets smaller as you get closer. No one models that, which is why RAID for example is worthless in DCS because all aircraft show up as discrete contacts even when they shouldn't. And this "sphere" is highly relevant for weapons performance. Imagine shooting your aamram in TWS against a sphere with like a 5 mile uncertainty. Well Mr AAMRAM can go, and hopefully when his radar is turned on he can find the enemy AC in that sphere still even though its not where the Viper radar told him its at, rather its like 3 miles to left and 3000ft higher. Which incidentally is why no one in the real world would want to use TWS for actual engagements, and used STT when possible, because the actual errors when using those modes were less due to much faster update times, i.e. Mr Radar has 5000 hits on a target to figure out exactly where it is instead of say 500 or 100. But again, uncertainty is not modeled in DCS, even though its highly relevant for modern modules. You have no idea what I have or havent seen, but I'll bet dollars to doughnuts its more than you. But the point being is that public data to check against is very useful for the community. And for modules much past the 90's you most folks aren't gonna see it, which makes checking accuracy of a module hard, meaning the overall quality of the model is lower. And its much easier and "safer" for everyone to work from public docs. That being said, even HB didn't model systems like the APX70 combat tree (VN era system) in the F14, because there likely aren't enough docs out there on it, or how it interacted with soviet era IFF systems. Even though literally most "in the know" people in the community know it has it. Does that effect how you employ the F14? You bet it does, and it leaves the module less capable than it was IRL. And if your expectation is they can do "better" with modern jets, well I guess you can think that but very few other people will. As for the EF pilot comment, do you think he is credibly giving up any actually classified info he can goto jail for? LOL... Nope. At best he's probably providing a level of disinformation making sure nothing too sensitive actually gets modeled too correctly since it has actual national security implications, cuz you know he took an oath at some point. So yeah, his commentary on the FM might be useful, or not. Since that matters alot less than say classified performance data on sensors or missiles etc. Esp since those are still in use.

Yeah its interesting but lord both the CPU and GPU frametimes are horrid. I assume its because of the high pixel count on the aero I guess. I wonder what these numbers would be like on a G2.

Not really, I just have an actual understanding of how alot of this works, and where DCS falls well short. I get it that they "try", but like on the whole MSI thing, that is literally what makes the hornet "good" IRL. But its entirely missing this system. "Modern" fighter combat isn't about dogfighting, and turning and burning etc. Its 100% about networked sensor systems and who sees the other guy first, gets a solution on them first and kills them BVR first. But literally no plane we have in DCS has any of that modeled from a sensor standpoint. Instead we have the 3rd generation part of 4th gen planes modeled but mostly not any of the 4th or 4th gen+ parts. What is sad is that most "modern" DCS players don't actually understand this.

Yeah thats the irony/issue right. Until ED actually manages to improve performance to "normal" game levels most of this is gonna be very small incremental improvements.

Who is marking all these things as a solution? None of them are.

Yeah, for me at least in MP, CPU is the issue. In SP with a light mission I can pull nearly 90fps with a 12900k/3090.

Clouds are synched in MP right now, just static. And yeah I forgot the AI sees through clouds too.

Will these finally block IR missiles/sensors from seeing through them?

Yeah and this is a very good point, its very hard to compare pancake to VR. IDK if the CPU issues DCS has cuz its literally having to draw the same frame twice, but I suspect thats part of the issue.

Yeah I mean it depends on your settings, I've optimized mine for online play and spotting. I'm sure I could do things to make it worse, but online its 100% a CPU problem for me not a GPU one. Offline yeah it would depend on the mission. At least in VR Id say you won't see much improvment. My 9700k and 2080ti to a 3090 maybe netted me 5fps in VR. Going from 9700k to 12900k (both at 5ghz) netted me like 25-30 fps. The performance difference between my 9700k and my 12900k in DCS was massive. And it makes sense if you look at the single core benchmarks. Its not just about GHZ, its things like Cache and IPC etc.

So, for reference, I went from a 9700k and a 2080ti to a 3090 and saw like maybe a few more frames like 5-10fps more. I then swapped the 9700k for a 12900k and saw a decent 20-30% frame increase. Going from a 1080 to a 3090 will alleviate any GPU bottlenecks you might have, but honestly you need at least a 12900k to get much out of a 3090. In VR rn online CPU is still the bottleneck, and I don't get the full performance from my 3090 it sits at 60-70% utilization.

Honestly IDK who the scalpers think they are going to be selling to. Previously the crypto miners I guess were willing to pay stupid money for cards since they were making money off them. But from what I read that market is basically dead at the moment. So I hope these scalpers loose their ass on the cards. I'm seeing 4090's north of 3k right now. And I doubt anyone is willing to pay that premium.

Well doing some optics math With 640x480 sensor and a 1.4x1.9 FOV and a 2x6m target, I'm getting a detect range of ~20km or so i.e. 2.6x4.3 pixels at 20km which is well within the "detect" Johnson Criteria. At 15km its 3.5x5 pixels which is a bit optimistic for "recognition for Johnson criteria. At 10km its 5.2x8.6 pixels which is on the ragged edge for "recognition", and at 8km its 6.5x10 pixel well within the "recognition" criteria. If I bump the sensor to a 1024x768 20km = 4.1x6.9 (detect) 15km = 5.5x9.2 (ragged edge of recognition) 10km = 8.3x13.8 (recognition) For reference this is a common "johnson critera" in terms of number of pixels for a man sized target. Detection 3.5x1 Recognition 13x5 Identification 28x8 So I guess depending on how optimistic you want to be about the sensor capabilities, and how much the manufacturer is over or understating those capabilites. Its fairly plausible the IR sensor is likely either 640x480 or possible better.