TESTED: Game mode and Hardware Accelerated GPU Scheduling (HAGS)

Interesting, I hadn’t thought about that. So in theory you could get higher framerates than when you flew it by hand to record it as some of the background simulation is not being calculated?

Just my guess - but not sure that’s what I actually see when I watch my replays back (I use FlightControlReplay) - I’ve not put the FPS counter on it - I’m more interested in smoothness. Given the sim still has to calculate everything else, I’m not sure if it is still doing the calcs (how would it know not to?) but effectively discarding the answers and replacing them with the fed data from the replay file.

Getting myself confused now. So maybe it would show a valid result!

1 Like

I’ve never used these flight recorders before, so I have to ask if you can change the weather conditions when you replay a recording? If so, and you could set up some strong winds for example, if the plan flies the recording just as it did before, then it isn’t calculating the winds effects on the plane at the very least. If it were I would expect the inputs made to the plane would no longer be valid, so unless it is emulating an autopilot it would deviate from the recorded route.

You can change anything other than the plane movement - so add/remove AI traffic for example, and change the weather. So you can record in zero wind and replay in 100 knots wind and the flight looks identical. I think this means that, at the very least, the calculations are being ignored - but I’m not seeing how the sim would know not to bother calculating the aircraft movement.

Simplistically (presumably - as I’ve no actual knowledge of this), the sim takes the current position of the plane, input controls, wind effects etc, does a momentary calculation and works out where the plane will be after a fraction of a second. It then uses that position plus all the other factors, as input to the next fraciton of a second and repeats from there. This is what causes the high CPU usage I think.

During a replay, I believe that some of the inputs - specifically those relating to the plane position/attitude etc are provided by the replay tool instead of the results of the previous calculation cycle. So I would think that the sim is still calculating where it thinks the plane will be after the next cycle, but then takes the actual values from the replay tool and ignores what it has just calculated. I don’t see how it could know not to do the calculation.

FCR does show things like gear up/down and shows switches moving and control surfaces moving - so it is presumably also providing control input values that affect the animation too.

Not sure where all this leaves us!

Realistically it might be best to have a defined route for the plane to fly on autopilot, starting from a defined waypoint, and measure from there. Remove weather from the equation, as well as any human interaction.

1 Like

Over lunch just now, I tried to test this with FlightRecorder (0.14). I setup a very short flight from Southend Pier to land at Southend Airport (EGMC), and allowed the copilot to have a crack it at. I recorded the result until wheels down (when the copilot had actually touched down off the right of the runway, and then veered violently off the left of the runway, ending up stopped on the grass perpendicular to the runway :rofl:), I then stopped the flight recording and played it back while capturing with CapFrameX.

At first I seemed to be getting quite consistent results, running the 5 minute flight over a few times without changing settings. I then started playing with HAGS and Game Mode, and it turned out that the best result was one of middle runs with both HAGS and GM on … so I went back and retested it, and got a fairly average result.

So, at least on this very short flight, the consistency isn’t really there. It’s worth noting by the way, from your discussion of whether or not the playback is easier on the sim or not … broadly speaking, it seems not. Again, this could just be general lack of consistency, but the first time I did the recording, immediately followed by playing it back to check FPS, the playback actually faired slightly worse than when the copilot was flying.

Later tonight if I have time, I’ll do my full 15 minute flight from parking to parking at 2 different airports (a route I often do), recording it as I go, without benchmarking. I’ll then do at least 2 runs of replaying that flight back (same options like multiplayer off, same starting time/weather, etc) on each mode to check it’s consistent. I’ll do this with the 4 combos of HAGS and Game Mode, and check that the 2 results of each test seem fairly consistent with each other. At the end, I’ll take the winner and re-run a 3rd time to triple check, and post my results here.

If anyone has any reccs for changes to that procedure, do let me know - as I don’t want to spend hours on this every night, I’d rather actually be playing :grin:

2 Likes

When I go into game mode and click on graphics settings, I don’t see any options for HAGS. Can someone please share a screenshot of this screen so I can see what I’m looking for?

it’s a Windows settings, not a game setting.

Settings → System → Display → Graphic Settings ( below ) → then you come to HAGS

2 Likes

Go to your start menu, and search for graphics:

Open that Graphics settings ‘app’, and it should look like this:

The headline option is Hardware-accelerated GPU scheduling … or ‘HAGS’. If you don’t see that, I imagine either your hardware doesn’t support it, or you’re using a version of Windows that doesn’t support it.

1 Like

Nope, it’s a driver question.
Some older Radeon drivers did enable hags, later on , AMD dropped support for it.
Must be the same for Nvidia…

1 Like

Wow, never would have thought to try HAGS off. It’s worked wonders on my system. At the FSDT Chicago ORD, I was getting 60+ FPS for the first time ever. Granted, it was in a Pitts, so not much going on CPU wise with that plane, but still. Couple of caveats - Photogrammatry off (because it’s ugly and really hurts performance), vegetation - Medium.

3 Likes

I’m aggregating my results now. I had some serious issues testing this, the main one last night being that without warning, something that hasn’t happened in probably 10 years for me, happened. I can’t even explain it now, but suddenly I discovered that I had completely run Scotch (whiskey) - none anywhere in the house. It made the process much more painful than it needed to be!

Then after almost 4 hours on this last night, I discovered there MAY have been a fault in my procedure. Namely that when using the replay tool, I may or may not have focussed the sim after clicking replay. I noticed that my animated wallpaper was still animating despite it being set to pause when a game is running. Seemingly it’s only when a game is running and focussed. When I realised on my final run of 8, I focussed the sim and FPS went up. I don’t know which, if any, of the runs I was focussing the sim or not. So I will likely discard those results in favour of the replacement runs I’ve been doing today :frowning:

1 Like

Right, here we go …

First off all the usual caveats:

  • This is my personal system, your mileage may vary.
  • This is a local flight for me, something I often fly in my local area. It is a suburban area, cities or mountains may have different results
  • This was done with the flight recorder and playback for consistency, but I don’t know what affect if any that has on CPU demand - although anecdotally, I would say it appears to be not that much, if anything it was lower FPS on playback than when I first recorded it, but these results are not presented here
  • This took me AGES, and I think I’m done, I can’t be bothered exploring this anymore

So this is a 15 minute flight in a C172s G1000 from EGSG to EGMC, it included startup, taxi, takeoff, cruise, descent, landing, taxi, shutdown. Flying time approx 14 minutes, end-to-end almost exactly 20 minutes.

I did this last night, and noticed toward the start of my FINAL run (which was HAGS off, Game Mode off) that I was not focussed on the sim, when I focussed the sim, the FPS jumped up a bit, and stayed higher throughout.

These are those results, I have used CapFrameX and aggregated both runs together:

Having dug into the data, I am somewhat confident that the only run that was affected by not focussing the sim (or phrased differently, the only run out of the 8 where I did focus the sim) was one of the 2 HAGS Off + GM Off, as a result, you can potentially disregard this entire dataset. But below, I present the amended set of data for the first set of 2 runs, but with that 1 run removed:


The *'d result was just the 1 run on it’s own, the other sets of bars are aggregated from 2 identical runs

Feel free to disregard these results as they are not perfect due to my fault, but the reason I present them at all, is because the amended results actually do tie in quite well with the results I got from today, where I was much more careful about procedure.

So I re-recorded a flight, double/triple checked every setting, ensured that I started the replay, then focussed the sim, then started the capture, which ran for precisely 1200 seconds (just enough time for the playback to finish at the other end). I also decided to try to stress the CPU more in the hopes that it would exaggerate any differences. For this, on top of background processes that I had running last night, I turned on my Flight Instrument Panels and ran them with SPAD.next, I also turned on the motion platform and software, along with NaviGraph charts, and even the G1000 bridge (for the G1000 iPad apps). This is pretty much what I normally have running when I’m playing the sim.

I am happy and confident with the follow set of data from doing all 4 tests twice, and aggregating the results:

You can see that in both the latest runs above, and the amended first set of runs, ‘HAGS on’ is in first and second place in both. It’s also worth pointing out while 1st and 2nd place are swapped between these 2 datasets, on both HAGS on + GM on actually had the highest 0.1% lows (and identical 1% lows). So in both runs, the ‘less stuttery’ experience is with both HAGS and Game Mode On.

Finally, here is the aggregate of all 4 runs (with the exception of HAGS off + GM off, which is the 3 consistent runs):

This seems to show that actually having both HAGS and Game Mode on, overall, is better for all of average, 1% low, and 0.1% low FPS. It also seems perfectly consistent across all these results that the worst option is Game Mode on, and HAGS off (very strange!!). Having both options off is also in 3rd place in all cases.

As I’ve said, these are my results, don’t necessarily expect the same if you were to try it yourself … but do absolutely feel free to do so, and let me know what you find!

For me personally, I’ll now be leaving both options On, and be happy that I’m probably getting the best overall experience when actually using the sim for the style of flying I normally do. As a bonus, it seems that opinion is with the latest nVidia drivers that HAGS on is essentially required for VR, which I do often use and so it’s mighty handy not to be worried about switching HAGS on/off and rebooting before playing. I had not yet tested VR with HAGS on/off, but I will say that I used to have it off, and noticed after the most recent driver update that it was more stuttery. I’ll try VR tonight or tomorrow, and with any luck, it will have smoothed out.

Congrats if you read all that, I hope you found it interesting!

12 Likes

Thanks for sharing the findings of your efforts.

Shouldn’t Asobo provide a system configuration guide that they recommend for the best results?!

3 Likes

HAGS is not a Asobo feature… it’s more or less a Beta Windows thing , which drivers and hardware must support too :wink:

I assume in future the new driver model will made this as default, but currently…

I know :wink:
My point is, Asobo should provide a system configuration guide with their recommendations for best performance.
Game mode, HAGS, timers, and so on.
Throwing a recommended processor and a GPU at us is not working. There are people here who have better than the ideal spec and still don’t see good performance. So, Asobo should tell us how to configure the system.

2 Likes

there are million different systems, they can’t do that :slight_smile:

Windows Default is: HAGS off , I assume there testers use it in this variant.

Whst do you mean a million systems?
I am not referring to hardware.
There is only one OS supported. Windows 10.
Tell us how to configure it.

Default Settings :slight_smile: … that I meant what I assume there testers use and for which recommendations about min/max is made. Special things or troubleshooting is mentioned then within FAQs.

Enabling HAGS is users own decision and can get beneffit or trouble.

edit: typo

1 Like

Well, something has definitely changed with the last patch. As I mentioned above, I was getting up to 41fps at gate 225 at EGLL.

I’m now getting around 23fps.

I was CPU limited before, and I’m not GPU limited for some reason.

2 Likes