For anyone that runs maxed out dynamic reflections, what GPU do you use?

Discussion in 'Computer Hardware' started by Piastavgo, Feb 14, 2019.

  1. Piastavgo

    Piastavgo
    Expand Collapse

    Joined:
    Nov 24, 2017
    Messages:
    0
    I'm playing BeamNG with the 1050ti with everything in high settings 1080p, full shadows but no SSAO and no dynamic reflections. With this settings I get in the "Italy" map mostly
    60fps with just one car.

    My main gripe is that I can't "afford" dynamic reflections so I don't have functional mirrors inside the car.
    Anything in the dynamic reflections menu butchers my fps and even if I set dynamic reflections to minimum, the mirrors look like they came straight from a PS1 game, plus the framerate is still in the unplayable territory.

    To the point:

    I would like to know what GPU do you use to maintain a 60fps lock at 1080p in the Italy map inside Castelleto, with high settings, all shadows and maxed everything on the dynamic reflections submenu.
    Also I welcome anyone with a GPU that can't maintain a 60fps lock, but is in the 50+fps range with the above settings to chime in.

    For reference, on my 1050ti in "Italy" with high settings preset, maxed dynamic reflections, all shadows, no SSAO, no DOF, I get 20-35fps in the day and 16-31fps in the night with the high beam lights on.
    On the Castelleto spawn location, without moving and looking straight ahead with the orbit camera mode and using the Piccolina 110 (base model), I get 31fps at night with high beams on and 35fps in the day without light.

    Thank you!
     
  2. fufsgfen

    fufsgfen
    Expand Collapse

    Joined:
    Jan 10, 2017
    Messages:
    6,782
    I can't run dynamic reflections maxed out everywhere with every vehicle, especially not with SSAO on.

    Piccolina is bit easier than some cars, I'm getting barely 60fps with i7-8086K and gtx1080 8GB:
    upload_2019-2-14_23-3-11.png

    However would still need bit faster GPU and faster CPU clock speed to keep those maxed out in Norte (lot harder than Casteletto) and especially with K-series (Lot harder on CPU than Piccolina).

    There is no point to keep dynamic reflections maxed out though, distance is for example just about same to keep half of maximum.

    With 2080Ti and golden sample of 9900K overclocked to over 5Ghz, I think you might be able to maximize dynamic reflections all times, but just might.

    With replacing your GPU to something really fast, you might not get full performance out from it if your CPU is not one of the fastest, more cores is nice when running more cars, but it does not help to get better FPS with single car.

    So depending which CPU you have, you might need to upgrade that too.

    That said, running dynamic reflections with lower distance and not at full updates is giving nice mirrors too, you don't need to max them out necessarily.

    gtx1070 or 2060 could be something that gives good performance, but if you absolutely need to max out dynamic reflections then 2080 or 2080Ti is best bet coupled with overclocked 9900K
     
    • Like Like x 1
  3. Piastavgo

    Piastavgo
    Expand Collapse

    Joined:
    Nov 24, 2017
    Messages:
    0
    Thanks for the reply!
    I truly thought maxed dynamic reflections would be easy with anything above 1070 judging by the massive gap in performance in every benchmark I see. In my mind I had it like "if a gtx 1070 is 80+% faster than the 1050ti, then it would guaranty locked 60fps".
    For my CPU I changed recently from an i3 6100 to a ryzen 5 2600 knowingly that it is subpar for gaming but still frametimes in BeamNG look solid. I wouldn't justify anything above a 2060 class of card for my system.

    So I did some testing in order to better see the impact of dynamic reflections and to possibly find an optimized middle ground that looks descent and has the least impact on the GPU.

    Test method and settings

    The test was pretty much controlled. Using the orbit camera looking dead ahead at all times, I started from the Norte spawn point, took the third turn left (town square), then the second turn left, then the first left in order to reach the crossroad in front of the spawn point (church), stopped and did a slow 360 degree camera rotation around the vehicle, reset camera position and save results. Casual driving with an xbox controller.

    I used RivaTuner's benchmark tool in order to record the framerates.
    Each run was recorded separately.

    Graphics were set to 1080p high from the presets, without SSAO or DOF. My lowly 1050ti was used.

    Test vehicles

    I chose 2 vehicles for different reasons:
    • T-series T65 cargo box, default color (for a single, large vehicle with many surfaces)
    • Burnside Special V8 manual, default color (for a car that is bigger than most others and has many curved and chromed surfaces)
    Notes: chominess didn't appear to impact the performance so I dropped the chromed out T-series Tank Up-Fit from the tests.

    Baseline results without Dynamic reflections

    • T-series T65 cargo box:
    Average framerate : 56.8 FPS
    Minimum framerate : 47.1 FPS
    Maximum framerate : 63.1 FPS
    1% low framerate : 47.0 FPS
    0.1% low framerate : 46.3 FPS
    • Burnside Special V8:
    Average framerate : 59.3 FPS
    Minimum framerate : 50.0 FPS
    Maximum framerate : 65.0 FPS
    1% low framerate : 50.0 FPS
    0.1% low framerate : 48.5 FPS

    The hard numbers look way worse than irl here.

    Test 1 | Maxed out Dynamic reflections (1000m Distance)

    • T-series T65 cargo box:
    Average framerate : 32.1 FPS
    Minimum framerate : 28.5 FPS
    Maximum framerate : 33.9 FPS
    1% low framerate : 27.2 FPS
    0.1% low framerate : 25.0 FPS
    • Burnside Special V8:
    Average framerate : 32.8 FPS
    Minimum framerate : 29.6 FPS
    Maximum framerate : 34.7 FPS
    1% low framerate : 27.4 FPS
    0.1% low framerate : 24.6 FPS

    Jesus Christ almighty, 56.6% slower average framerate compared to the baseline without Dynamic reflections. RTX ON kinda situation!

    Test 2 | Maxed out Dynamic reflections (200m Distance)

    • T-series T65 cargo box:
    Average framerate : 33.9 FPS
    Minimum framerate : 29.9 FPS
    Maximum framerate : 36.2 FPS
    1% low framerate : 29.9 FPS
    0.1% low framerate : 25.5 FPS
    • Burnside Special V8:
    Average framerate : 35.0 FPS
    Minimum framerate : 31.0 FPS
    Maximum framerate : 37.0 FPS
    1% low framerate : 30.9 FPS
    0.1% low framerate : 27.4 FPS

    800 meters shorter Dynamic reflections but we gained only 6% better average framerate.

    Test 3 | Faces per update 3/5, detail 9/10, max texture size (100m Distance) | OPTIMIZED SETTINGS

    • T-series T65 cargo box:
    Average framerate : 39.8 FPS
    Minimum framerate : 34.5 FPS
    Maximum framerate : 42.9 FPS
    1% low framerate : 29.9 FPS
    0.1% low framerate : 29.9 FPS
    • Burnside Special V8:
    Average framerate : 41.1 FPS
    Minimum framerate : 35.9 FPS
    Maximum framerate : 44.1 FPS
    1% low framerate : 34.6 FPS
    0.1% low framerate : 34.3 FPS

    With this settings you get mirrors that are descent looking and pretty convincing I think, as well as non-jarring reflections off the vehicle's body.
    The GPU is happy so I am happy.
    With this optimized settings I see an average improvement of 22% over maxed out dynamic reflections (1000m distance) and 16.5% improvement over maxed out dynamic reflections (200m distance).

    My thoughts and observations

    Vehicle complexity definitely impacts performance as expected. Of course more vehicles will have further impact on the framerate.
    Dynamic reflection distance is not as expensive as faces per update.
    The test 3 optimized settings is the bare minimum that I would use if I could achieve anywhere close to 60fps.
    With the optimized settings on the 1050ti I see a 36% drop in framerate compared to not using dynamic reflections. Still better than the huge 56.6% with maxed out dynamic reflections.

    Screen space reflections would be a godsend in this game for body panel reflections..
    I have no clue what could be done to have better and less expensive mirrors tho.

    This took way longer than expected. Hats off to the guys and gals making YouTube videos and with 200 times more tests than I did here.

    Now if my assumptions are correct, a 1070/ vega 56/ 2060 (potentially the upcoming 1660ti) can run for the vast majority of times at 60fps with the optimized settings above (if not even locked). Is that the case?
     
  4. Deleted member 126452

    Deleted member 126452
    Expand Collapse
    Guest

    You're pretty much done here but I'm still chiming in even though my research is way less in-depth.
    With the settings cranked up on 1440p on an RTX2070 that is overclocked as well I usually don't se consistent framerates below 50fps and usually it's at a steady 60, well, I have seen pretty consistent ca. 45fps on Italy under some circumstances while driving around. Sometimes the framerate dips pretty randomly for a second but I guess that's BeamNG being weird every once in a while on certain, mostly modded, maps.

    And you're absolutely correct. You can overclock your Ryzen for more fps in BeamNG specifically, that's a giant bottleneck in this particular game as well.
     
  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice