I’ve been reaching around 30-60% GPU usage on my build with FiB ShadowMaps enabled.
I am playing with the graphics settings maxed, but with my rig should be able to handle it no problem…
(Currently using DirectX11.)
i7-8700k (Usually overclocked, I have another bench with an i7-8086k that has this same problem.)
32 GBs of DDR4-3000 MHz RAM
Windows 10 Preview (Insider, 1903 - 18890.1) (64-bit of course.)
RX 590 8GB V-RAM (Overclocked.)
I can run other games pretty well with this rig so its confusing when running both Roblox Client and Studio at around 20-60% GPU usage. (CPU is usually fine, Voxel is at about 20-30%)
(My actual FPS stay at 60, so I’m not really lagging.)
Edit: I had previously said on a baseplate, sorry.
Microprofile dump on a map with ShadowMaps: (While in-game on the client, I dumped 128 frames.) microprofile-20190509-150109.html (6.1 MB)
The profile you attached indicates that we are spending ~3ms per frame on rendering the map, which should correspond to ~20% GPU utilization (~20% of the frame we’re using the GPU for rendering, the rest is spent idling - your map could be 5x more expensive to render and GPU would still be able to render it at 60 FPS). I’m not sure why the reported utilization in task manager is 60%…
We have discussed this before many times, and again, for the current phase we aren’t planning to implement fine grained quality controls (it’s very possible that we’d have to separate draw distance from quality in phase 3 - we’ll see). There are many reasons for this, “most players wouldn’t use this feature” is not how we think about it and it’s a oversimplification. For your specific example, quality level 9 has ~9k stud draw distance with no SSAO and a minimum level of antialiasing, which seems like a reasonable workaround.
It’s different on the client and on the studio but might be due to using a preview build on Windows, they have a few bugs when it comes to task manager and the API. I’ll try running a VM on a stable build to see if it’s any different. (Both my computers are on preview builds so its pretty likely that’s the problem.)
Thanks though for putting in effort to help me.
( Also apparently object names aren’t filtered in dumps, e.g: an object with a new line in its name will break the html profiler dump and make it un-openable, any chance you could forward this as a bug report? )
Perhaps I read something else that said otherwise and I would like to apologize.
I still don’t agree with anti aliasing being apart of a quality level - the performance impact for most computers is simply too much, coupled with the simple fact that it overrides Radeon or NVidia settings - for me at least. Anti aliasing is just too taxing on my system for any noticeable graphical changes which is why I’ve been stuck with Quality 8, which isn’t terrible, but it still doesn’t allow me to see as far as I could if anti aliasing and SSAO wasn’t forced on Quality 9 & 10.
If anything, Draw Distance, SSAO, and AA should all be separate settings like they were before I believe 2013. There was a year where Roblox gutted all of the graphical settings under Studio and made everything into a single bar in Roblox’s client.
I just updated my game with future is bright phase 2 and now get a barrage of messages from players in my discord saying how they used to get a playable fps but now it’s just terrible for them(even at lower settings?) . I’m trying to ask them for specifics but there are a lot of them. These are all players that ran on poor machines to begin with but I do think there’s some performance issue somewhere
Also getting a few reports of my game taking forever to load now. I haven’t changed anything this update besides enable shadow map
Please ask them to share computer specs and microprofile data. The issue may be specific to your game as well - so leaving a link to the game would help.
Shadow map should not impact loading time. I don’t think this is related.
Try making some of the parts, like smaller ones, not cast shadows.
I did this for Redshift Arena, which has a ton of pickups, which are MeshParts and it boosted the performance.
The property is BasePart.CastShadow.
I seen that your game uses a lot of grass meshes, which could be the root cause of the performance problem for players with lower-end machines. I would definitely turn off shadows for these.
Just feel like it’s necessary for me to contribute here:
On my Mac Pro, I am still getting buttery smooth gameplay on max settings. When opening up the “Performance Stats”, I notice no difference from Voxel lighting. Mind you, I haven’t checked the micro profiler, but, on the outside, everything looks completely fine. Hopefully this can help you out in some way.
Edit: I played @tyridge77’s game for at least 10 minutes just to make sure I was done downloading things. Then, I dumped my microprofiler. Here is the file:
I casually took a picture of some clothing I just designed (I rarely ever do that lol, that’s why every time I do I test it a lot) and noticed weird zig-zag shadows on my garage’s walls, cast by my garage’s roof.
Image, put it in a dropdown to make it easier for others to scroll.
That’s because shadow maps aren’t ‘antialiased’ (I know that’s the wrong term but it’s similar to that) so you need to increase shadow softness if you don’t want to have that aliasing effect.
I had users test my airport, same thing, for some reason they are getting 30-20 fps, and it randomly increases, however before they maintained a solid 50-60. I have asked them to send me their microprofile data and computer specs, ill reply with the results once they reply.
I don’t know if this is meant to be this way but I just noticed that even if the server is on Voxel and the GlobalShadow property is disabled, there is an accurate shadow coming from the dragon:
