Anti-Aliasing is usually a downgrade in graphical quality: washed out edges, blurring contours, foggy appearance. And on top it eats performance. You know, the thing EDO struggles with.
Anti-Aliasing in a Nutshell - Odyssey
- Thread starter Koben
- Start date
A few years back, one of the possibilities for reducing indentation o a 1080p monitor was exactly what you described above... but now we have way smaller pixels and a bunch of them (4K), so Aliasing is no longer that noticible.
Also we have now AI-driven Anti-Aliasing solutions like DLSS, which can provide even better image quality and details with a better performance than plain 4K rendering. Belive me, I use DLSS in Fortnite and the final result is much better than anything else I have tried before!
Properly(!) implemented AA is always an improvement and was there long before "1080p". I've seen better AA in games from almost two decades ago than in Odyssey.
Also, DLSS is slightly more than an AA solution.
Try 1.78x and more sub-sampling (supersampling multiplier below 1.00) with the in-game settings.
Good AA implementations clean up jaggies without touching the rest of the image much and generally aren't that costly in terms of performance.
A big reason EDO looks like relative crap at times is the blatant aliasing, especially of high-contrast lines, and a big reason I still have performance complaints is that the only good way to mitigate this on most of my hardware is via super sampling, which does eat a lot of performance.
Older games tend to have MSAA, which is pretty good at mitigating jaggies but cost less than SSAA/FSAA, and modern games tend to have a solid SMAA or T(X)AA implementation that are generally better at preserving sharpness than MSAA, with similar edge anti-aliasing, at even lower cost.
Look at this image:
That's from the suit selection page on a system running 4k ultra-ish, with no in-game supersampling, no FSR, no sharpening, and the in-game SMAA.
It's clear that SMAA is working much more strongly on some edges than on others. In particular, the AA looks pretty good on the character's non-directly illuminated right side. That's the sort of AA I expect, everywhere, from a decent SMAA implementation.
Now look at the illuminated side of the character, the gun, and any edge that seems to have specular highlights. It's barely anti-aliased at all and looks like crap.
This is the overriding issue with this game's AA. It also has some curious problem with single pixel thick lines on newer AMD hardware.
I don't know how directly related this is to the issue the OP is experiencing with certain texture edges, but I would not be surprised if it were all interconnected. SMAA and FXAA are shader based post-process filters and EDO has had all sorts of strange issues with it's shaders. Maybe they are doing the anti-aliasing in the wrong order, or taking a sub-optimal path with some hardware...I can only guess at what the underlying cause is. All I know for sure is that it looks like crap and it really shouldn't have to.
That's from the suit selection page on a system running 4k ultra-ish, with no in-game supersampling, no FSR, no sharpening, and the in-game SMAA.
It's clear that SMAA is working much more strongly on some edges than on others. In particular, the AA looks pretty good on the character's non-directly illuminated right side. That's the sort of AA I expect, everywhere, from a decent SMAA implementation.
Now look at the illuminated side of the character, the gun, and any edge that seems to have specular highlights. It's barely anti-aliased at all and looks like crap.
This is the overriding issue with this game's AA. It also has some curious problem with single pixel thick lines on newer AMD hardware.
I don't know how directly related this is to the issue the OP is experiencing with certain texture edges, but I would not be surprised if it were all interconnected. SMAA and FXAA are shader based post-process filters and EDO has had all sorts of strange issues with it's shaders. Maybe they are doing the anti-aliasing in the wrong order, or taking a sub-optimal path with some hardware...I can only guess at what the underlying cause is. All I know for sure is that it looks like crap and it really shouldn't have to.
This is what I've been trying to show... maybe what is being reported as "bad AA" might be something else giving the impression of it.
My only hope is that anyone at Frontier would take a look at this and get a full diagnosis on what is going on besides just AA issues.
Just a follow up: I've now completed my regular slew of benchmarking with this and can say I'm thoroughly impressed. Zero impact to framerates in ground conflict zones using this particular DSR + subsampling method (further evidence that ground CZs are still horribly CPU bottlenecked), and only minor framerate impacts elsewhere.Just wanna say I gave this a try and I'm loving the results so far. Significant decrease in the jaggies without too much performance impact.
I've yet to test it in a conflict zone, but this is showing far more promise than all the experimenting I did with Reshade last week. The only decent result I could get out of that program absolutely murdered the HUD, and blurred everything else. Your method straight up improves everything in the image, while being even kinder on the framerate than Reshade was.
My hat's off to you friend. Best AA "fix" I've found so far. I might even have to edit my signature.
While there's only a small (but still noticeable) improvement in visual quality when comparing screenshots, the real difference here is the reduction in flickering while in motion, which is the most annoying part of Odyssey's aliasing issues. Doesn't remove them completely, but it's a massive improvement.
I can't thank you enough for sharing this, Morbad. The game feels so much better to play now on the ground.
Was in a CZ on foot just a few hours ago and got about 47FPS... stood still and did some testing:
CPU (7820X 4.5GHz 16Threads) load was at 25% - quite a bit of headroom there;
GPU (3090Strix) load at 70-75% (undervolting and underclocked to 1620MHzCore - 9500MHz VRAM);
GPU at 1800MHz Core - 11000MHz VRAM and GPU load decreased to 66% - CPU still at 25% flat.
FPS was still at 47FPS!! This seems not at all related to hardware... or how powerfull/weak it is.
CPU (7820X 4.5GHz 16Threads) load was at 25% - quite a bit of headroom there;
GPU (3090Strix) load at 70-75% (undervolting and underclocked to 1620MHzCore - 9500MHz VRAM);
GPU at 1800MHz Core - 11000MHz VRAM and GPU load decreased to 66% - CPU still at 25% flat.
FPS was still at 47FPS!! This seems not at all related to hardware... or how powerfull/weak it is.
That doesnt tell much. Except maybe there are things really badly optimized for multithreading.
I get 60 FPS in 4K on my 6900XT in on-foot CZ. But do have a 5900X, maybe that helps?
Missing the resolution they are running in, if it is 8K or higher...
Maybe your other settings allows you to have a better framrate...
But the point I was trying to make is, overclocking my GPU just made it more idle... and CPU doesn't show signs to be the bottleneck.
Because Odyssey doesn't use all threads a CPU can offer.
Probably, they are all mostly Ultra.
ETA: @Morbad made a few posts regarding CPU bottlenecking - perhaps they might drop in and add a few more words of wisdom here, apparently windows Task Manager is way too slow to show the issue.
Last edited:
Seems that way... 8 Threads might be what Cobra engine can handle at once.
If you're not limiting the frame rate and the GPU is not being fully utilized then either there are SW issues such as bad threading or similar, or the CPU is limiting the frames. ODY does appear to be much more CPU intensive then Horizons, and not just at a ground CZ either, although in that scenario it would be logical to see higher CPU load.
Which doesn't seem to be sufficient for the amount of work Odyssey demands.
It won't help with CPU-Bound issues.
I know... it was just an attempt to bring the "8-thread" back to the original subject... "Anti-Aliasing"