• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Why does Nvidia brag about FXAA? It makes most of my games look like crap...

newtekie1

Semi-Retired Folder
Joined
Nov 22, 2005
Messages
28,472 (4.25/day)
Location
Indiana, USA
Processor Intel Core i7 10850K@5.2GHz
Motherboard AsRock Z470 Taichi
Cooling Corsair H115i Pro w/ Noctua NF-A14 Fans
Memory 32GB DDR4-3600
Video Card(s) RTX 2070 Super
Storage 500GB SX8200 Pro + 8TB with 1TB SSD Cache
Display(s) Acer Nitro VG280K 4K 28"
Case Fractal Design Define S
Audio Device(s) Onboard is good enough for me
Power Supply eVGA SuperNOVA 1000w G3
Software Windows 10 Pro x64
it doesn't seem like FXAA can tell the difference between the edge of an object and a sharp line on a texture so it makes everything look blurred...

It doesn't matter if the line is the edge of an object or a line in a texture, it should be softened by AA, that is the point of AA. It is supposed to remove jaggy lines, and the only way to do that is to soften them, or blur if you prefer.
 
Joined
Feb 26, 2008
Messages
4,876 (0.83/day)
Location
Joplin, Mo
System Name Ultrabeast GX2
Processor Intel Core 2 Duo E8500 @ 4.0GHZ 24/7
Motherboard Gigabit P35-DS3L
Cooling Rosewill RX24, Dual Slot Vid, Fan control
Memory 2x1gb 1066mhz@850MHZ DDR2
Video Card(s) 9800GX2 @ 690/1040
Storage 750/250/250/200 all WD 7200
Display(s) 24" DCLCD 2ms 1200p
Case Apevia
Audio Device(s) 7.1 Digital on-board, 5.1 digital hooked up
Power Supply 700W RAIDMAXXX SLI
Software winXP Pro
Benchmark Scores 17749 3DM06
so...
¿whatcha doin with that 7970?
 

STCNE

New Member
Joined
Sep 1, 2010
Messages
236 (0.05/day)
Location
Connecticut USA
System Name Lycanthropy
Processor i5 2500K
Motherboard ECS p67-h2a2
Memory 8GB Muskin Radioactive Series
Video Card(s) MSI R7770
Case Lancool K62
Audio Device(s) Ultra-Fi Dac41, Vintage Sherwood amp, STAX SR-5 Headphones
Power Supply XFX 850W
Software Win 7 64-Bit(Tried Win8 and never going back)
Benchmark Scores Reboots in 35 seconds flat, with a standard HDD. I'm a bit of an OCD tweaker. Using less than 100GB
Agreed, I don't really like the new 'low performance hit' AA that AMD and NVIDIA have now. It just blurs things too much. I used MSAA back when I played GTAIV because it looked really bad without AA of any kind(the update that added AA didn't work for me, needed a mod to bypass broken GFWL, only worked on 1.4)

What I really don't like is how NVIDIA advertised SSAA when it didn't work on any game I tried it with when I had a NVIDIA card. It's one of the few things that drove me back to AMD where I could force it on for any non-dx11 game I had. Kind of ironic really, shortly after I switched back I stopped being so picky about the graphics. When I realized I would have needed an SSD to play New Vegas with the massive texture pack I downloaded I realized how much I was putting into the GPU so I could go completly overboard with the settings and downgraded to put my money to use elsewhere.
 

Binge

Overclocking Surrealism
Joined
Sep 15, 2008
Messages
6,979 (1.23/day)
Location
PA, USA
System Name Molly
Processor i5 3570K
Motherboard Z77 ASRock
Cooling CooliT Eco
Memory 2x4GB Mushkin Redline Ridgebacks
Video Card(s) Gigabyte GTX 680
Case Coolermaster CM690 II Advanced
Power Supply Corsair HX-1000

Benetanegia

New Member
Joined
Sep 11, 2009
Messages
2,680 (0.50/day)
Location
Reaching your left retina.
It doesn't matter if the line is the edge of an object or a line in a texture, it should be softened by AA, that is the point of AA. It is supposed to remove jaggy lines, and the only way to do that is to soften them, or blur if you prefer.

Of course it matters. Lines inside of textures don't have jaggies, providing that AF is used. That's what anisotropic filtering is for, well texture filtering in general.

Any post-process AA blurs the entire screen and although the algorithms have become better and better, they are still nowhere near as accurate as MSAA because they don't perform depth tests and so they cannot tell the difference between an edge or a line in a texture.

That's not to say that FXAA is useless. For the grand mayority of people and in most games it's better than 0xAA, so for slower cards is perfect.
 

Phusius

New Member
Joined
Mar 23, 2012
Messages
1,316 (0.30/day)
Processor i5-2500k @ 4.5
Motherboard Asus Z68 Deluxe
Cooling Noctua NH-D14
Memory 16GB DDR3 1600
Video Card(s) Sapphire 7950 @ 1200/1475 @ 1175v
Display(s) Dell 23" 2ms
Case Carbide 500r
Audio Device(s) Asus Xonar DG
If you designed IP wouldn't you talk about it if your goal was to sell product?

Yes, but now I have no shame in being an AMD man anymore, at least they don't feed me lies. Just give me pure power and say enjoy.
 

Benetanegia

New Member
Joined
Sep 11, 2009
Messages
2,680 (0.50/day)
Location
Reaching your left retina.
Yes, but now I have no shame in being an AMD man anymore, at least they don't feed me lies. Just give me pure power and say enjoy.

AMD did the exact same thing with MLAA, exact same claims. It looks worse, it runs slower. So are you gonna be an Intel guy now?

http://sites.amd.com/us/game/technology/Pages/morphological-aa.aspx

Small performance impact

Truth: greater impact than FXAA.

Stunning image quality

... you thread, your call.

EDIT: And woah woah woah, wait a minute:

The stunning image quality enabled by MLAA is broadly comparable to that of Supersample Anti-Aliasing (SSAA)

That's when marketing guys cross the line...
 

OneMoar

There is Always Moar
Joined
Apr 9, 2010
Messages
8,741 (1.71/day)
Location
Rochester area
System Name RPC MK2.5
Processor Ryzen 5800x
Motherboard Gigabyte Aorus Pro V2
Cooling Enermax ETX-T50RGB
Memory CL16 BL2K16G36C16U4RL 3600 1:1 micron e-die
Video Card(s) GIGABYTE RTX 3070 Ti GAMING OC
Storage ADATA SX8200PRO NVME 512GB, Intel 545s 500GBSSD, ADATA SU800 SSD, 3TB Spinner
Display(s) LG Ultra Gear 32 1440p 165hz Dell 1440p 75hz
Case Phanteks P300 /w 300A front panel conversion
Audio Device(s) onboard
Power Supply SeaSonic Focus+ Platinum 750W
Mouse Kone burst Pro
Keyboard EVGA Z15
Software Windows 11 +startisallback
because obviously forcing a post-process effect onto games that where never written with it in mind is gonna end well
 
Joined
May 22, 2010
Messages
2,516 (0.50/day)
Location
Canada
System Name m1dg3t | DeathBox | HairPi 3
Processor 3570k @ 4.0 1.15v BIOS | q9550 @ 3.77 1.325v BIOS
Motherboard Asrock z77e iTX | p5q Dlx 2301 BIOS
Cooling Custom Water | D-14 & HR-03gt | Passive HSF
Memory Samsung MV-3V4G3D 4g x 2 @ 1866 1.35v | OcZ RpR 2g x 4 @ 1067 2.2v
Video Card(s) MSi 7950 tf3 @1000 / 1350 | Asus 5870 V2 @ 900 / 1275
Storage Adata sx900 256Gb / WD 2500 HHTZ | WD 1001 FALS x 2
Display(s) BenQ gw2750hm | 46" Sharp Quatron
Case BitFenix Prodigy - m0dd3d | Antec Fusion Remote MAX
Audio Device(s) Onboard Toslink > Yamaha HTR 6290 | Xonar HDAV1.3 > Yamaha DSP z7
Power Supply Ocz mXp700w | Ocz zx850w | Cannakit 5v 2.5a
Mouse Logitech G700s | Logitech G9x - Cable Repaired
Keyboard TT Meka G1 - Black w Cherry Blacks| Logitech G11
Software Win7 Home | Xp sp3 & Vista ultimate | Raspbian
Benchmark Scores Epeen!! Who needs epeen??
Simple answer; To sell cards.

/Thread

:rolleyes:
 

Benetanegia

New Member
Joined
Sep 11, 2009
Messages
2,680 (0.50/day)
Location
Reaching your left retina.
because obviously forcing a post-process effect onto games that where never written with it in mind is gonna end well

Nah it's not a problem of being a post-process per se. Because it's post process there's no problem applying later, quality wise, performance is better if applied within the app.

The problem is the nature of the beast. Post-process AA does not have extra information. For a 1920x1080 frame, you get only that many pixels of meaningful information. Then that info is dilluted when for every pixel, nearby pixels are mixed with the original. So it's blurring, plain and simple.

With 4xMSAA you get 4 color and 4 alpha coverage samples per every pixel and it's this info that is mixed for the final picture. Nearby final pixels are not mixed with each other, blurring does not occur.

Of course 4xSSAA is much better, since you get 4 full pixels per output pixel. (with shaders, lighting, shadowing, everything already calculated)
 

OneMoar

There is Always Moar
Joined
Apr 9, 2010
Messages
8,741 (1.71/day)
Location
Rochester area
System Name RPC MK2.5
Processor Ryzen 5800x
Motherboard Gigabyte Aorus Pro V2
Cooling Enermax ETX-T50RGB
Memory CL16 BL2K16G36C16U4RL 3600 1:1 micron e-die
Video Card(s) GIGABYTE RTX 3070 Ti GAMING OC
Storage ADATA SX8200PRO NVME 512GB, Intel 545s 500GBSSD, ADATA SU800 SSD, 3TB Spinner
Display(s) LG Ultra Gear 32 1440p 165hz Dell 1440p 75hz
Case Phanteks P300 /w 300A front panel conversion
Audio Device(s) onboard
Power Supply SeaSonic Focus+ Platinum 750W
Mouse Kone burst Pro
Keyboard EVGA Z15
Software Windows 11 +startisallback
Nah it's not a problem of being a post-process per se. Because it's post process there's no problem applying later, quality wise, performance is better if applied within the app.

The problem is the nature of the beast. Post-process AA does not have extra information. For a 1920x1080 frame, you get only that many pixels of meaningful information. Then that info is dilluted when for every pixel, nearby pixels are mixed with the original. So it's blurring, plain and simple.

With 4xMSAA you get 4 color and 4 alpha coverage samples per every pixel and it's this info that is mixed for the final picture. Nearby final pixels are not mixed with each other, blurring does not occur.

Of course 4xSSAA is much better, since you get 4 full pixels per output pixel. (with shaders, lighting, shadowing, everything already calculated)

exactly there is nothing inherently wrong with fxaa its self fxaa/smaa work best when you provide it with all the information about the scene when you just slap something on down the pipe it endsbadly
 

Benetanegia

New Member
Joined
Sep 11, 2009
Messages
2,680 (0.50/day)
Location
Reaching your left retina.
exactly there is nothing inherently wrong with fxaa its self fxaa/smaa work best when you provide it with all the information about the scene when you just slap something on down the pipe it endsbadly

But it doesn't get any more info whether it's built into the game or applied later. The algorithm is run AFTER the final composition is made. When the aliased frame is ready to be sent to the monitor, it's then when FXAA or MLAA take that picture and basically blur it "intelligently*". There's no extra information passed by the games engine, no further calculations other than what the algorithm does on those 1920x1080 pixels. That's why it's so fast.

* The algorithm suposedly takes care of trying to guess what needs to be blurred most (i.e polygon edges) and what info needs to be kept (i.e a line or dots in a texture).
 

Phusius

New Member
Joined
Mar 23, 2012
Messages
1,316 (0.30/day)
Processor i5-2500k @ 4.5
Motherboard Asus Z68 Deluxe
Cooling Noctua NH-D14
Memory 16GB DDR3 1600
Video Card(s) Sapphire 7950 @ 1200/1475 @ 1175v
Display(s) Dell 23" 2ms
Case Carbide 500r
Audio Device(s) Asus Xonar DG
Fair enough bene, MLAA, never heard of it before now. Heh, well I am happy with my 660 non-ti version, it runs everything very well so far and stays ice cold and silent. so it all depends if I end up fixing the vrms on my 7970 whether or not I keep it. may just throw away my arctic 7970 cooler and use my reference Msi cooler that my 7970 came with originally, i hear the ref should cool vrms much better
 

OneMoar

There is Always Moar
Joined
Apr 9, 2010
Messages
8,741 (1.71/day)
Location
Rochester area
System Name RPC MK2.5
Processor Ryzen 5800x
Motherboard Gigabyte Aorus Pro V2
Cooling Enermax ETX-T50RGB
Memory CL16 BL2K16G36C16U4RL 3600 1:1 micron e-die
Video Card(s) GIGABYTE RTX 3070 Ti GAMING OC
Storage ADATA SX8200PRO NVME 512GB, Intel 545s 500GBSSD, ADATA SU800 SSD, 3TB Spinner
Display(s) LG Ultra Gear 32 1440p 165hz Dell 1440p 75hz
Case Phanteks P300 /w 300A front panel conversion
Audio Device(s) onboard
Power Supply SeaSonic Focus+ Platinum 750W
Mouse Kone burst Pro
Keyboard EVGA Z15
Software Windows 11 +startisallback
method != implementation it is very possible to apply fxaa to certain elements of the scene and not to the entier frame (bf3 does this sorta)
 

Benetanegia

New Member
Joined
Sep 11, 2009
Messages
2,680 (0.50/day)
Location
Reaching your left retina.
method != implementation it is very possible to apply fxaa to certain elements of the scene and not to the entier frame (bf3 does this sorta)

Then it's not really fxaa. Honestly I don't know how it works and I have no interest in the game, so I can't even see it with my eyes. Google doesn't help either.

I've seen that it's used alongside MSAA or something like that, so it's posible that it uses some kind of Edge Detect AA (kind of like MSAA but only coverage, without 4x color info), bake that alpha coverage info into a texture and then use that texture to let the algorithm know where to enhance the blurring on edges, and less on textures. Fair enough, but since alpha tests are performed anyway, MSAA could be used in the first place (or CSAA, because that's basically what CSAA is). Seems like a moot point. idk.
 

OneMoar

There is Always Moar
Joined
Apr 9, 2010
Messages
8,741 (1.71/day)
Location
Rochester area
System Name RPC MK2.5
Processor Ryzen 5800x
Motherboard Gigabyte Aorus Pro V2
Cooling Enermax ETX-T50RGB
Memory CL16 BL2K16G36C16U4RL 3600 1:1 micron e-die
Video Card(s) GIGABYTE RTX 3070 Ti GAMING OC
Storage ADATA SX8200PRO NVME 512GB, Intel 545s 500GBSSD, ADATA SU800 SSD, 3TB Spinner
Display(s) LG Ultra Gear 32 1440p 165hz Dell 1440p 75hz
Case Phanteks P300 /w 300A front panel conversion
Audio Device(s) onboard
Power Supply SeaSonic Focus+ Platinum 750W
Mouse Kone burst Pro
Keyboard EVGA Z15
Software Windows 11 +startisallback
I guss that point is that if I got out and buy a brandnew sports car and they claim it goes 0-60 in under 4 seconds and every time I have a chance I have to get up to 60 I gun the engine and then the engine blows up in ~30k that's not exactly the car dealers problem really bad analogy I know but its late and ./care
 
Joined
Jun 12, 2007
Messages
4,815 (0.78/day)
Location
Wangas, New Zealand
System Name Darth Obsidious
Processor Intel i5 2500K
Motherboard ASUS P8Z68-V/Gen3
Cooling Cooler Master Hyper 212+ in Push Pull
Memory 2X4GB Corsair Vengeance DDR3 1600
Video Card(s) ASUS R9 270x TOP
Storage 128GB Samsung 830 SSD, 1TB WD Black, 2TB WD Green
Display(s) LG IPS234V-PN
Case Corsair Obsidian 650D
Audio Device(s) Infrasonic Quartet
Power Supply Corsair HX650w
Software Windows 7 64bit and Windows XP Home
Benchmark Scores 2cm mark on bench with a razor blade.
As said in a previous post, certain types of AA can make a game look worse or cause graphical glitches.

Only way to find out is try the different types of AA if you can be bothered and see which types appeal to you in certain games and which do not.

Some people like types of AA which make the game look like it is played through VGA to remove the jaggies.
Some people see the jaggies as making the image look crisp.
 

Benetanegia

New Member
Joined
Sep 11, 2009
Messages
2,680 (0.50/day)
Location
Reaching your left retina.
I guss that point is that if I got out and buy a brandnew sports car and they claim it goes 0-60 in under 4 seconds and every time I have a chance I have to get up to 60 I gun the engine and then the engine blows up in ~30k that's not exactly the car dealers problem really bad analogy I know but its late and ./care

I'm not very good with analogies either, but for what we are discussing: post-process AA is like painting the car. The end result is mostly the same whether you paint the body before the car is assembled or afterwards. Painting it afterwards means you have to cover the parts you don't want painted and so it will take more time (analogy to the greater performance hit when applied outside the app).
 

johnspack

Here For Good!
Joined
Oct 6, 2007
Messages
5,980 (0.99/day)
Location
Nelson B.C. Canada
System Name System2 Blacknet , System1 Blacknet2
Processor System2 Threadripper 1920x, System1 2699 v3
Motherboard System2 Asrock Fatality x399 Professional Gaming, System1 Asus X99-A
Cooling System2 Noctua NH-U14 TR4-SP3 Dual 140mm fans, System1 AIO
Memory System2 64GBS DDR4 3000, System1 32gbs DDR4 2400
Video Card(s) System2 GTX 980Ti System1 GTX 970
Storage System2 4x SSDs + NVme= 2.250TB 2xStorage Drives=8TB System1 3x SSDs=2TB
Display(s) 2x 24" 1080 displays
Case System2 Some Nzxt case with soundproofing...
Audio Device(s) Asus Xonar U7 MKII
Power Supply System2 EVGA 750 Watt, System1 XFX XTR 750 Watt
Mouse Logitech G900 Chaos Spectrum
Keyboard Ducky
Software Manjaro, Windows 10, Kubuntu 23.10
Benchmark Scores It's linux baby!
fxaa actually worked great on my sli 285s, but 32x csaa on some games, upto 8x supersampling in most, works better on my 480. I stopped using fxaa....
 

johnspack

Here For Good!
Joined
Oct 6, 2007
Messages
5,980 (0.99/day)
Location
Nelson B.C. Canada
System Name System2 Blacknet , System1 Blacknet2
Processor System2 Threadripper 1920x, System1 2699 v3
Motherboard System2 Asrock Fatality x399 Professional Gaming, System1 Asus X99-A
Cooling System2 Noctua NH-U14 TR4-SP3 Dual 140mm fans, System1 AIO
Memory System2 64GBS DDR4 3000, System1 32gbs DDR4 2400
Video Card(s) System2 GTX 980Ti System1 GTX 970
Storage System2 4x SSDs + NVme= 2.250TB 2xStorage Drives=8TB System1 3x SSDs=2TB
Display(s) 2x 24" 1080 displays
Case System2 Some Nzxt case with soundproofing...
Audio Device(s) Asus Xonar U7 MKII
Power Supply System2 EVGA 750 Watt, System1 XFX XTR 750 Watt
Mouse Logitech G900 Chaos Spectrum
Keyboard Ducky
Software Manjaro, Windows 10, Kubuntu 23.10
Benchmark Scores It's linux baby!
FXAA worked great with my sli 285s. Nice bandaid for weak cards. With my 480 I use 32x CSAA, SuperSampling, and enforced real AA. I'm looking at sli 480s soon, which is the min to go past 660ti in performance....
 
Joined
Apr 10, 2012
Messages
1,400 (0.32/day)
Location
78°55' N, 11°56' E
System Name -aLiEn beaTs-
Processor Intel i7 11700kf @ 5.055Ghz
Motherboard MSI Z490 Unify
Cooling Corsair H115i Pro RGB
Memory G.skill Royal Silver 4400 cl17 @ 4403mhz
Video Card(s) Zotac GTX 980TI AMP!Omega Factory OC 1418MHz
Storage Intel SSD 330, Crucial SSD MX300 & MX500
Display(s) Samsung C24FG73 144HZ
Case CoolerMaster HAF 932 USB3.0
Audio Device(s) X-Fi Titanium HD @ 2.1 Bose acoustimass 5
Power Supply CoolerMaster 850W v2 gold atx 2.52
Mouse Razer viper 8k
Keyboard Logitech G19s
Software Windows 11 Pro 21h2 64Bit
Benchmark Scores ► ♪♫♪♩♬♫♪♭
If you think its crap now, then download older 300.xx & 301.xx drivers and you will see what's crap. :D


From what I saw they changed its algorithm since early 304.xx and it actually looks "ok now.
 

newtekie1

Semi-Retired Folder
Joined
Nov 22, 2005
Messages
28,472 (4.25/day)
Location
Indiana, USA
Processor Intel Core i7 10850K@5.2GHz
Motherboard AsRock Z470 Taichi
Cooling Corsair H115i Pro w/ Noctua NF-A14 Fans
Memory 32GB DDR4-3600
Video Card(s) RTX 2070 Super
Storage 500GB SX8200 Pro + 8TB with 1TB SSD Cache
Display(s) Acer Nitro VG280K 4K 28"
Case Fractal Design Define S
Audio Device(s) Onboard is good enough for me
Power Supply eVGA SuperNOVA 1000w G3
Software Windows 10 Pro x64
Lines inside of textures don't have jaggies, providing that AF is used. That's what anisotropic filtering is for, well texture filtering in general.

They most certainly do have jaggies, any time a line doesn't line up exactly with the pixels of the image there will be jaggies, AF helps but only to a point, AA does way more to remove jaggies.
 
Last edited:

Benetanegia

New Member
Joined
Sep 11, 2009
Messages
2,680 (0.50/day)
Location
Reaching your left retina.
They most certainly do have jaggies, any time a line doesn't line up exactly with the pixels of the image there will be jaggies, AF helps but only to a point, AA does way more to remove jaggies.

AF does EXACTLY the same as AA (more so MLAA or FXAA) except it includes a perspective correction on the algorithm (read is better), so what are you talking about? If AF does not get rid of jaggies AA certainly won't. That being said, try to use AF because if you still see jaggies on textures, you are not enabling it properly. I've not seen jaggies on textures in a long long really long time.
 
Joined
Feb 26, 2008
Messages
4,876 (0.83/day)
Location
Joplin, Mo
System Name Ultrabeast GX2
Processor Intel Core 2 Duo E8500 @ 4.0GHZ 24/7
Motherboard Gigabit P35-DS3L
Cooling Rosewill RX24, Dual Slot Vid, Fan control
Memory 2x1gb 1066mhz@850MHZ DDR2
Video Card(s) 9800GX2 @ 690/1040
Storage 750/250/250/200 all WD 7200
Display(s) 24" DCLCD 2ms 1200p
Case Apevia
Audio Device(s) 7.1 Digital on-board, 5.1 digital hooked up
Power Supply 700W RAIDMAXXX SLI
Software winXP Pro
Benchmark Scores 17749 3DM06
AF does EXACTLY the same as AA (more so MLAA or FXAA) except it includes a perspective correction on the algorithm (read is better), so what are you talking about? If AF does not get rid of jaggies AA certainly won't. That being said, try to use AF because if you still see jaggies on textures, you are not enabling it properly. I've not seen jaggies on textures in a long long really long time.

AF is anisotropic filtering, which filters actual textures. Maybe you mean something else than anisotropic filtering.
 
Top