GPU Ray Tracing Performance Comparisons [2021-2022]

[DavidGraham]

More RT 4K Max settings testings from ..

Cyberpunk 2077:
6950XT: 100%/11fps
3090Ti: 200%/22fps
4080: 236%/26fps
4090: 345%/38fps

The 4080 is 18% faster than 3090Ti and 2.36x times faster than 6950XT.

Dying Light 2:
6950XT: 100%/12fps
3090Ti: 216%/26fps
4080: 258%/31fps
4090: 390%/47fps

The 4080 is 20% faster than 3090Ti and 2.58x times faster than 6950XT.

Metro Exodus:
6950XT: 100%/32fps
3090Ti: 137%/44fps
4080: 165%/53fps
4090: 225%/72fps

The 4080 is 20% faster than 3090Ti and 65% faster than 6950XT.

Test kart graficznych NVIDIA GeForce RTX 4080 vs GeForce RTX 3090 Ti - Byłaby rewelacja, gdyby nie zaporowa cena | PurePC.pl

Test kart graficznych NVIDIA GeForce RTX 4080 vs GeForce RTX 3090 Ti - Byłaby rewelacja, gdyby nie zaporowa cena (strona 17) Test i recenzja kart graficznych NVIDIA GeForce RTX 4080 vs GeForce RTX 3090 Ti - Byłaby rewelacja, gdyby nie wysoka cena. Ponad 2 razy wyższa od GeForce RTX 3080.

www.purepc.pl

Chernobylite:
6950XT: 100%/18fps
3090Ti: 194%/35fps
4080: 233%/42fps
4090: 344%/62fps

The 4080 is 20% faster than 3090Ti and 2.3x times faster than 6950XT.


Guardians of the Galaxy:
6950XT: 100%/31fps
3090Ti: 193%/60fps
4080: 222%/69fps
4090: 300%/93fps

The 4080 is 15% faster than 3090Ti and 2.2x times faster than 6950XT.

Watch Dogs Legion:
6950XT: 100%/27fps
3090Ti: 140%/39fps
4080: 174%/47fps
4090: 237%/64fps

The 4080 is 20% faster than 3090Ti and 74% faster than 6950XT.

Cyberpunk 2077:
6950XT: 100%/12fps
3090Ti: 191%/23fps
4080: 241%/29fps
4090: 358%/43fps

The 4080 is 25% faster than 3090Ti and 2.4x times faster than 6950XT.

Nvidia Geforce RTX 4080 – Förväntat dyrt och överraskande bra - Test - Test: Ray tracing

Efter en händelserik månad är det dags att syna Nvidia Geforce RTX 4080 i sömmarna och se vad det går för.

www.sweclockers.com

 

[Silent_Buddha]

Ouch only a single game in there is playable (for me) on a 4080 with RT max settings with my current setup. However, once I have an HDMI 2.1 card, it's 120 Hz and nothing lower if possible. So, none of those would be playable. At that point only the 4090 appears to be almost playable.

That said, I'd, of course, be lowering settings until 120 Hz locked (or as close as possible), so I'd certainly be happy with any of those cards if I didn't have to pay for them.

I have to say that after playing some games at 120 FPS over at my friend's place, I find myself just not really enjoying playing any action games on my home rig (limited to 60 FPS) anymore. I knew I shouldn't have tried it before getting an HDMI 2.1 card, but I couldn't help myself and now 60 FPS is almost entirely unenjoyable and horrible to look at (not smooth in motion).

Regards,
SB

 

[yamaci17]

Yeah honestly, 30-50 is perfectly playable for me, especially in TPS games with a gamepad. I don't really care if some other people deems them unplayable or anything, if a GPU renders 30+ in a specific game, I deem it playable and I deem the GPU capable of running that game at those settings (whatever they may be). Because IT is playable, I can play them just fine, enjoy, gather memories, gather all achivements and achieve all challenges the game may present. Just... stay away from Vsync and that's it.

 

[Frenetic Pony]

Projecting out the 50-70% increase in the 6950-7900xtx, looks like the card is pretty close to the same price/perf as the 4080 in RT titles even with careful normalized benchmarks.

As for Nvidia vs Nvidia, you can go buy a 3090 on Ebay for $800 or less. As this certainly isn't a 50% performance increase over that still fine card, I can see why people without overflowing cash supposedly don't care about buying this thing. Nvidia outcompeting Nvidia isn't the best look, they're going to have to drop the "4070ti" from it's original price by at least $100 or more to get anyone to care about that thing.

 

[Cyan]

Ouch only a single game in there is playable (for me) on a 4080 with RT max settings with my current setup. However, once I have an HDMI 2.1 card, it's 120 Hz and nothing lower if possible. So, none of those would be playable. At that point only the 4090 appears to be almost playable.

That said, I'd, of course, be lowering settings until 120 Hz locked (or as close as possible), so I'd certainly be happy with any of those cards if I didn't have to pay for them.

I have to say that after playing some games at 120 FPS over at my friend's place, I find myself just not really enjoying playing any action games on my home rig (limited to 60 FPS) anymore. I knew I shouldn't have tried it before getting an HDMI 2.1 card, but I couldn't help myself and now 60 FPS is almost entirely unenjoyable and horrible to look at (not smooth in motion).

Regards,
SB

what would you be eager to sacrifice to achieve that goal is the question. 60fps with a solid VRR is a good experience. For 120Hz you need a bit of everything. Tbh, I managed to play quite a few games at 164Hz on my old GTX 1060 3GB -my least favourite GPU to date, of those I've had-. However the sacrifices that were made stopped any possibilities for enthusiasm.

 

[Dampf]

It really depends how you lock that 30 FPS. If you play on a 60 Hz monitor and lock a rock solid 30 FPS with perfect frame pacing, then its smooth in my eyes. I prefer that actually over a higher framerate but with uneven frame pacing (like 50 fps on a 144 hz display without G-Sync/FreeSync)

 

[DavidGraham]

Excessive motion blur (per object and camera motion blur) can make a 30fps game feel very smooth, it can make 60fps feel even smoother, VRR will add even more smoothness to either of them.

For 120Hz though, DLSS/FSR is your best hope in heavy RT titles/heavy Raster titles. Or drop down to 1440p if you prefer native quality.

 

[DavidGraham]

The new episode in The Dark Picture Anthology series is released, it's called The Devil in Me and supports RT AO and RT Reflections.

At 4K max settings, the 2080Ti ties the 6900XT (no pun intended), the 3080Ti is 50% faster than 6900XT, the 4090 is 2.9x times faster than 6900XT.

2080Ti: 33fps
6900XT: 33fps
3080Ti: 50fps
4090: 98fps

At 1440p, the 3080Ti 43% faster and the 4090 is 2.6x times faster.

6900XT: 59fps
3080Ti: 84fps
4090: 154fps

The Dark Pictures Anthology: The Devil in Me тест GPU/CPU | RPG/Ролевые | Тест GPU

Студия, подарившая нам Until Dawn, представляет серию не связанных между собой кинематографичных игр с разветвл

gamegpu.com

 

[DegustatoR]

Ouch only a single game in there is playable (for me) on a 4080 with RT max settings with my current setup.

All of them can be "playable" with the help of DLSS/FSR2. There's also a world of difference in performance usually between RT medium/high and RT ultra. These while affecting the relative fps in a significant way will not affect the relative standing of the cards.

That said, I'd, of course, be lowering settings until 120 Hz locked (or as close as possible)

This is rather pointless on a VRR display as you're unlikely to even notice fps fluctuations between 90 and 120 in 99% of games.

 

[Cyan]

It really depends how you lock that 30 FPS. If you play on a 60 Hz monitor and lock a rock solid 30 FPS with perfect frame pacing, then its smooth in my eyes. I prefer that actually over a higher framerate but with uneven frame pacing (like 50 fps on a 144 hz display without G-Sync/FreeSync)

some people can play really well under those conditions (like in the video below, FX 6300 GTX 1660 High quality, XeSS enabled), but I guess some kind of VRR makes it more palatable, which imo is the actual issue, keeping the image stable.

 

[DavidGraham]

Miles Morales RT at 4K max settings.

The 2080Ti is 12% faster than 6950XT, 3080Ti is 60% faster, the 4080 is 2.27x times faster, and the 4090 is 3x times faster.

6950XT: 24fps
2080Ti: 27fps
3080Ti: 38fps
4080: 55fps
4090: 72fps

Spider-Man: Miles Morales im Technik-Test: Wunderschöne Winter-Weihnachts-Erweiterung zu Sonys Spinnen-Mann-Hit

Diese Seite beschäftigt sich mit der GPU-Performance. Sie finden hier Benchmarks und weitere Analysen zum Spiel.

www.pcgameshardware.de

3080Ti is 60% faster than 6900XT, the 4090 is 2.8x times faster.

6900XT: 29fps
3080Ti: 46fps
4080: 62fps
4090: 83fps

Test wydajności Marvel's Spider-Man: Miles Morales PC - Porównanie kart graficznych NVIDIA GeForce i AMD Radeon | PurePC.pl

Test wydajności Marvel's Spider-Man: Miles Morales PC - Porównanie kart graficznych NVIDIA GeForce i AMD Radeon (strona 13) Czy pójdzie mi Marvel's Spider-Man: Miles Morales PC? Test wydajności kart graficznych NVIDIA GeForce i AMD Radeon w grze Marvel's Spider-Man: Miles Morales PC

www.purepc.pl

 

[DavidGraham]

In the Quake 1 Path Tracing mod, the regular 3080 is 2X times faster than the 6800XT at 4K.

 

[Subtlesnake]

Alex says in the video he suspects AMD cards are not optimized for the game because they perform poorly even with FSR 2.0 enabled.

 

[Rootax]

Alex says in the video he suspects AMD cards are not optimized for the game because they perform poorly even with FSR 2.0 enabled.

I was perplex hearing that. I mean, if RT is the main bottleneck, lowering resolution will help but not solve everything right ? So yeah it's struggling with FSR2 too, I'm not surprised ?

 

[DavidGraham]

Alex says in the video he suspects AMD cards are not optimized for the game because they perform poorly even with FSR 2.0 enabled.

At native 4K, the 3080 is 2X the 6800XT. It's really nothing new here, we've seen this with the optimized Vulkan Quake 2 RTX, where the 3080 is 2.1x times faster than the 6800XT at native 4K, and the 3090 is 2.5x times faster. We've also seen this in Minecraft RTX, where the 3080 is 2.3x times faster than 6800XT and the 3090 is 2.7x times faster.

Test • Nvidia GeForce RTX 4090

Test de la nouvelle carte graphique GeForce RTX 4090 au travers du modèle Founders Edition de Nvidia. Au programme, 26 jeux testés, dont 20 en rastérisation et 15 en Ray Tracing, 13 benchmarks, des mesures de consommation, nuisances sonores, températures et imagerie infrarouge.

www.comptoir-hardware.com

As for the bug Alex mentioned, it happens specifically with FSR2, which probably means the FSR2 implementation is bugged and is not applied correctly. But, native numbers seem within the normal performance range for the 6800XT in any path traced game.

 

[Kaotik]

At native 4K, the 3080 is 2X the 6800XT. It's really nothing new here, we've seen this with the optimized Vulkan Quake 2 RTX, where the 3080 is 2.1x times faster than the 6800XT at native 4K, and the 3090 is 2.5x times faster. We've also seen this in Minecraft RTX, where the 3080 is 2.3x times faster than 6800XT and the 3090 is 2.7x times faster.

Test • Nvidia GeForce RTX 4090

Test de la nouvelle carte graphique GeForce RTX 4090 au travers du modèle Founders Edition de Nvidia. Au programme, 26 jeux testés, dont 20 en rastérisation et 15 en Ray Tracing, 13 benchmarks, des mesures de consommation, nuisances sonores, températures et imagerie infrarouge.

www.comptoir-hardware.com

As for the bug Alex mentioned, it happens specifically with FSR2, which probably means the FSR2 implementation is bugged and is not applied correctly. But, native numbers seem within the normal performance range for the 6800XT in any path traced game.

If FSR numbers are off the problem is elsewhere. It's literally running it at lower res

 

[Inuhanyou]

If throwing more gpu power at a problem solved RT, then FSR2 would have to be bugged since your giving the GPU more headroom

 

[DavidGraham]

If FSR numbers are off the problem is elsewhere. It's literally running it at lower res

It could be not applying the lower resolution correctly, and again native numbers for the 6800XT are in line with Quake 2 RTX and Minecraft RTX, the new variable here is FSR2, which means it's probably the source of the bug now.

 

[Kaotik]

It could be not applying the lower resolution correctly, and again native numbers for the 6800XT are in line with Quake 2 RTX and Minecraft RTX, the new variable here is FSR2, which means it's probably the source of the bug now.

Nope, that's not an option. It literally runs the game at lower resolution. There's no ifs and buts there.

 

[CarstenS]

It could be not applying the lower resolution correctly, and again native numbers for the 6800XT are in line with Quake 2 RTX and Minecraft RTX, the new variable here is FSR2, which means it's probably the source of the bug now.

In the beginning of the Upscaler comparison, the AMD card does hit 60 fps, coming from a measly 17 fps in native 4K. So FSR is working as it should - at least in principle. When the scene gets more complex and more dynamic updates to lighting and possibly more divergent rays, performance tanks on the Radeon, going down to 26 fps.

Remember, this is a purely ray/path traced (and denoised) image. So not much here which you can build perf upon like a healthy baseline of raster performance as in many hybrid games.

FWIW, I could easily achieve unlocked fps with DLSS / FSR ultra performance modes or the Vintage "CRT" or low res setting.