GPU Ray Tracing Performance Comparisons [2021-2022]

6900 XT is on par with 2080 Ti here because it has way better raster performance, obviously, not because of the nonsense "first gen RT".

1) That was not the discussion you think you were replying to. There were meritless accusations from the same people who are back pating themselves on the back and pressed "like" in your post just now, that AMD RE8 RT performance was due to manufacturer tampering the game. - Fanboyism - if you will.



I responded that there are other games that imply the real performance difference is neither just 30% worse nor 300% worse. Someone came back two weeks later and posted a 74% result.



2) The 6900XT is on par with the 2080 Ti here because both manufacturers used different methods, with different hardware, with different advantages, that resulted in both GPU's being their respective first generation RT products. One achieved the result from being superior in one aspect of the rendering pipeline, and the other from another aspect. What are you trying to say? that water is wet? nonsense what?
 
I responded that there are other games that imply the real performance difference is neither just 30% worse nor 300% worse. Someone came back two weeks later and posted a 74% result.

The church scene shows the same performance delta like Control and Cyberpunk. Something you thought would be more outliners than normal: https://forum.beyond3d.com/posts/2203009/

I came back because i saw the numbers and remembered your posting.
 
Games that are light (er) on ray tracing obviously wont have as large a gap as the ones who have heavy use of the feature. Also, as pointed out, its kinda obvious that something like a 6900XT is offering a much higher raster performance as opposed to a 2080Ti, as raster performance still matters since theres not many games abandoning rasterization completely yet.
 
The church scene shows the same performance delta like Control and Cyberpunk. Something you thought would be more outliners than normal: https://forum.beyond3d.com/posts/2203009/

I came back because i saw the numbers and remembered your posting.

I see your point of view, but I presented Metro as an example of unfair criticism towards RE8 which was being massacred for its better than expected performance. Not as an example of how good AMD *really* is. Completely different things.
 
In Metro Exodus, the 3090 is 67% faster than 6900XT @1440p Ultra RT, according to HUB test. I expect the difference to grow beyond 80% @4K.

Who said Metro is disturbing the RT performance detla trend again?
The 3090 has a much bigger lead in RT over the 3080 than the 6900 XT has over the 6800 XT (And the 3090 has a significant lead at 4K even in rasterization!). So it's pretty easy to find huge leads even in relatively better performing titles. In that context, you should look at the 6800 XT in comparison to the 3080/3070. And here we see that the 6800 XT is close to the 3070/ 2080 Ti at 1440p, which is a notably better performance than seen in Cyberpunk/Control.

Eurogamer even mention it in their review:

https://www.eurogamer.net/articles/digitalfoundry-2021-nvidia-geforce-rtx-3080-ti-review?page=6
 
Last edited:
The 3090 has a much bigger lead in RT over the 3080 than the 6900 XT has over the 6800 XT (And the 3090 has a significant lead at 4K even in rasterization!). So it's pretty easy to find huge leads even in relatively better performing titles. In that context, you should look at the 6800 XT in comparison to the 3080/3070. And here we see that the 6800 XT is close to the 3070/ 2080 Ti at 1440p, which is a notably better performance than seen in Cyberpunk/Control.
Benchmarking RT is highly scene dependent, what Eurogamer tests is different from HUB or other outlets, which yields different results.

Generally speaking, the more you use RT in any scene the higher the difference between AMD and NVIDIA. This remains true irrespective of the game or the RT implementation.
 

Wow that video was really salty. How many times did he need to say that Nvidia is releasing the 3080 Ti to make more money. Of course this is a radical idea compared to the other reasons companies sell stuff.

The best part was at the end when he offered that Nvidia was “blown out of the water” by FSR at computex and that DLSS is dead. Lol what? Did he just get back from the future where FSR has already been tested?

I like HWUB but that video wasn’t their best work.
 
The best part was at the end when he offered that Nvidia was “blown out of the water” by FSR at computex and that DLSS is dead. Lol what? Did he just get back from the future where FSR has already been tested?
Once more, they are SEVERELY AMD biased, downplaying RT and DLSS to the best of their abilities. Even now, they just can't admit that their choice of 5700XT over RTX 2070 Super was a grave mistake.
 
Once more, they are SEVERELY AMD biased, downplaying RT and DLSS to the best of their abilities. Even now, they just can't admit that their choice of 5700XT over RTX 2070 Super was a grave mistake.

It's rough love, they really love nvidia. I hope the eth limiter will help supply. Slightly more expensive price can also help as it slims down the profit margin for scalpers/miners which could potentially help cards ending up into consumers hands. I'm not trying to justify expensive prices. I wish situation would return to normal but we are not living normal times just yet.


It's convenient to forget yields. Not every chip yields 3090 or 3080ti. But hey, makes a good click bait to dumb things down.

 
RT put more stress on the cpu too, right ? And the more the scene is complex, the more cpu power will be requiered ?
 
Once more, they are SEVERELY AMD biased, downplaying RT and DLSS to the best of their abilities. Even now, they just can't admit that their choice of 5700XT over RTX 2070 Super was a grave mistake.


I wouldnt call it a grave mistake. They were not cards in the same price bracket. 5700Xt was a hundred dollars cheaper. And back when they launched, the RT and DLSS advantages were not that widespread for them to be a very big factor. Today, for sure, a 2070 is a much better card for what's available.

But that stuff at the end, Steve actually sounds a bit insane there. He's not even trying to hide it anymore. Im sure it wont stop him from trying to be smug on reddit. The entire internet is laughing at amd's FSR and he's saying amd delivered a killer blow to nvidia and killed DLSS ? At a time where DLSS is a plugin in UE4 and multiple games are coming out monthly, since last year now.
 
Benchmarking RT is highly scene dependent, what Eurogamer tests is different from HUB or other outlets, which yields different results.


Generally speaking, the more you use RT in any scene the higher the difference between AMD and NVIDIA. This remains true irrespective of the game or the RT implementation.
That can be true and it can also be the case that on average Metro Exodus is less reliant on RT than some other titles, leading to better average performance on AMD cards. And that the "gap" between AMD and Nvidia depends on which product SKU you are comparing, since performance scaling on AMD and Nvidia is different.

The 67% faster seems like a lot in isolation, but we have Cyberpunk benchmarks showing the 3090 87% faster at 1440p (https://www.kitguru.net/components/...x-3080-ti-review-ft-gigabyte-inno3d-palit/21/). So if it's the case that the most RT heavy scenes in Metro are still less demanding than the most RT heavy scenes in Cyberpunk, then Metro can still be an "outlier", without doing anything to disprove the basic relationship you describe.
 
That can be true and it can also be the case that on average Metro Exodus is less reliant on RT than some other titles, leading to better average performance on AMD cards. And that the "gap" between AMD and Nvidia depends on which product SKU you are comparing, since performance scaling on AMD and Nvidia is different.

The 67% faster seems like a lot in isolation, but we have Cyberpunk benchmarks showing the 3090 87% faster at 1440p (https://www.kitguru.net/components/...x-3080-ti-review-ft-gigabyte-inno3d-palit/21/). So if it's the case that the most RT heavy scenes in Metro are still less demanding than the most RT heavy scenes in Cyberpunk, then Metro can still be an "outlier", without doing anything to disprove the basic relationship you describe.

Metro enhanced PC edition is unique in a sense that it requires RT and doesn't need to make a compromise to run without. If anything maybe metro is showing that going all into RT allows better optimization than trying to forcefully bolt RT on top of existing game engine. Enhanced edition runs surprisingly well with all the additional RT stuff when comparing against the non enhanced edition with its more limited RT implementation. dlss2.0 to replace dlss1 in enhanced edition is also a welcome addition.
 
Last edited:
Enhanced edition runs surprisingly well with all the additional RT stuff when comparing against the non enhanced edition with its more limited RT implementation.
I dunno seem to run about the same in my benchmarks, maybe slightly slower. I'd say that it shows the progress which was made in applying RT to games during the last couple of years - being smarter in how you use RT h/w. The fact that they've dropped non-RT h/w support is not necessarily relevant here.
 
I dunno seem to run about the same in my benchmarks, maybe slightly slower. I'd say that it shows the progress which was made in applying RT to games during the last couple of years - being smarter in how you use RT h/w. The fact that they've dropped non-RT h/w support is not necessarily relevant here.

Enhanced edition has a ton more RT stuff going on than the original metro exodus had.
 
Back
Top