-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
really getting fed up with the posts stating RTX/DLSS does not work this gen
- Thread starter sunshinewelly
- Start date
Caporegime
DLSS is great, but it surely has a finite lifespan? All the new cards will do native 4K easily and the generations that follow will do 8K too. Surely there will be a time when it’s simply no longer needed and economically worth coding into the game?
When looking at a bunch of the Control comparisons, tons of 2.0 upscaled 1440p shots look significantly better than native 4k. I'd still opt for 4k DLSS than 4k native in Control.
Perhaps the future will just be AI based fidelity enhancements.
Soldato
It'll be ages before any card can do native 4k with ray tracing though.
Caporegime
I must admit, even though my 2070S can handle ultra SOTTR at 1440p to an acceptable FPS level, I still think with DLSS, it looks better and while obviously running better than native 1440p. I think it takes my average FPS from 53 to 68 FPS.
Caporegime
True if you want minimum 60fps. Kind of makes a mockery of the graphics card industry when you consider what the new consoles can do for under £500.
Man of Honour
One thing I have noticed about Ray Tracing on the 3XXX cards is it is no where near as big an improvement over the 2XXX cards as some of the outrageous claims that were made before the Ampere launch.
A 3090 is about 35% faster in Ray Tracing than a 2080 Ti/RTX Titan on the Port Royal bench, not bad but no where near the X4 that some were predicting.
A 3090 is about 35% faster in Ray Tracing than a 2080 Ti/RTX Titan on the Port Royal bench, not bad but no where near the X4 that some were predicting.
It is capable of an upto 4x speed up in applications that use ray tracing techniques for all features and the latest DLSS (such as the nVidia marbles demo). Though they overplayed it in the presentation.
That's dlss 1.0, it should look much better if you disable it!
How did you calculate that?
https://www.guru3d.com/news-story/3d-mark-port-royal-raytracing-benchmarks-(geforce-rtx-series).html
https://www.guru3d.com/articles-pages/geforce-rtx-3090-founder-review,28.html
2080ti FE = 7969 pts (36fps)
3090FE = 13545 pts (63fps)
Improvement: 75%
As for the 4x number, yeah its kinda BS but we knew it would be. With RTX2000 they said its 6x faster than the 1080ti but was it in real games? nope only in their star wars tech demo.
Same applies here, is the 3090 4x faster in RT games? Nope, only 4x faster in a tech demo marbles
https://www.youtube.com/watch?v=E98hC9e__Xs
RTX8000 Marbles: 720p 25fps average
RTX3090 Marbles: 1440p 30fps locked
This "4x" number comes from the fact that this Path Tracing demo is able to run at slightly higher fps and 4 times more pixels on the RTX3090.
Last edited:
Man of Honour
Look in the Port Royal bench thread which is all about Ray Tracing and compare the scores, there is about a 35% difference between the TU102 cards and the 3090.
Thats overclocked high end 2080ti's on water vs probably stock (or mild oc) air cooled 3090's at this point. I'd give it a bit more time before you use the Port Royal thread because the 3090 scores will improve when people put them under water and we get high power limit cards.
edit: Yeah the top 2080ti score there is a water cooled Kingpin card and that 11k is freaking massive, thats 22% greater than my 2080ti which is overclocked and air cooled. An 11k 2080ti is not representative of the average 2080ti's owner score, its way above it.
So maybe wait until we see the score for a water cooled kingpin 3090.
As you can see in the Guru3d reviews, when the 2080ti and 3090 are both stock there is a 75% difference.
And as per my other explanation, even though the 3090 may be able to produce 4x pixels compared to Turing in a fully path traced tech demo - it will never translate like that into games because there is a lot of other things going on for the GPU to render. These differences become larger the more of the visuals are ray traced.
edit: Yeah the top 2080ti score there is a water cooled Kingpin card and that 11k is freaking massive, thats 22% greater than my 2080ti which is overclocked and air cooled. An 11k 2080ti is not representative of the average 2080ti's owner score, its way above it.
So maybe wait until we see the score for a water cooled kingpin 3090.
As you can see in the Guru3d reviews, when the 2080ti and 3090 are both stock there is a 75% difference.
And as per my other explanation, even though the 3090 may be able to produce 4x pixels compared to Turing in a fully path traced tech demo - it will never translate like that into games because there is a lot of other things going on for the GPU to render. These differences become larger the more of the visuals are ray traced.
Last edited:
I predicted at least double the performance from a few year ago and ended up a little disappointed in that regard. Wanted them to be powerful enough so turning RT did not tank performance so much.
If hopper ends up an MCM design like rumoured then I think that will see a big leap in RT performance.
That said I do think as new games start coming out we will see the gap between Turding and Ampere get bigger.
It's pretty much following the same history as Tesselation, exactly as I've been predicting for the last year.
Tesselation v1 (AMD 5870): Looks cool, terrible performance
Ray Tracing v1 (Nvidia RTX2080ti): Looks cool, terrible performance
Tesselation v2 (AMD 6970): Much better performance but still a significant hit to framerate.
Ray Tracing v2 (Nvidia RTX3090): Much better performance but still a significant hit to framerate.
Tesselation v3 (AMD 7970): Little to no performance hit anymore.
Ray Tracing v3 (Nvidia RTX4090): ???
Yeah. My expectations of something better was because I expected them to use more die area for it and increase the number of cores quite a bit more. Still, can’t complain too much. Better than 2080Ti for half the price
Yes one reason I'm not upgrading, besides being skint! Third gen on I think ray tracing will start to take off. Once I can do similar or better FPS to my 5700XT at 1440p with RT on I'll jump.
Soldato
Yeah. My expectations of something better was because I expected them to use more die area for it and increase the number of cores quite a bit more. Still, can’t complain too much. Better than 2080Ti for half the price
Now we have the 30 series testing done, it would appear Ampere has fallen a bit short. Like you mention it should be there next gen, just disappointing they were not honest but if Jensen didnt beguile you it would be another Turing flop (Turing - mainly overpriced, people could have stomached the RTX if prices were lower) so they need to shift the units again this gen, albeit power hungry too.
Now let’s see what AMD have to bring to the table.
Do you think it will be better RT performance?