GPU Ray Tracing Performance Comparisons [2021] *spawn*

Like Nokia mobile phones 14 years ago?

 

Oh, come on...

EDIT: Like, seriously, what the hell kind of post is this. Can we maintain some kind of decorum on these forums like we used to have?

 

It's called a "comparision". Kaotik defined his own "real world". Nokia phones were great for calls and messages. Problem was the iPhone has changed the landscape even when most people used their mobile phone for calls and text messages.

in Doom Eternal Ampere is ~50% faster than RDNA2. Doom Eternal is playable with >100FPS on a 3060. And dont let me start with DLSS which makes the "real world" a more surreal enviroment...

 

Turing holds its own quite well against RDNA2 in non ray tracing environments. Ampere even in non-ray tracing has the more advanced features over RDNA2, being ahead in compute aswell.

 

That's an interesting "real world", cause in my real world everything RDNA2 based below 6800 is losing to Ampere about everywhere - if we compare the cards by their MSRPs.
I've said it already - if not for the cryptoshit pricing situation RDNA2 lineup would be very weak in gaming in comparison to Ampere.

 

Err, what? RDNA2 cards go neck to neck with Ampere at least up to 4K in Doom Eternal, higher than that no clue. 6900 XT is leading the pack at up to 1440p, 3080 Ti and 3090 get the crown at 4K. 6800 XT is beating 3080 no matter the res and 3080 beats 6800 while 6800 beats 3070 Ti etc.
Depending on resolution you can go back at least couple gens and still get over 100 FPS, not sure how that's relevant.

Apples to apples, not apples to rehydrated oranges.

So instead of looking at the whole stack you pick arbituary card(s) supporting your view based on MSRP which everyone knows you can't get any of them at anyway.

 

Ampere is faster in every resolution: https://www.pcgameshardware.de/Doom...ng-RTX-Update-DLSS-Benchmarks-Review-1374898/

Ampere will always be faster with ML approaches. I think it will be around 30% with TensorCores over DP4a. So add this to the ~40% advantage with Raytracing.

 

Doom Eternal...?
Do you see how laughable your argument is..? Today's $1k video cards are not viable Ray-Tracing solutions.

Does anyone play Warzone with RT on....? Out of 30 million, how many play with it on?

The silence is deafening

 

I didn't specify "without RT" for nothing. Take a look at the same game without RT.

Maybe the difference is 30%, maybe not, it's irrelevant at this time.

 

You are confusing gaming and RT.

Again, people don't really game with RT, they just try it out. RDNA2 is more advanced architecture for Games. Ampere is just a discounted quadro meant for content creation, that NV sells at a discount to gamers, until they can engineer a game specific architecture.

And without Pandemic RDNA2 cards would outsell the more expensive and outdated Ampere. More human beings the planet-wide are AAA gaming on RDNA2 than Ampere. More games will be developed for RDNA2 than Ampere.

Ray tracing is not a buying metric, it is marketing hype...

 

Can people not interested in discussing the topic (GPU Ray Tracing Performance Comparisons) GTFO and post elsewhere? Make a new thread and crap on RT all you want, just not here, please.

 

You said "real world". I dont care about more than 144fps because my 4K display is capped at 144Hz. So Doom Eternal with Raytracing is full playable in 4K with Raytracing on nVidia GPUs.

Like DLSS was irrelevant two years ago? I remember that people said you should not buy a Turing card because of DLSS 1.0. That was really a great advice...

 

Sums it up.

 

It helps to read the whole sentence instead of picking just the words that you can fit into your narrative.

And still is for me and many, many others. Just like every other scaling crap.

 

Does that mean native + TAA is also crap because it often looks worse than the better implementations of DLSS? Seems like a losing proposition.

 

Yes, no doubt the other 17% of the discrete GPU market.

 

I am looking at the whole stack, it's you who picks something out of it.
The whole RDNA2 stack is weak at its official MSRPs.
Higher end is on par with Ampere until RT hits it which means that it's losing at the same MSRP due to the lack of RT and DLSS.
Lower end is losing to Ampere MSRP vs MSRP even without RT.
Essentially all RDNA2 lineup is overpriced - but we don't see the results because, well, MSRPs aren't relevant.
Still, if we look at where prices are right now in some countries RDNA2 cards are actually more overpriced than Ampere ones - which means that they are in an even worse relative position than where they'd be at MSRPs.
So yeah, real world and all.

To get this train back on track:


https://news.xbox.com/en-us/2021/10/21/halo-infinite-the-great-journey-on-pc/

 

Nice. Didn't realize Halo Infinite was going to have RT. It's probably going to be a light-touch implementation, but hopefully it makes a positive impact (however small).

 

Every little bit helps.

 

The numbers for RT performance loss look pretty decent for AMD right now, 55% loss on 6900XT vs. 42% on 3090. It's even better if you consider that the Navi21 is actually similar to GA104 which also takes a 50% hit with RT.

Even 3dcenter note that AMD can make up for it if they've higher rasterization performance than nvidia:

The MCM design might probably pull off 30% over 4090, at which point the RT performance would be equal if things remain the same.