Radeon 6000 Series Raytracing Performance Review

https://i.imgur.com/FTyN6UF.jpg
Radeon 6000 Series Raytracing Performance Review


Introduction

Good afternoon readers, or at least, it’s afternoon when I’ve started typing this, today we are going to be performing a people’s review of AMDs Raytracing performance with their 6000 series of GPUs. What’s so different about this you might be wondering, well you have all more than likely seen a smattering of Raytracing performance numbers here and there but most of them are synthetic, the likes of 3Dmark’s RT test and Cryteks Neon Noir, but here today we are going to be looking at AMDs Raytracing performance in actual games to see how it performs, we aren’t going to be using half measures either from 1080p to 4K Raytracing will be utilised to the fullest extent each tested game allows.

Right then, let’s get down to it.


Test Setup

CPU: AMD Zen 3700X @ 4.2GHz 1.2v
Mainboard: MSI X570 Tomahawk
RAM: 2x16GB Kevv BoltX 3600MHz, optimised timings
GPU: Powercolor 16GB 6800XT Red Devil & 12GB 6700XT Hellhound
Storage: 250GB Hynix SL301 SATA SSD, 250GB & 500GB Asgard NVME PCI-E 3.0 SSDs, 2TB Seagate Barracuda
Opticals: 24x Lite-On iHAS324 DVD-RW, 16x HP BH40N Blu-Ray
Sound: Xonar DX 7.1, Realtek ALC1200
PSU: EVGA 1000w Supernova G2
OS: Windows 10 Pro x64 (latest ISO) and all updates
Case: NZXT Phantom 530

Adrenalin drivers used will be 21.5.2 as I’ve been able to quite thoroughly vet this driver set over time to ensure there are not any unexpected abnormalities in testing.


Benchmarks

We’re off to the races then it’s time to dig in to some performance numbers.

https://i.imgur.com/SwP1Caw.jpg

Before getting to the Raytracing results themselves we will first get to the preset I chose to use as some of you may well be wondering, the explanation is simple; Metro combines many (too many) options into the presets and ultimately the only real difference between the “High” and “Extreme” presets is the shading rate which anyone who has experimented with can attest to it hitting framerate hard for very little visual gain, the other reason, and this is true for any more modern title tested today, is that we want to look at RT performance here without overburdening the GPU with higher settings as doing so wouldn’t give us as true of a reading on RT performance impacts.

Ok, to the results then. Both the 6800XT and 6700XT show good numbers here with the former being playable all the way up to 4K with Raytracing at “Normal” while the 6700XT musters 1440p with RT set to “High”. AMD are off to a good start here for their first foray into RT, the 6700XT actually looks a little better in this title with it managing the second Raytracing tier for the advertised 1440p resolution the card is good for while the 6800XT manages the first tier of the targeted 4K resolution. I’d like to say there is a significant difference between RT at “Normal” vs. “Extreme” but I really couldn’t see any visual difference so you can safely use the lowest setting and profit from the performance improvement.

https://i.imgur.com/EchkKnL.jpg

Control is considered by many to be one of the better examples of Raytracing use however it doesn’t have a built-in benchmark like Metro does so we have to improvise and make our own using FRAPS, these results were obtained from the start of the game where lots of RT effects are on full display.

Results aren’t as impressive here with Raytracing on both cards being a little disappointing realistically the 6800XT maintains playable frames up to 1440p and a “High” RT setting while the 6700XT is just a notch behind holding 1440p and a “Normal” RT setting. I couldn’t see any difference between the “Normal” and “High” RT settings making the “High” quality setting seem more like feature creep than necessity and again in terms of targeted resolutions with RT the 6700XT looks just a mite better but at this point it is important to note that while the 6700XT is managing “acceptable” performance in both tests so far that performance is sub 60FPS with minimal amounts of Raytracing in Metro, which is not ideal for a card AMD are marketing as ideal for 1440p.

https://i.imgur.com/sb9pcRf.jpg

Up now is Cyberpunk, a game which is controversial for many reasons but none the less it does remain popular as a RT performance yardstick and a point of curiosity for many, it is for these reasons why Cyberpunk is used as part of these tests. The “Ultra” preset was chosen because the game isn’t exactly a looker nor is it particularly demanding, disable RT and even with the “Ultra” preset you’ll be seeing performance roughly triple or even quadrouple. With the 6800XT for instance without RT you’ll easily see a fairly constant 60FPS at 4K. Cyberpunk does not have a built-in benchmark so once again improvisation comes into play and results were recorded with FRAPS while making a couple laps of Jig-Jig Street at night.

As the results show Cyberpunk is not a good title for AMD, RT performance is poor with all RT features enabled, I wouldn’t necessarily say this is where there is anything wrong on AMDs side though but rather more an indication of just how unfinished Cyberpunk currently is. Suffice to say if you choose to enable RT in this title leave Raytraced reflections and shadows disabled, even then you’ll find yourself having to reduce other graphical settings to even get playable frames at 1440p on both the 6800XT and 6700XT hence leaving RT disabled altogether is the better choice here, not that you’re really going to notice a visual difference anyway with RT disabled.

https://i.imgur.com/MS0wXwh.jpg

While being an older title the latest patched version of SotTR includes Raytraced shadows and mercifully a built-in benchmark so despite being a slightly older title still proves itself useful.

On to the final RT results then, across the board frames are playable for the 6800XT with most people likely to stick to 4K “Medium”, or perhaps “High” if you don’t mind sub 60FPS. To a lesser degree the 6700XT is also comfortable at 4K “Medium” and as with all the other tested titles for this article I was hard pressed to tell any difference between minimum and maximum RT settings.


Conclusion

Well here we have it, a short but useful analysis and review of Raytracing performance on AMDs 6000 series of GPUs. Games with Raytracing are still rather thin on the ground but there are at least enough available now to look at RT and see what it can, or potentially can, offer.

With the exception of Control where the developers apparently just did not incorporate the usual developer tricks to simulate reflections such as screenspace or cube maps disabling RT has a very noticeable impact on visuals. Fire up Cyberpunk or SotTR though and you really won’t be able to tell the difference with RT enabled vs. disabled unless you stop and really look for it and even then the visual improvement is minimal.

Metro Exodus Enhanced Edition is a bit of a special case however and why it has been included today being the first title to require an RT capable card to play... and that was a mistake by 4A in my opinion. Why? Because Raytracing really doesn’t add all that much even here, in fact the “original” version of Metro Exodus, in my opinion, looks better as you’ll see things in the benchmark like the moonlight being reflected off of the water surface to the nearby bank when you turn RT on, a very noticeable effect bizarrely and ironically absent from the Enhanced Edition’s benchmark, with RT disabled in the older version of Metro outside of that effect you will again be challenged to notice any significant difference.

The problem for all of these titles is that developers have gotten so good at simulating these types of effects is that the only really noticeable difference right now is the massive performance impact enabling Raytracing incurs. In the long game I can see RT being beneficial but, and this is the crux of it all, only if tools are developed, the technology continues to evolve and does not stagnate, and the time is put in to develop RT as a true evolution and replacement for old developers tricks which in time will benefit developers by allowing them to add these types of effects much easier and faster than the old methods.

Speaking of performance leads us nicely to the final part of this analysis and review, the heavy toll required for the use of RT for many makes it not worth while, there is some truth to this, namely RT is going to find its place in single player games, most notably RPGs such as Elder Scrolls or The Witcher 3 (an RT edition will be released for TW3 in time), where immersing the player is most critical and performance less of an issue. In multiplayer titles though ask a gamer if they want pretty visuals or FPS and you’ll be told preferably both but if they have to choose they will take FPS, meaning screenspace and cube map reflections are far from dead until RT performance can be improved and the technology evolves. All of these things considered the 6700XT is a reasonable card but all is not sunshine and roses for it while rasterisation performance is quite strong RT performance is lacking in all but the oldest tested game, Shadow of the Tomb Raider, but we will get to the 6700XT in detail in the forthcoming review for it so keep your eyes peeled for that.