• 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.

The RT Related Games, Benchmarks, Software, Etc Thread.

Status
Not open for further replies.
I think he means full path tracing is 20 years away if we keep increasing RT performance as we are seeing.

Fair enough on that but still just a very rough guess on his end if all he is looking at is hardware advancement. I think there is still a heck of a lot that can be improved via software/drivers methods both from nvidia and amds end as well as developers getting the best from their end, same way they have done with raster over the years.

That and upscaling plus frame generation kind of tech. will be pivotal to bring it much sooner.
 
Last edited:
I think he means full path tracing is 20 years away if we keep increasing RT performance as we are seeing.
Which kinda makes sense given that Quake II is a 25-year-old game :D

The current 'state-of-the-art' in RT is excellent but as beautiful as Cyberpunk and other games can be, we're a long way from running them in the manner of Quake II RTX or Minecraft (even Quake II requires a denoiser - try turning it off!).

*Edit* I believe the only path-traced game that *doesn't* require a denoiser is Lego: Builder's Journey - and that it was an intentional choice to preserve the fine details. That's essentially a mobile game that *should* run at 1000's of frames per second but on my 3070 @1440p with DLSS I only just got 60-ish fps (haven't tried it on my 3090 Ti) - the underlying game isn't causing those slowdowns, it's the insane demands of path-tracing.
 
Last edited:
Speaking of quake RTX, there has been quite a lot of updates lately where amd have had big performance jump:

Radeon RADV Ray-Tracing Performance Further Improved - PLOC BVH Builder


In recent months there has been a lot of RADV ray-tracing optimizations and improvements for maturing the ray-tracing support by this open-source Radeon Vulkan driver within Mesa. The RADV ray-tracing performance is about to take another step forward with a pending merge request providing a PLOC BVH builder that can improve the Quake II RTX performance by around 33%.

Friedrich Vock and Bas Nieuwenhuizen have been working on this Parallel Locally-Ordered Clustering (PLOC) support for building bounding volume hierarchy nodes (BVHs). Vock commented that it uses in-shader global synchronization and a task partitioning algorithm similar to GPURT. GPURT is notable in being AMD's open-source GPU ray-tracing project that recently enabled Vulkan ray-tracing for AMDVLK, AMD's official open-source Vulkan Linux driver and an alternative to Mesa's RADV. It's good to see GPURT being helpful to RADV at least for ideas.

My benchmarking on a 6700XT indicates a 33% uplift for Q2RTX and the nvpro demos. Control suffers a bit from the increased build times, but still gets a tiny bit better (around 47 -> 51 FPS).

Mesa's RADV Radeon Vulkan Ray-Tracing Performance Continues Improving


Hence why I said there is far more to RT than just hardware :p IMO, given RT came about with turing but has only really taken of over the last 2 years, it's quite impressive how quickly and how far it has all come along. But agree, to get the likes of cp 2077 full path tracing is probably a long way of.
 
  • Like
Reactions: TNA
Which kinda makes sense given that Quake II is a 25-year-old game :D

The current 'state-of-the-art' in RT is excellent but as beautiful as Cyberpunk and other games can be, we're a long way from running them in the manner of Quake II RTX or Minecraft (even Quake II requires a denoiser - try turning it off!).

Exactly people forget Quake is a quarter of a century old... One of the few path traced games that actually runs on dual 3090/ti's in NVLINK/SLI and really does scale well too. Now reality it still runs a full settings not so well and games we have now running at full path tracing is really a long way off and his guess of 20 years maybe even a lot off unless you enjoy 5 fps to run games like cyberpunk fully path traced.

I think a lot of people don't realise how much compute you really need for full real time path tracing of high fidelity images.

A lot of these cartoon movies we see from disney/pixar take years to render in large server farms all over the world. The micky mouse power in our computers even the highest current card can't even compete with a simple server setup with gpus and these servers take years to get the results from sometimes. Our little space heaters are good for one thing creating the work that needs to be sent off to the big boys sometimes, unless you want to literally waste your life waiting for the results or even be dead before it even is near finishing.
 
Last edited:
Always told them 10gb isn't enough, poor guys didn't listen

Like I said from the start. 10GB would be fine until next gen cards were out and at most we would have a handful of games that would need more. That is exactly what happened. Was even better than that actually as none of those games were must play ones. Most notable was Far Cry 6 that some went on about, and let’s be honest that was kind of crap. Lol


It's pretty much the exact same game, just new content and snowy winter setting.

Only played for bit so far but so far but seems they need to do some work as:

- dlss and fsr don't look as good as they did in the first one especially if using lesser presets, main issue I have is the moire effect around likes of Miles
- I know this has got RT shadows now but aside from that, game seems to not run anywhere as well as the first one despite the rest of the visuals being the exact same, only thing that could explain this is the different setting i.e. snow/winter vs summer? Maybe the weather particle effect might be hampering performance more? Suppose DF will confirm when they bring their video out

Even reading on frame generation, it seems to not work as well as it did in the first one?

Have no doubt we'll see more patches to improve things though.

Is the whole game in snow? Not sure I would like that, I do prefer the summer setting overall. Based on the issues I will wait and buy it later when the price drops further.
 
Well if you think about it, 'this generation's' path tracing showcase will be Portal RTX which, whilst released in 2007, uses an engine that debuted in 2004 - so Quake II RTX (based on a 25-year old game), debuted in 2019, nearly 4 years later we get a path traced version of a game whose engine debuted *9* years after Quake II.

If we continue along that trajectory we'll have a path traced Cyberpunk 2077 in approximately 10 years - expect it not to be very playable until 2-4 years later though :D
 
Is the whole game in snow? Not sure I would like that, I do prefer the summer setting overall. Based on the issues I will wait and buy it later when the price drops further.

Performance is definitely very weird, same spot of overlooking central park:

SYxeFVk.png

Notice CPU usage too.

Now this isn't like for like, dlss quality 3440x1440 vs 4k dlss quality in miles, pretty sure I did have this bit in one of my 4k videos for the original spiderman game but seems not, either way.... 99% sure in 4k/dlss, fps was nowhere as low in the original as it is in miles for this scene:

BIR663k.png

Don't know if just me but reflections seem higher quality in miles so could be that hitting performance too.

Also, what's weird is going back to 3440x1440, performance doesn't seem to improve as much either. Going to wait for a patch or 2 for miles and finish plague tale requirem for now :)

All in all, Alex from Df will have his video up this weekend so that should confirm what is going on.

As for snow setting and the entire game, not sure, being Christmas setting, I imagine it will be. Snow setting in NY at xmas is the best! :cool: Reminds me of division 1.



PS. Just saying.... FC 6 played fine on my end except for when I ran benchmark mode with no FSR and forced rebar on:cry: :p :D
 
Last edited:
  • Haha
Reactions: TNA
Another great in depth video by Alex/DF :cool:


However, not sure how he has come to the conclusion performance is similar to Spiderman, it is definitely not just a couple or even few fps difference on my end :cry:
 
Another great in depth video by Alex/DF :cool:


However, not sure how he has come to the conclusion performance is similar to Spiderman, it is definitely not just a couple or even few fps difference on my end :cry:

Just had a play, both Spideys maxed out, DLSS on quality, 3440x1440, run roughly the same on my rig, 12700K @ 5.1, 3080 Ti FE under-volted to 1950/0.9V.

Slightly weird though, just as the action starts in the intro (MM) I changed DLSS to balanced, and got lower frames!

Restarted, kept it at quality, and all good, not sure what happened there.
 
Just had a play, both Spideys maxed out, DLSS on quality, 3440x1440, run roughly the same on my rig, 12700K @ 5.1, 3080 Ti FE under-volted to 1950/0.9V.

Slightly weird though, just as the action starts in the intro (MM) I changed DLSS to balanced, and got lower frames!

Restarted, kept it at quality, and all good, not sure what happened there.

Yeah once I get past the intro section, performance improves a bit more, seems to be similar to the first game in that sense with the intro. also being the most demanding bit.

Also had the same with this and the first game when any settings are changed especially dlss, a restart is required. Surprised they haven't fixed that yet.
 
Definitely something not right with my CPU usage in miles morale, went back to check original game to make sure and compare each one in same area (different time of day, miles one being at night), ignore washed out look, was playing in hdr:

A4wNviH.jpg

P05pDCA.jpg

vZEgT1Z.jpg

hmWaE9d.jpg

Had a quick google and seems to be mixed reports, some saying no issues and others also noticing weird performance.

EDIT:

And before you say vram @TNA, it's not as they are both using the same amount there :p
 
Last edited:
  • Haha
Reactions: TNA
Performance in miles morales is quite weird at 8k

Whether you use:

* Native 8k no RT
* Native 8k with RT
* 8k with DLSS2 and with RT
* 8k with DLSS3 and FG and with RT

It doesn't matter, the performance is the same in all of these.

Can someone explain why? My theory is that 8k just requires an insane amount of memory bandwidth to get decent framerates and since the 4090 only has 1TB/s it's not enough no matter how GPU bound you make the game. DLSS3 with Frame Generation doesn't help improve performance at all.
 
Last edited:
Status
Not open for further replies.
Back
Top Bottom