GPU Ray Tracing Performance Comparisons [2021-2022]

And I wonder what is it you're doing in compute for VLIW4 to be 5x faster than scalar.
Most likely he was using something highly not optimal on nv graphics like the shared memory atomics (they were emulated via slow global memory atomics).
Highly unoptimized (for Kepler) compute shaders were a common thing in the Kepler/GCN generation due to GCN presence in consoles and some specific to Kepler opts requirements, gladly it was fixed just in 2 years in Maxwell.
 
Yep, but the reason it's fast is something new. I should get done to limit the risk it will die with me in secret... :)

Waiting for reasonable prices. Though, it's just compute and nothing has change here since ages aside subgroup functions and fp16.

No, 59 series was 1st or maybe 2nd gen of GCN on PC, Vulkan works. Though maybe i remember the model number wrongly. Could be 5950, or even 5870. But i'm sure it's a 3TF GCN model. I still use it for benchmarks.

Yes, often but not extensively. I have learned Kepler did emulate this in VRAM(?), so i assume it's the major reason of it being so slow for me.
IDK what Fermi did, but 670 was exactly as fast as 480 overall, while showing better perf. in cases where atomics to LDS were used.
Whatever, when NV dropped driver support for Kepler recently, i did celebrate it's death :D

I think you confuse 5970 and 7970, 7970 is GCN 1.0.
 
Last edited:
Highly unoptimized (for Kepler) compute shaders
Ofc. i used atomics only if alternatives like scan algorithms were slower, so that's no real argument. Maybe NV lacked behind with atomics, like they did with async compute as well. Seems only Ampere now finally gets this right, after so many years.

I think you confuse 5970 and 7970, 7970 is GCN 1.0-
Exactly - you're right, but it's a 7950 then.
 
Latest new HW i have tasted was FuryX vs. GTX1070. AMD still had a 10% lead if normalizing on teraflops
In absolute terms which one was faster in general? I am asking because FuryX aged horribly in the last two/three years of gaming.
Whatever, when NV dropped driver support for Kepler recently
Coincidentally, AMD dropped FuryX's support, and the entire family of GPUs before it last month too.
 
No, 59 series was 1st or maybe 2nd gen of GCN on PC, Vulkan works. Though maybe i remember the model number wrongly. Could be 5950, or even 5870. But i'm sure it's a 3TF GCN model. I still use it for benchmarks.
7950 probably? 5000 series was VLIW5, 6000 was VLIW4. Neither support Vulkan or DX12.
If your code is faster on GCN1 than on Kepler then it's not a surprise since Kepler (lower end one especially) wasn't that good in compute, especially when it was being ran alongside graphics.
This was "fixed" in Maxwell, and Turing has gone beyond GCN and even RDNA while Ampere added a lot of performance.
If your compute code is still faster on modern AMD h/w then it is doing something really weird.

Whatever, when NV dropped driver support for Kepler recently, i did celebrate it's death
They didn't. R470 will provide support for Kepler till ~Oct (Win11 launch would be a cut off point if I had to guess) and patches to R470 branch will still be released till the end of 2024.
 
They didn't. R470 will provide support for Kepler till ~Oct (Win11 launch would be a cut off point if I had to guess) and patches to R470 branch will still be released till the end of 2024.
They announced retirement though: October 2021.
https://nvidia.custhelp.com/app/answers/detail/a_id/5202

Coincidentally, AMD dropped FuryX's support, and the entire family of GPUs before it last month too.
My 7970 silently weeps all over its still competitive 6 GByte.
 
They announced retirement though: October 2021.
https://nvidia.custhelp.com/app/answers/detail/a_id/5202
From active development branch (aka "Game Ready" drivers).
R470 will remain in LTSB till the end of 2024 or so.
Thus there will still be R470 driver releases with Kepler support which will fix newly discovered vulnerabilities at least.
So it's still on support, just more of "life" type than "full".
 
In absolute terms which one was faster in general? I am asking because FuryX aged horribly in the last two/three years of gaming.
FuryX was faster, it also has more TF (8.6 vs. 6.4). IIRC 1070 had 0.6 times perf. of FuryX.
I always wondered why GCN did not shine with games. People say rasterization can't saturate, and frontend was wrong. But i have only basic experience with graphics pipeline.
If your compute code is still faster on modern AMD h/w then it is doing something really weird.
Will see. TBH, i have a hard time believing anything could beat GCN, not even RDNA2 ;) So let's i hope i'm wrong in any case.
 
Will see. TBH, i have a hard time believing anything could beat GCN, not even RDNA2 ;) So let's i hope i'm wrong in any case.
RDNA2 hasn't really improved its compute over GCN so it's very possible that RVII will in fact beat 6900XT in a purely compute workload.
AMD is kinda confirming this themselves by using GCN as their CDNA base.
 
RDNA2 hasn't really improved its compute over GCN so it's very possible that RVII will in fact beat 6900XT in a purely compute workload.
With 64 CUs running at ~1.7 GHz vs. 80 CUs @ 2.3-ish? It's not like there was an obvious regression in capabilities elsewhere in the CU.

In memory bound workloads, maybe. But then you're not measuring compute, but memory.
 
With 64 CUs running at ~1.7 GHz vs. 80 CUs @ 2.3-ish? It's not like there was an obvious regression in capabilities elsewhere in the CU.

You mean 40 WGPs?

Edit: It doesn't matter, I might just prefer the new identifier because it's a strong remainder that it's not really the same unit here (as in metric).
 
Last edited:
You should definitely try to get your hands on some newer hardware. Fury and Pascal aren’t really relevant today.

Was reading his findings (intresting), and was also thinking, get some newer hardware :p I really hope Joej gets ahold of RDNA2/Ampere hardware soon enough.

I did not notice how my 'it's better' would be perceived, and when i realized, it was too late. Sorry for the fuzz and rudeness.

Oh ok doesnt matter, you dont have to be sorry either ;) It doesnt matter if someones wrong or not, aslong as the discussions are fun and intresting, which i think it is now.

Now it's clear i meant API restrictions, not HW performance. Sadly for me the HW is only as good as the API, because there is no other way to use it.

I can imagine. I also think API's can be improved upon (they have done so), and i think they will over time.

For me there is this magic number of about 10. If perf difference is lower, we can scale. If it's higher, it enables new things not possible on the low end. Usually, within a console generation, we keep below this number. And til yet that's also the case with raytracing.

While true, its also true that scaling really has come long ways already with last generation, and its due to get even better, in special that were in for the rolling generations and cross-gen thing.

The topic of 'console holding back PC' (and others) feels pretty artificial. It's like men arguing which football team is better, although they both love football and watch the same games.
Contrary to that, API limits are holding back for real. Independent of HW, better APIs can enable better performance. On the other hand, performance benchmarks always fall back to platform or vendor wars, with no general profit. So it's beyond me why the former is ignored, while the latter becomes public religion. (Well, i see i post this in the wrong thread, but still.)

Somewhere true, somewhere not :p Ofcourse the consoles set the baseline, which does limit what would be possible if every dev would target the latest and greatest hardware available (crysis, SC etc). Not saying its feasible to target high end pc hardware, but its quite logical to me that consoles 'hold back' newer hardware.

Benchmarks are one tool to measure performance, then we have real world measurements in actual games. In both, theres quite a large gap between console/pc, in special ray tracing. bad API's or not ;)

Devs want to be able to be creative on PC as well. Also this argument is independent of the initial performance we start from.

Your going to be having abit more creative (optimizing) for a 5700XT raw perf class machine vs say the pc market wheres 36TF compute power available. As said, i think scaling is the future anyway, theres not much around it i think.

Whatever, when NV dropped driver support for Kepler recently, i did celebrate it's death

Those GPU's have lived a very, very long supported life though. And even when driver support ends, you can still play most new games for a good while to come.

Get some modern Turing or even Ampere hardware whenever you can, NV has improved alot in the compute department these days.
 
Ofcourse the consoles set the baseline, which does limit what would be possible if every dev would target the latest and greatest hardware available (crysis, SC etc). Not saying its feasible to target high end pc hardware, but its quite logical to me that consoles 'hold back' newer hardware.
It's logical, but not practical. Proof: SC is a myth. If it ever comes out, consoles will likely be able to run it, and it may even look dated. To me it already does.
Crysis 2 was criticized about consoles holding it back, while IMO it was technically way more impressive than the first.
At current day we just lack examples of consoles holding back, because there are no glorious PC exclusive games to proof that.
Which is very sad i think. But not because of tech, it's the game design around 'gamepads', which is more like a TV remote than a proper interface 'to be in the game'. It's impossible to make games worth to play. Immersion is impossible. Fun? Impossible. Skill and challenge? Impossible.
I speak about 3D games. Super Mario was super good! But the critically acclaimed Super Mario 64? Frustrating camera controls. A need to control a damn camera already rules out immersion. Meh - no thanks.

So i fully agree consoles hold back PC games. But the issue is much worse than technical downgrades or compromises. Quake 3 looks shit in comparison to CoD, but the bots were moving and shooting at the same time.
I wonder if we should extend the term 'scaling' to game design as well, to address such shortcomings and adapt to platform. No wonder PC players only play MP games like Counterstrike or Fortnite, while being hungry for good SP games and not getting them.
 
It's logical, but not practical. Proof: SC is a myth. If it ever comes out, consoles will likely be able to run it, and it may even look dated. To me it already does.
Crysis 2 was criticized about consoles holding it back, while IMO it was technically way more impressive than the first.
At current day we just lack examples of consoles holding back, because there are no glorious PC exclusive games to proof that.
Which is very sad i think. But not because of tech, it's the game design around 'gamepads', which is more like a TV remote than a proper interface 'to be in the game'. It's impossible to make games worth to play. Immersion is impossible. Fun? Impossible. Skill and challenge? Impossible.
I speak about 3D games. Super Mario was super good! But the critically acclaimed Super Mario 64? Frustrating camera controls. A need to control a damn camera already rules out immersion. Meh - no thanks.

I think things are just different these days. The 'crysis' (insert HL2, Doom 3 etc) days are over i think, its just not feasible anymore. Were shifting to great scalability instead.
I dont think we really need glorious pc games either, just because of scaling. CP2077 on pc is still the ultimate experience for that game, and quite much so if you have the hardware for it. I agree on the gamepads, though i think its all up to what you like personally and what kind of games you play. I use the controller quite much for my main pc, mostly SP games ofcourse :p

So i fully agree consoles hold back PC games. But the issue is much worse than technical downgrades or compromises. Quake 3 looks shit in comparison to CoD, but the bots were moving and shooting at the same time.
I wonder if we should extend the term 'scaling' to game design as well, to address such shortcomings and adapt to platform. No wonder PC players only play MP games like Counterstrike or Fortnite, while being hungry for good SP games and not getting them.

Those games are by far the most popular on consoles aswell.... good SP experiences for sure exist on pc too. Not to forget Xbox/MS=PC. Whatever exclusive or Xbox game is on that console series will be on PC aswell. + Sony dropping more and more of their games on PC, aswell as showing intrest in the platform in general. I think its a nice middle-land ;)

Technically, visually etc, you have the best versions aswell, and as the gen progresses, the differences become larger and larger.
 
CP2077 on pc is still the ultimate experience for that game
I love watching the world, but shooting is no fun. Enemies are bullet sponges and barely move. It's more of an interruption between the cutscenes to me. The longer i play, the more i understand all the critique. Maybe i should invest some of my skill points. Refused to do so til yet :)
A good example would be Doom Eternal. Glory kills feel put on to compensate close range combat issues from gamepads. The game works really well and design is very good, but still it feels compromised 'superpower on mashing the right button'. I prefer something like Amid Evil where i hit what i aim at, although it's really retro.
There are some AAA titles i look forward to, like Scorn or Uncharted, but i don't expect much. Mostly the games i really like are smaller indie games coming out of nothing.
 
I love watching the world, but shooting is no fun. Enemies are bullet sponges and barely move. It's more of an interruption between the cutscenes to me. The longer i play, the more i understand all the critique. Maybe i should invest some of my skill points. Refused to do so til yet :)
A good example would be Doom Eternal. Glory kills feel put on to compensate close range combat issues from gamepads. The game works really well and design is very good, but still it feels compromised 'superpower on mashing the right button'. I prefer something like Amid Evil where i hit what i aim at, although it's really retro.
There are some AAA titles i look forward to, like Scorn or Uncharted, but i don't expect much. Mostly the games i really like are smaller indie games coming out of nothing.

I you think CP2077 is a FPS, you should look into the tabletop version..just saying.
The game follows it own rules quite well.

I could reverse the argument with that if you think any FPS has anything to do with "real" shooting...you have gone of the deep end.
 
You mean 40 WGPs?

Edit: It doesn't matter, I might just prefer the new identifier because it's a strong remainder that it's not really the same unit here (as in metric).
Yes, if anything, they should be more efficient by being able to pool certain ressources from two old-style Vega NCUs. Except maybe if you've got tons and tons of independent 16-wide workgroups which for some reason cannot be combined in half as much 32-wide ones.
 
If you're able to force ray tracing shaders to run at wave32 instead of wave64 you can get better performance out of AMD HW like they did in Ratchet & Clank Rift Apart ...


AFAIK there's currently no way to explicitly control the wave size for ray tracing in any of the public APIs. On D3D12 with SM 6.6 you can only specify the wave size on compute shaders. On Vulkan there's EXT_subgroup_size_control but it doesn't do anything on AMD drivers ...
 
Except maybe if you've got tons and tons of independent 16-wide workgroups which for some reason cannot be combined in half as much 32-wide ones.
You would never use 16-wide WGs in practice because half of threads would remain idle.
If i understood this correctly, RDNA can fill its SIMDs faster with new work, which matters after switching shaders, so also if shaders are short, and if lots of stuff is to schedule (e.g. async compute).
Maybe it also matters after switching WGs because of a wait on VRAM access, texture fetch, RT, etc. Would be a huge win then.
On GCN this takes longer because the pipelining of 64 threads over 16-wide SIMDs. RDNA with 32 threads and 32-wide SIMDs can fill those SIMDs twice as fast.
A downside is higher latency. Many instructions take 4 cycles on GCN, but 5 on RDNA.
But i'm really not sure about any of this!

I you think CP2077 is a FPS, you should look into the tabletop version..just saying.
That's another topic. Indeed i don't like tabletop mechanics in realtime video games, because it means GUI interface clutter not conforming the reality simulation going on beside that. So i have ignored the RPG genre as a whole and expect some issues on my side with playing CP.
But obviously it implements all FPS mechanics: Free roam of 3D world, and shooting at enemies which shoot back. And even ignoring the RPG mechanics, i can judge this implementation is bad and way below to what other modern (console) games achieve.
To address aiming does not work so well using gamepads, there are many options:
Aim Assist. See a style shooter like Deathloop does not even have a crosshair. Just like Doom had no crosshair because it had no mouselook yet.
Cover shooter. Solving the problem by alternating movement and shooting.
Melee mechanics. Button mash to attack close enemies, not requiring aim.
Special abilities. See Dishonored, Control, etc. Usually married with the style shooter genre.

But all of those solutions have issues, which are mostly annoying to me:
Aim Assist: No problem with that, but game ends up too shallow and simple, so to add some challenge, we add some of the other mechanics as well.
Cover Shooter: Disables free roam. Constatly i get sucked into some cover, then i'm trapped into that cover and try to break free, ending up doing movements i did not intend.
Melee: That's just too dumb to be fun, mostly.
Special abilities: I have to learn them, remember buttons, and when i come back to the game weeks later i have it all forgotten. I hate it - it's a bolt on cancer. Remembering Controls levitation mode where you're intended to control flying up and flying down with just one button gives me the creeps. I've fallen to death more often than not.

So, in short, it's all a form of complexity bloat. I want simple controls, and complexity should manifest in the games world and challenges, not in GUI pages or by mastering combos of a dozen of buttons.
If Doom Eternal had a 'Classic' mode, without glory kills and stuff, just simple but challenging shooting, i would replay this game 5 times. As is, i won't, although i'd actually like to. Not willing to relearn its mechanics again.

I could reverse the argument with that if you think any FPS has anything to do with "real" shooting...you have gone of the deep end.

It's not about shooting. The best genre of recent years to me was indie horror, as invented from Frictional Games Penumbra or Amnesia. Those games have no shooting at all, but free roam and interaction with a properly simulated and interactive world, and this gives great immersion, which is what i'm after.
So actually i think to evolve games respecting all platforms, i would sacrifice that 'shooting' all together, or tone it down. But if we do so, it turns out too quickly game worlds are just smoke and mirrors and not interesting enough on its own. Thus my believe in pushing physics to lift the technical base here.
 
Back
Top