GPU Ray Tracing Performance Comparisons [2021-2022]

[davis.anthony]

4 hardware threads directly from the paper it is good better than single threaded It means 4 CPU cores. This is great and it stay 4 cores for the CPU to do other stuff on an 8 cores CPU. Maybe it it is possible to use a little bit of CPU time using SMT on the 4 cores doing BVH construction. 4 cores are multiples cores.

I love how you think you've won by bolding a comment of mine that doesn't mention RT

Funny that in the sentence directly below it I said this.....

Do you not think it's a coincidence that pretty much every game with RT tanks only a few CPU cores with RT enabled?

Oh look, I talk about RT using more than one core.

Nice try and have a nice day.

 

[chris1515]

I love how you think you've won by bolding a comment of mine that doesn't mention RT

Funny that in the sentence directly below it I said this.....



Oh look, I talk about RT using more than one core.

Nice try and have a nice day.

No it means this is inefficient because the CPU have enough core to do other stuff even if 4 cores are fully occupied by building the BVH but continue to show you don't understand.

 

[davis.anthony]

No it means this is inefficient because the CPU have enough core to do other stuff even if 4 cores are fully occupied by building the BVH but continue to show you don't understand.

You failed and now you're chatting wobble as you always do when you don't get the victory you want.

Good day sir.

 

[Cyan]

I've played around with the config file and can not find any options to reduce the objects in the BVH, the range of the BVH or any options that would bring CPU performance up.

Can you point me in the right direction to these settings?

I can't, I've just glanced at it for other reasons, but there are a few tweaks there that might affect RT. In the end it's all in the code. That being said, this video hints at BVH being done in the GPU.

Also, I run full-fledged Quake II RTX at a stable 60fps -it's like 14-15 fps at native 4K but I use a resolution scale target fps, and the game while not as crisp as native 4K looks okay- which is path traced game.

 

[davis.anthony]

I can't, I've just glanced at it for other reasons, but there are a few tweaks there that might affect RT. In the end it's all in the code. That being said, this video hints at BVH being done in the GPU.

Also, I run full-fledged Quake II RTX at a stable 60fps -it's like 14-15 fps at native 4K but I use a resolution scale target fps, and the game while not as crisp as native 4K looks okay- which is path traced game.

It can be done on the GPU but the set-up of the high level BVH is typically done on the CPU which is why RT tanks CPU performance. And if a developer does it on the CPU it means they don't have to do per GPU architecture optimisations for it.

Once GPU's get the relevant fixed function hardware required to do it all on the GPU then the CPU will be left alone, but the question is how long will that take?

Quake 2 RTX runs good, have you tried the RT mod for the first game? I recommend it if you haven't.

 

[Cyan]

It can be done on the GPU but the set-up of the high level BVH is typically done on the CPU which is why RT tanks CPU performance. And if a developer does it on the CPU it means they don't have to do per GPU architecture optimisations for it.

Once GPU's get the relevant fixed function hardware required to do it all on the GPU then the CPU will be left alone, but the question is how long will that take?

Quake 2 RTX runs good, have you tried the RT mod for the first game? I recommend it if you haven't.

tried that mod yes, but I can't run it as is. There is some set up that must be done. I gotta go but there are two articles that I find interesting. I haven't read the second one except for a few lines, but in the first one they say this: "Our preliminary results show that current GPU architectures can compete with CPU implementations of hierarchy construction running on multicore systems. In practice, we can construct hierarchies of models with up to several million triangles and use them for fast ray tracing or other applications.". -the article is from 2009-

Fast BVH Construction on GPUs | Research

We present two novel parallel algorithms for rapidly constructing bounding volume hierarchies on manycore GPUs. The first uses a linear ordering derived from spatial Morton codes to build hierarchies extremely quickly and with high parallel scalability. The second is a top-down approach that...

research.nvidia.com

Also this one, from June 2022.

How to build a BVH – part 9a: To the GPU – Jacco’s Blog

jacco.ompf2.com

 

[pjbliverpool]

Man...I can't imagine being that into fps, 60fps sure is appreciated though this gen. I feel like if I ever got a PC I'd be absolutely paralyzed by choice and spend more time adjusting settings than playing games.

What you talking about dude, adjusting settings and testing performance is what makes PC gaming so much more fun than console gaming

On the 3700x argument - I'm probably biased because I have one, but I'd class it as still a very capable gaming CPU. More capable than the console CPU's if dealing with the same workload (which won't always be the case) but still a way off the really high end on PC.

Is it better than a 12400 for gaming? In general no, because games are still single thread heavy.... but I'd argue it has more potential gaming performance for a very well threaded engine due to having double the cores but also more than half the per core performance of the 12400. Problem is that theoretical advantage likely won't materialise in the majority of cases.

 

[Inuhanyou]

Question about CPU parallelism to the experts. Didn't most devs already figure out paralell coding for the last gen? I was under the impression console CPUs were so weak that you were forced to parallelize code to get results out of them.

So why do so many engines still seem to rely so heavily on IPC?

Or am I getting multi core parallelism confused with multithreading? I assumed they were similar except within each core you could effectively have two mini cores doing seperate tasks?

Am I totally off base with this? Help a total casual out!

 

[DavidGraham]

So why do so many engines still seem to rely so heavily on IPC?

Many gaming workloads run on the CPU are serial in nature, and are difficult to extract parallelism from, especially if you are a developer limited by time and resources, and don't have the luxury to rewrite major parts of your game/engjne from scratch.

Also many gaming workloads are latency sensitive, and parallelism incurs a certain latency, so developers run these workloads serially to acheive the lowest latency possible.

 

[Inuhanyou]

Many gaming workloads run on the CPU are serial in nature, and are difficult to extract parallelism from, especially if you are a developer limited by time and resources, and don't have the luxury to rewrite major parts of your game/engjne from scratch.

Also many gaming workloads are latency sensitive, and parallelism incurs a certain latency, so developers run these workloads serially to acheive the lowest latency possible.

I see. So Gotham knights example that Alex showed as having bad CPU utilization may not be something that could have even been helped...

 

[Rootax]

I see. So Gotham knights example that Alex showed as having bad CPU utilization may not be something that could have even been helped...

Some games are doing better. Even the previous batman game. So either the engine is not very good, or the devs using it aren't. Probably both ?

 

[Inuhanyou]

Some games are doing better. Even the previous batman game. So either the engine is not very good, or the devs using it aren't. Probably both ?

Well unreal engine is pretty the same in both cases

 

[Rootax]

Well unreal engine is pretty the same in both cases

Well yeah but UE is a big engine/framework/tool. What you do with it can define the performance level a lot. So maybe UE was not the right choice for them and their skills. We would need some insiders to know what went wrong.

 

[PSman1700]

Again young padawan, I see you joined in June 2021

What has joining date to do with expertise and knowledge? I sure hope thats not the route were going to take here. Terrible choice to counter someone in a technical discussion.

Man...I can't imagine being that into fps, 60fps sure is appreciated though this gen. I feel like if I ever got a PC I'd be absolutely paralyzed by choice and spend more time adjusting settings than playing games.

Its all optional, most casual pc gamers dont tinker that much with game settings.

 

[chris1515]

What has joining date to do with expertise and knowledge? I sure hope thats not the route were going to take here. Terrible choice to counter someone in a technical discussion.



Its all optional, most casual pc gamers dont tinker that much with game settings.

No but I never see so much nonsense here and there for someone knowing about what he is talking or at least searching about the subject and now he has change of camp and gone from a PS5 fan to a PC fan and what he is saying doesn't make more sense before too.

I never see this level of knowledge from a long time member.

 

[PSman1700]

No but I never see so much nonsense here and there for someone knowing about what he is talking or at least searching about the subject and now he has change of camp and gone from a PS5 fan to a PC fan and what he is saying doesn't make more sense before too.

I never see this level of knowledge from a long time member.

There is never a reason to attack someone based on registration date. Some users that joined this summer have shown to be very reasonable, reliable and potent users. The RTX Remix topic has spawned many intresting new forum users, probaly 10 to 20's new users just last week. All are not intrested in this pro-PS5 kind of stuff to begin with.
No, theres no reason to deem anyones competence based on registration date. Your registration date is 2005, yet your being on another long-timers ingore list aswell as being called a troll by such long-time member. Your contradicting yourself.

 

[chris1515]

There is never a reason to attack someone based on registration date. Some users that joined this summer have shown to be very reasonable, reliable and potent users. The RTX Remix topic has spawned many intresting new forum users, probaly 10 to 20's new users just last week. All are not intrested in this pro-PS5 kind of stuff to begin with.
No, theres no reason to deem anyones competence based on registration date. Your registration date is 2005, yet your being on another long-timers ingore list aswell as being called a troll by such long-time member. Your contradicting yourself.

I don't care to be make to ignore list by Albuquerque because basically he call a dev a liar and a guy making marketing for a machine because Elan Ruskin told there are loading a full level after each portal at the end it is true.

I would say I am even happy to be on his ignore list. I don't care about this.

 

[PSman1700]

I don't care to be make to ignore list by Albuquerque because basically he call a dev a liar and a guy making marketing for a machine because of this tweet.

I would say I am even happy to be on his ignore list. I don't care about this.

Let him be the one on this forum to be knowledgeable regarding file systems. Go figure.

 

[chris1515]

Let him be the one on this forum to be knowledgeable regarding file systems. Go figure.

Again he was wrong they load the full level inside a portal and this is possible with any machine having a fast I/O like PC with Direct Storage, Xbox Series or PS5. They did like this because it was easier to manage if was not the case they would have job to manage loading only part of the level and so on. I don't care if he worked on I/O management because he is a PC fanboy which deter his judgment.

EDIT:
You know I like precise information. I will not said some name but when a game load in 1.5 seconds on PS5 and 8.5 seconds on Xbox Series X, some people I won't give pseudo were saying this is the real difference in I/O between the two machines and this is false too. Searching the tweet I wanted to find I saw some craziness here on all side but mostly consoles.

 

[Cyan]

What you talking about dude, adjusting settings and testing performance is what makes PC gaming so much more fun than console gaming

On the 3700x argument - I'm probably biased because I have one, but I'd class it as still a very capable gaming CPU. More capable than the console CPU's if dealing with the same workload (which won't always be the case) but still a way off the really high end on PC.

Is it better than a 12400 for gaming? In general no, because games are still single thread heavy.... but I'd argue it has more potential gaming performance for a very well threaded engine due to having double the cores but also more than half the per core performance of the 12400. Problem is that theoretical advantage likely won't materialise in the majority of cases.

do you use stock settings? I decided to undervolt the 3700X to 1.0875V and lock it to 4GHz. I did this to avoid the huge and sudden temp spikes and crazy fan spinning. I loved the first Ryzen CPUs and the Zen 2 generation, but both Ryzen had the same issue. Nowadays, alas, I wouldn't recommend a Ryzen to anyone. It's just that Intel has better gaming -and overall- CPUs, imho. I am still happy with the 3700X though, it was a very nice CPU at the time and still kicking.