What is AMD's Radeon™ Rays software?

[BootyDustBandit]

I understand what it is and what they are trying to get it to do, but why? and how well does it even perform?

[MedievalMatt]

2 minutes ago, BootyDustBandit said:

I understand what it is and what they are trying to get it to do, but why? and how well does it even perform?

They need some way to compete with nVidia's Ray tracing gear.

 

though given the state of ray tracing performance on the 20 series.  Any AMD card will melt itself trying to ray trace anything in realtime.

[BootyDustBandit]

3 minutes ago, MedievalMatt said:

They need some way to compete with nVidia's Ray tracing gear.

 

though given the state of ray tracing performance on the 20 series.  Any AMD card will melt itself trying to ray trace anything in realtime.

how would an RX Vega 64-LC handle it?

[MedievalMatt]

1 minute ago, BootyDustBandit said:

how would an RX Vega 64-LC handle it?

I can only go based on nVidia's performance.

 

but a 2080ti at 4K in Battlefield 5 normally does 150+ FPS IIRC.  With Ray tracing on that drops to 35(ish) FPS.

 

Thats a massive kick in the balls.  and considering that a Vega 64 is like a 1070ti, and the 1080ti and 2080ti are similar in performance (ray tracing aside) im going to go ahead and say it would shit itself.

[Zando_]

2 minutes ago, MedievalMatt said:

I can only go based on nVidia's performance.

 

but a 2080ti at 4K in Battlefield 5 normally does 150+ FPS IIRC.  With Ray tracing on that drops to 35(ish) FPS.

 

Thats a massive kick in the balls.  and considering that a Vega 64 is like a 1070ti, and the 1080ti and 2080ti are similar in performance (ray tracing aside) im going to go ahead and say it would shit itself.

Also, Vega cards don't have the RT cores that Nvidia put on their GPUs, specifically designed to handle Ray-Tracing, so they'd do even worse. 

[MedievalMatt]

5 minutes ago, Zando Bob said:

Also, Vega cards don't have the RT cores that Nvidia put on their GPUs, specifically designed to handle Ray-Tracing, so they'd do even worse. 

With the advent of the AI Anti-Aliasing.  I wonder how long itll be before your GPU gradually learns the games you play and self optimizes to give you greater performance.  

 

Probably will lead to a massive database of datapoints on your PC somewhere.  but if it means i can get more out of my GPU as i play certain games longer, thatd be neat.

[Franck]

17 hours ago, MedievalMatt said:

With the advent of the AI Anti-Aliasing.  I wonder how long itll be before your GPU gradually learns the games you play and self optimizes to give you greater performance.  

 

Probably will lead to a massive database of datapoints on your PC somewhere.  but if it means i can get more out of my GPU as i play certain games longer, thatd be neat.

You would be surprised how little the data can be for AI machine learning. Depends on the precision you need. Especially for fast pace game where it flicker so fast in the screen that even if it's off you won't have the time to notice it.

[BootyDustBandit]

1 hour ago, Franck said:

You would be surprised how little the data can be for AI machine learning. Depends on the precision you need. Especially for fast pace game where it flicker so fast in the screen that even if it's off you won't have the time to notice it.

this is actually super interesting. I think i'm going to do some more of my own research on AI learning when it comes to graphics.

[Franck]

3 minutes ago, BootyDustBandit said:

this is actually super interesting. I think i'm going to do some more of my own research on AI learning when it comes to graphics.

the nice thing about the learning field it's the same thing that you apply to manufacturing, graphic, music or food recipes. You have many main principle and bunch of smaller derivation base of those main principle then you adapt to your model. Some are as easy as crunching bunch of numbers to find the pattern and get a formula that when used return you the intelligent answer based on the past results. Conceptually very very easy.

[BootyDustBandit]

5 minutes ago, Franck said:

the nice thing about the learning field it's the same thing that you apply to manufacturing, graphic, music or food recipes. You have many main principle and bunch of smaller derivation base of those main principle then you adapt to your model. Some are as easy as crunching bunch of numbers to find the pattern and get a formula that when used return you the intelligent answer based on the past results. Conceptually very very easy.

but..it isn't easy to make a true AI, most "AI" people make have to have a specific data or command given to it by a human to actualy do anything, which is safer but not AI. I would say a true AI would just need a command to "Wake up" from a human and then it should be able to learn whatever it wants all on its own. (keyword; WANTS)