• 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.

Rx 6800 good enough?

This is another discussion: do you use part of the chip for dedicated hardware or do you put more compute units for native rendering and use them for upscaling?
Nvidia bet on dedicated hardware but that is also because you can manipulate things with dedicated hardware: sponsor games that will only run decent at 1080p unless upscaled or use a denoiser that is made to only work well with your dedicated hardware in RT games. You can do a lot of things when you have a proprietary solution and a big wallet. Everything will change once Direct ML becomes a standard but for now, Nvidia has a big advantage in gaming and they know how to use it, even against their own clients (those who own a Pascal card or even a weaker Turing card ).

AMD have already said they will dedicate more silicon to machine learning with RDNA3, while Nvidia will run Microsoft's Direct ML on their Tensor cores.
 
AMD have already said they will dedicate more silicon to machine learning with RDNA3, while Nvidia will run Microsoft's Direct ML on their Tensor cores.
Yes because they need to close the gap with Nvidia in upscaling and denoising but that does not mean you can't do these things without dedicated hardware. And again the machine learning thing is done in the servers, the dedicated hardware in the cards ( like the tensor cores ) are used to calculate maths that is called surprisingly tensor. :D
Your card doesn't learn anything.
 
Lots of people on there missing that you do the work on servers. I think theres much confusion from people thinking tensor cores actually upscale when they don't. They are just rendering the end equation. In any case looks like the marketing worked. :) we have ai upscaling on the Nvidia shield for example without tensor cores. Granted video is cheapwr to run the a game, but that's the point isn't it, compare a 3080 to a shield GPU wise.
 
Yes because they need to close the gap with Nvidia in upscaling and denoising but that does not mean you can't do these things without dedicated hardware.

If you don't use dedicated hardware then you have to use general hardware that is used for everything else, everything else then slows down as a result due to the greater workload on that general hardware. I'm sure AMD only took the route they did due to budget restrictions on creating a console chip, RDNA2.

And again the machine learning thing is done in the servers, the dedicated hardware in the cards ( like the tensor cores ) are used to calculate maths that is called surprisingly tensor. :D

Your card doesn't learn anything.

There is a lot of info out there on Deep Learning if you are interested in how it works and how the Tensor cores are used, just Google it. In DLSS for example, the Tensor cores are used to reconstruct the image from training done originally elsewhere. The data from said training is supplied via the driver. There is no reason that you can't do such training on consumer cards.
 
If you don't use dedicated hardware then you have to use general hardware that is used for everything else, everything else then slows down as a result due to the greater workload on that general hardware. I'm sure AMD only took the route they did due to budget restrictions on creating a console chip, RDNA2.
I already said there is another trade off. Take for example a console like X box X. What do you put inside the chip? Do you put more compute units that help rendering in every game or do you put specialized cores that are only used for upscaling/denoising, especially since 99% of games don't need upscaling or denoising and you can do the same maths the specialized cores are doing, on shader cores with some performance impact?
Once DirectML becomes a standard and it will be implemented in future games, the Xbox X will be able to run any game at upscaled 4k-60FPS without any specialized cores. So that part of the chip is almost useless and it can't do anything else.

Dedicated hardware is a thing for Nvidia because they can use it right now to get a huge advantage over AMD and to force older Nvidia cards to become obsolete. But for such tactics, you need a big team and a big wallet, to pay the developers to give you access to their code and to put your code inside their games.

Because on PC it matters if you get 80 FPS instead of 100, even if both results are obtained at a fake resolution, AMD will probably use dedicated hardware in their next generation knowing or hoping that directML becomes a standard and that way they can compete with Nvidia. Without a standard they won't be able to compete as they don't have the manpower or the money to support their own proprietary technology.
I am sure Nvidia is not happy either because they have to spend a lot of resources for every game that gets RT and/or DLSS and that is why is trying to push the narrative and hope the developers will use their own money to add RT and/or DLSS support in their games. But at least Nvidia is using their proprietary tech as a marketing tool and they are doing a good job at it so these are money well spend. And i am sure that once they lose the advantage in upscaling/denoising, they will focus on something else and promote that thing.

What i want to say is that none of these companies are stupid. There are reasons for why Nvidia has dedicated hardware (marketing) and there are reasons for why AMD does not have it atm(lack of resources and no standard). There is a reason for why Microsoft and Sony don't use it in consoles either. They simply won't need it for their future upscaling plans.
 
AMD have already said they will dedicate more silicon to machine learning with RDNA3, while Nvidia will run Microsoft's Direct ML on their Tensor cores.
No AMD didn't, say that explicitly. that is your interpretation of a vague quote that you have been pedalling all over the forums.
 
Last edited:
Back
Top Bottom