Carzy Zarx’s PC Gaming Emporium - Catch Up on All the Latest PC Gaming Related News

[Jizz_Beard_thePirate]

Bofferbrauer2 said:

Jizz_Beard_thePirate said:

The thing is that if their only choice is to develop an open standard, it needs to be competitive against Nvidia instead of being much worse. Intel has made a pretty worthy competitor to DLSS with their XeSS in their first attempt. It may not be as good as DLSS but it is still better than FSR and it works with Radeon GPUs. I highly doubt Intels GPU team is anywhere as big as Nvidia or Radeon so to me it doesn't make any sense as to why Radeon doesn't have their own Ai based upscaling solution.

I think it's simply a matter of priorities. Radeon would rather have a check mark than spend the time to fully develop a product. Instead of being like, here is frame generation, if they instead came out with ML based upscaling for all their GPUs, it would have been much better imo. Because if AMD comes out with an actual Ai upscaler that is competitive against DLSS that is open source and works with all GPUs and consoles, at that point they can really make the argument... Okay, the need for DLSS might actually not be there. But these days it's like, we have FSR that's worse than DLSS in a lot of different areas so people just buy Nvidia because reviewers tell them DLSS is better and game devs are increasing releasing badly optimized games that rely on upscaling.

So if they want to play the "Nvidia features but open standard" game, they need to fully bake their products even if they release it later. Otherwise they need to innovate. They need to do something that's not the current trajectory because it's not working.

NVidia already had the hardware for ML in their products, AMD didn't. They couldn't include ML without having something to gather the data for ML first, so they had to go another route or wait until about right now before they could release any product. Now imagine the sales of AMD if they didn't have *anything* to counter DLSS for 4 years straight and then only add the current additions several years later down the road. AMD could just as well quit the GPU business entirely if they don't have anything.

Case in point, frame generation. How many times have testers, critics and other people (including you) criticized AMD for not having their own frame generation tech. Now imaging if they would come and say today "we'll have it ready for release by 2026". That's suicide for the company.

Long story short, what you're proposing is simply not realistic or feasible. AMD needed something to counter, to show they care. It's not nearly as good, but at least it's something and not nothing.

I am not saying don't have an upscaling solution until now. If they needed to have FSR1/2 as a stepping stone, fair enough. But it's almost 2024 and they still do not have ML based upscaling even though Intel managed it on their first try. That is utterly insane that a first time dGPU vendor managed an Ai upscaling solution while Radeon still doesn't have one. And yea the Radeon technically does not have something like dedicated tensor cores until RDNA 3 but Intel showed with XeSS that you don't need dedicated Ai cores to make ML upscaling solution work. XeSS works on RDNA 2/1 because it uses dp4a. It's not like XeSS works any better on Nvidia than it does on Radeon. XeSS just works better on Intel only.

And yea many people wanted AMD to have FG and now that it's out, look at what they are saying. "It's not ready, it's not as good as Nvidia's solution, needs more time to get the kinks worked out, AMD released it too early" etc. Just by having a checkmark doesn't mean it's a good thing because now it's basically free advertising for Nvidia. As more and more comparisons and reviews come out, people are looking at how superior DLSS is and all those FSR reviews are giving people reasons to avoid AMD.

Where as if AMD worked on ML solution instead like XeSS and it turned out to be similar to DLSS that worked on every modern GPU like XeSS does. Critics and such will be singing it's praises and declaring DLSS to be dead.

At the end of the day, I am not saying there is an easy solution for Radeon but they need to change their strategy. Because if Intel comes out with Frame Generation with Battlemage early next year and the reviews show DLSS 3 as #1, XeSS + FG as #2, FSR 3 as #3. It will not be a good outlook for the red team.

Last edited by Jizz_Beard_thePirate - on 06 October 2023