AMD's FidelityFX Super Resolution reportedly Doubles Game Performance - A DLSS killer

[WYP]

Radeon's multi-platform performance booster!



Read more about AMD's FidelityFX Super Resolution feature reportedly doubling game performance.

 

[Dicehunter]

DLSS killer ? That's an enormous claim, It'll be a welcome alternative but I'll take this with a truck load of salt.

 

[Dark NighT]

Bit sensational i agree, when the cold hard numbers of this come in then we can see if it really does what it says and if it's a "killer" or not.

 

[WYP]

DLSS killer ? That's an enormous claim, It'll be a welcome alternative but I'll take this with a truck load of salt.

If it works as intended and is fully cross-platform (including Nvidia), it will diminish the appeal of DLSS.

If AMD can match Nvidia in terms of quality, Nvidia will have a hard time keeping developers interested.

 

[Dicehunter]

If it works as intended and is fully cross-platform (including Nvidia), it will diminish the appeal of DLSS.

If AMD can match Nvidia in terms of quality, Nvidia will have a hard time keeping developers interested.

It'll be interesting to see how it's implemented as Nvidia use dedicated hardware for it but from what I have read AMD do not have dedicated hardware for this type of thing which going by what I know of RDNA2 it will use a cards compute ability meaning it wouldn't give the same level of performance uplift as Nvidia's Tensor cores.

Could be totally wrong though and AMD have some secret hidden special sauce hardware.

 

[ImprovizoR]

At this point, who the hell cares. Unless crypto market crashes, I see no point in even following the GPU market.

 

[AngryGoldfish]

I think it's fair to say it could be a DLSS killer in that it could offer greater performance uplift and reach a wider audience. DLSS may still end up being visually superior, especially as it ages, but the main issue with DLSS at the moment is implementation. Even if AMD are late, if they come out with something at least somewhat visually comparable, with greater performance gains, and wider implementation, that could 'kill' DLSS in the same way that Freesync 'killed' Gsync. It obviously won't just as Freesync didn't actually kill Gsync, because the technology is very good and Nvidia has its foot in many companies' doors, but Freesync as it currently is is effectively superior. It's cheaper, easier to implement, and works across multiple systems. If AMD had the financial clout and foothold as Intel, Gsync would have no place.

It seems to me that DLSS vs FidelityFX Super Resolution is a repeat of that. DLSS isn't used for anything other than gaming, it can't be used by any other manufacturer, it has to be implemented by the game developers but isn't available on the two main consoles, and it has costly hardware demands. Those are some pretty big drawbacks.

DLSS is awesome and it's in Nvidia's future for sure, but I do think AMD's answer could 'kill' it.

 

[Dicehunter]

I think it's fair to say it could be a DLSS killer in that it could offer greater performance uplift and reach a wider audience. DLSS may still end up being visually superior, especially as it ages, but the main issue with DLSS at the moment is implementation. Even if AMD are late, if they come out with something at least somewhat visually comparable, with greater performance gains, and wider implementation, that could 'kill' DLSS in the same way that Freesync 'killed' Gsync. It obviously won't just as Freesync didn't actually kill Gsync, because the technology is very good and Nvidia has its foot in many companies' doors, but Freesync as it currently is is effectively superior. It's cheaper, easier to implement, and works across multiple systems. If AMD had the financial clout and foothold as Intel, Gsync would have no place.

It seems to me that DLSS vs FidelityFX Super Resolution is a repeat of that. DLSS isn't used for anything other than gaming, it can't be used by any other manufacturer, it has to be implemented by the game developers but isn't available on the two main consoles, and it has costly hardware demands. Those are some pretty big drawbacks.

DLSS is awesome and it's in Nvidia's future for sure, but I do think AMD's answer could 'kill' it.

From the bits I've read, AMD's version will be heavily based on Microsofts Direct ML image upscaling so for games on PC this should be incredibly easy to implement, Compared to DLSS.

 

[NeverBackDown]

DLSS code is provided as a toolkit. It's not like they each do it from scratch.

Remains to be seen for FidelityFX SR. Wouldn't surprise me but this does create issues still with segmentation. Even if they adopt it, if based on DXML it will not be cross compatible with PS. Which then begs the question of if we already support DLSS, why bother adding this for one platform(XBOX)? It wouldn't be worth the time to implement and remove DLSS(from PC) to add it to both Xbox and PC considering most studios only just have gotten DLSS.

 

[looz]

Source: YouTuber says so

 

[AngryGoldfish]

DLSS code is provided as a toolkit. It's not like they each do it from scratch.

Remains to be seen for FidelityFX SR. Wouldn't surprise me but this does create issues still with segmentation. Even if they adopt it, if based on DXML it will not be cross compatible with PS. Which then begs the question of if we already support DLSS, why bother adding this for one platform(XBOX)? It wouldn't be worth the time to implement and remove DLSS(from PC) to add it to both Xbox and PC considering most studios only just have gotten DLSS.

Not sure if I'm right here, but according to the video by RGT, Super Resolution is not the same as DirectML. It could be based on it like Vulkan is based on Mantle, but if I'm understanding RTG rightly, it's not the same thing.

Source: YouTuber says so

In fairness, that dude's channel has a solid track record for accurate leaks.

 

[NeverBackDown]

Not sure if I'm right here, but according to the video by RGT, Super Resolution is not the same as DirectML. It could be based on it like Vulkan is based on Mantle, but if I'm understanding RTG rightly, it's not the same thing

You use ML to get the results you want. How you apply it is up to AMD and their optimizations they want to focus on. If they are basing their predictions off of DirectML then it effectively is the same thing. They are just using it in a specific way that others may not need. DirectML is pretty big and does more than just upscaling tech.