Valve's Dota 2 Adds AMD FidelityFX Super Resolution

[900k]

Yes, but keep in mind that 3D rendering on screen is badly undersampled by definition, unlike video. You can't fix this by hallucinating stuff, you need samples. And the only samples that are both not expensive and of which there is a plentiful reservior are temporal ones. Unlike oiaohm thinks, the key element is reducing artifacting when using them, and not resort to pointless solutions like FSR 1.0 that try to cope without them. Probably even AMD knows that their attempt has no future (I'd say it even has no present), but they couldn't do any better until now.

I dont disagree, I think that the value of fsr comes in the "it doesn't look nasty" factor, resolution scaling isn't a new feature for games, but it almost always looks terrible, so its nice to see something that is a little cleaner .

[cb88]

Bad news Nvidia has taken out 50+ patents and one of them that got approved is exactly what you just wrote. Patent office is really useless at checking for the too obvious issue. Great fun if this end up with a patent troll.

IBM's patent rate is vastly higher than Nvidia's measley 7338 patents... IBM has 150,000 patents. So... yeah they may have patented alot of ideas, but they aren't generating impregnable patent thickets.

[tildearrow]

You mean paid. AAA developing these days is complex and expensive enough without experimental proprietary features. Nvidia is famous for sending their minions to implement their stuff in selected, high profile titles.

Yes, I exactly mean that.

[Quackdoc]

FSR is amazing, it's not the absolute best quality, but it far exceeds the acceptable line in terms of quality and performance. I can see this essentially free performance boost being widely implemented.

[bple2137]

I tested that out of curiosity and it works pretty well. I run that game on Linux on 4K TV and at 90% I can barely see any difference while there's clearly an improvement when it comes to FPS. At 70% it start to look not so good, but I can only tell because I'm sitting close to the screen. Looking forward to see some games having this feature that I would actually play.

[leipero]

Bad news Nvidia has taken out 50+ patents and one of them that got approved is exactly what you just wrote. Patent office is really useless at checking for the too obvious issue. Great fun if this end up with a patent troll.

Now that's nonsense really, you can't patent "using the previous frame and current frame to generate an upscaled frame" as cb88 explained, the reason is, it's not nvidia who used that method first, it is in use for a long time in image/frame algorithms AFAIK, so really, they can only patent their own specific implementation(s), there's basically infinite methods how you can approach that issue, 50 patents is meaningless. Plus, it really is a patent-trolling in order to gain "competitive advantage" and I'm sure there are laws against that as well, so if one find it worth to bother with it, one could potentially win and remove those patents.

On the topic of FSR, I'm not impressed from what I saw, but it's OK-ish. The core issue here is that it doesn't really solve the problem of "bad/lazy developer", you can have a relatively good game with not so good graphic implementation, and you are basically at square one again, because if FSR/whatever is not implemented by developer, you have no reasonable option.

[cb88]

On the topic of FSR, I'm not impressed from what I saw, but it's OK-ish. The core issue here is that it doesn't really solve the problem of "bad/lazy developer", you can have a relatively good game with not so good graphic implementation, and you are basically at square one again, because if FSR/whatever is not implemented by developer, you have no reasonable option.

I tend to agree, that said FSR at 90% resolution seems worth turning on as it makes thing noticeably sharper even above native.

[jaxa]

Sadly, lack of temporal component means it has no chance of competing with DLSS and will always look like shit quality-wise.

Not sure what they were thinking. Many games feature custom temporal antialiasing solutions that look better than this.

It looks better than DLSS 1.0 did. AMD already has improvements to FSR working internally. This is just the beginning.

[oiaohm]

IBM's patent rate is vastly higher than Nvidia's measley 7338 patents... IBM has 150,000 patents. So... yeah they may have patented alot of ideas, but they aren't generating impregnable patent thickets.

This is not understanding IBM. Yes IBM patent rate look high but that is only tip of iceberg of what IBM has done.

https://en.wikipedia.org/wiki/Cloem

When you are making a patent thicket you are not always using humans to generate the patent applications or defensive publications. For every individual ibm patent IBM does at least 10 defensive publications. IBM these days still uses the same patent thicket tech they developed decades ago but instead of spending the money on patents for the possible generated invalid they use defensive publications to restrict what other parties could invent/patent around what they have already patented.

Now that's nonsense really, you can't patent "using the previous frame and current frame to generate an upscaled frame" as cb88 explained, the reason is, it's not nvidia who used that method first, it is in use for a long time in image/frame algorithms AFAIK, so really, they can only patent their own specific implementation(s), there's basically infinite methods how you can approach that issue, 50 patents is meaningless. Plus, it really is a patent-trolling in order to gain "competitive advantage" and I'm sure there are laws against that as well, so if one find it worth to bother with it, one could potentially win and remove those patents..

This miss understand patent law look a Cloem link. Patent law does not mandate you patent something you have invented at all this is how computer generated patent applications covering basically everything that is not prior art is legal. Big problem in the current patent system is lack of requirement to have a presentable implementation to take out patent. Like it or not lot of parties party stuff that is not their own specific implementation because its legal to-do so. Yes I do agree this mass of patent applications by Nvidia over what should be just a handful has the clear markings of a company planing patent trolling for competitive advantage and really should be forbid.

On the topic of FSR, I'm not impressed from what I saw, but it's OK-ish. The core issue here is that it doesn't really solve the problem of "bad/lazy developer", you can have a relatively good game with not so good graphic implementation, and you are basically at square one again, because if FSR/whatever is not implemented by developer, you have no reasonable option.

We don't know yet on this 100 percent yet. FSR being open source nothing says that some FSR method cannot be used by gamescope or in wine work for wayland scaler. Basically it still early days for FSR. Yes the wine work for wayland adding a scaler does add in the possibility of adding something like FSR without game developer involvement. Yes the fact that DLSS has not been on Linux until recent has also mean this third party ways have not been on the table either.

Of course FSR cannot make a game that looks garbage in native implementation at native support resultion look good. But a relatively good game with a relatively good engine making a relatively good render FSR in time may help with as the developers behind valve gamescope and wine work for wayland to have time to look over how usable FSR is to them.

[flashmozzg]

I don't get the TAA fad. It looks good on static screenshot but in practice I'd prefer even blurry FXAA over that crap. It's a shame game don't include at least SMAA by default.