Testing AMD's FidelityFX Super Resolution (FSR): Image Analysis and Preliminary Performance

Status
Not open for further replies.

Watzupken

Posts: 315   +303
This is very encouraging result and seem to be consistent with some other reviews. If this indeed works well with minimal visual degrade, I think AMD have a winner. But for now, I would prefer to see if AAA titles look as good because there are too little games to conclude this. Especially some of the game titles don't seem to have very high quality visuals to begin with. But seeing how much performance you can claw back even with the ultra quality settings, I think I am happy to run it at this quality settings and get a good 20 to 30+ percentage performance back. Thanks for the great and quick turnaround review.
 

Dimitriid

Posts: 953   +1,763
Even though the situation looks rather bland in terms of game support for FSR, the fact that Microsoft and Sony desperately need this on their consoles to kind of, sort of deliver on their promises of 4k visuals or (Well they said "and" but we know that's not happening) high refresh rates.

This means that eventually, new console titles that launch with FSR should have a relatively straight forward port to PC with FSR included.
 

Irata

Posts: 1,673   +2,805
I‘m actually very positively surprised by the results. Had expected OK quality (I.e. not great but good enough), so seeing that ultra quality is very close to native while providing a noticeable fps boost ist great, particularly for owners of older GPU.

Two things I‘d really like to see are iGPU results, as well as a quality comparison between FSR at the target resolution vs. no FSR at the resolution and quality settings that would give me the same performance. Imho, this is much more relevant because for cards that don‘t quite cut it, the alternative often is not 1440p with FSR on or off but rather 1440p with FSR vs 1080p without.
 

Puiu

Posts: 4,864   +3,749
TechSpot Elite
I'm really liking the results for Ultra Quality on 4K and 1440p. I wasn't expecting much more from FSR than some sort of improved CAS. it's a good start, but they really need the big AAA games to come now. (and esports titles too)

If devs or AMD can manage to make proper tweaks on a per game basis and improve the results then it should be just free performance for all GPUs.
 

George Keech

Posts: 54   +66
Love the article.

However This might be a bad take, but I find it soo funny that DLSS and FSR are so popular and being touted when I remember how much crap the PS4 Pro and Xbonex got for only upscaling to 4K by PC gamers and now that's what we want. I understand there is a difference due to FPS differences and the type of upscaling but there is a bit or Irony in there
 

NeoMorpheus

Posts: 883   +1,677
Considering that it took the anti-consumer d!cks at nvidia 3 tries with dlss (v1, v2, v2-rereleased with the same version number), this is very encouraging from AMD part.

Lets see how this will translate on consoles and the upcoming AMD powered Exynos SOC, since it looks like AMD lost the pc due to superior marketing brainwashing from nvidia and sheep mentality of those in there.
 

quadibloc

Posts: 307   +194
Since DLSS and FSR were both touted as ways to make ray-tracing usable, I would still have liked to see benchmarks on FSR on Godfall with ray-tracing enabled, even if ray-tracing in that game doesn't seem to produce the desired benefits in visual quality.
After all, the benchmarks we've seen shows that the higher modes of FSR provide more performance benefits on less-powerful video cards, so making the game as demanding on the video card as possible should show it providing performance benefits.
 

Shadowboxer

Posts: 1,717   +1,322
This website is so disgustingly biased. Techspot hurled negative article after negative article at DLSS when it launched, because among other reasons it didn’t have many supported games for example (how dare devs not make games for techs that don’t exist yet). They concluded DLSS to be a fail three months into its life and then proceeded to ignore DLSS 2.0 when it released and I believe this may be the first mention of it.

Oh and to all the fanboys praising AMD for making FSR work on Nvidia. If they didn’t then devs wouldn’t bother putting it in their games as AMD have a pitiful GPU market share due to the fact that most people choose to buy Nvidia solutions.

 

Kosmoz

Posts: 382   +687
I'll probably be against the popular belief, but I have no issues with FSR at 1080p Ultra.
It's really good enough for me for extra fps, if I need those fps, of course.

My GTX 1080 is happy with how FSR 1.0 looks and works :)

Also I have no worries at all that FSR adoption will explode with both consoles and Intel joining the party.
In 1 year time FSR could be in more titles than DLSS, it really could be...

Thanks AMD!

*now only if the prices would drop even harder so I can buy a new AMD GPU too.
 

Dimitriid

Posts: 953   +1,763
I‘m actually very positively surprised by the results. Had expected OK quality (I.e. not great but good enough), so seeing that ultra quality is very close to native while providing a noticeable fps boost ist great, particularly for owners of older GPU.

Two things I‘d really like to see are iGPU results, as well as a quality comparison between FSR at the target resolution vs. no FSR at the resolution and quality settings that would give me the same performance. Imho, this is much more relevant because for cards that don‘t quite cut it, the alternative often is not 1440p with FSR on or off but rather 1440p with FSR vs 1080p without.
Ditto: igpu and apu results could be quite promising: I'd make 1080p viable not just for "esports" and "indies" as with the current paradigm but reasonable visuals and reasonable 1080p performance for even "AAA" games too.

It should also be interesting to see how that ends up affecting lower end chips because I think the performance hit from this algorithm should come from somewhere so if it's not putting any extra stress on the compute cores like DLSS then it might be using raw CPU cycles to do it.

If we just test with even a 6/12 CPU then we wouldn't see it but we could potentially see some more utilization at what I think might be the modest integrated low end cpu offerings at 4/4 and graphics or 4/8 and graphics chips.

If the hit comes to the regular raster workload however well then this probably won't work very well on integrated graphics.
 

McMurdeR

Posts: 330   +327
Love the article.

However This might be a bad take, but I find it soo funny that DLSS and FSR are so popular and being touted when I remember how much crap the PS4 Pro and Xbonex got for only upscaling to 4K by PC gamers and now that's what we want. I understand there is a difference due to FPS differences and the type of upscaling but there is a bit or Irony in there

I think the ire was directed at the 'claim' of 4k coming from a starting position at the time where almost all console games were targeting 1080p30 at best. PC gamers are well positioned to understand the jump in hardware capability required to step up to an intermediate resolution, not to mention 4k.

The first chequerboarding methods, while hardly perfect, actually made a certain amount of sense owing to the fact that console gamers don't sit on top of their displays to the same extent that PC gamers do, and that their displays (usually TVs) are generally native 4k. But no matter what way you dress it, it wasn't as good as true 4k.

What's changed here is that the upscaling methods are genuinely evolving into something both consoles and PCs can use with less loss of perceived quality.
 

RaXelliX

Posts: 18   +21
This website is so disgustingly biased. Techspot hurled negative article after negative article at DLSS when it launched, because among other reasons it didn’t have many supported games for example (how dare devs not make games for techs that don’t exist yet). They concluded DLSS to be a fail three months into its life and then proceeded to ignore DLSS 2.0 when it released and I believe this may be the first mention of it.

Oh and to all the fanboys praising AMD for making FSR work on Nvidia. If they didn’t then devs wouldn’t bother putting it in their games as AMD have a pitiful GPU market share due to the fact that most people choose to buy Nvidia solutions.
Yes. So biased that they praised 2.0 and have continued to do so:

DLSS 1.0 was a fail. Nvidia themselves realised that.
AMD could have easily skipped FSR on Nvidia cards and developers would have still implemented it because of consoles and ease of integration.
 
Status
Not open for further replies.