RedGamingTech AMD Super Resolution DOUBLES Performance - FidelityFX Leak

This is and a few other youtubers are full on AMD fanboys. Hell it's even in his channel name.
It wouldn't be so bad if they didn't constantly bad mouth nVidia as "the ultimate evil" like many of them do.
 
We will see if this turns true.

Don't have high hopes for RDNA1 though, as its not capable of INT8 instruction set, so it has to rely on FP16 which will slow it down massively, that in turn means massive sacrifices in terms of image quality.

Just focus on RDNA2. RDNA1 at this point without Raytracing and DX12 Ultimate is electronic waste which is holding back game development and progress.
 
Last edited:
RGT isn't what I would consider a completely reliable source, honestly.

Also, performance is only half of the pie. We need to know what it looks like. DLSS 1.0 improved performance but looked like shit. DLSS 2.0 looks great.

We need to see this in action to truly know anything worthwhile.
 
Welp.

Never trust anything from this guy or any other """"""leaker"""""" on the internet.

Some could have some legits sources, but every single one don't resist the attention and always start to mix fake shit in between.

So, you really can't blindly believe anyone. They are all worthless.
 
lol this is the guy who said the ps5 had zen 3 and rdna 3 features lmao.

the guy needs to be banned and needs to start washing his hair with head & shoulders.
 
OMG are some people still confident about rdna3 stuff after this new batch of things thrown to the wall ?
Maybe , maybe 3080 ti one will stick...the others good luck.
 
Last edited:
Some could have some legits sources, but every single one don't resist the attention and always start to mix fake shit in between.

So, you really can't blindly believe anyone. They are all worthless.

I like RGT since he's constantly simply saying "take this with a huge grain of salt". It's fun to speculate on how things could go if rumors end up being true.
 
Personally I don't trust this guy, but I'm willing to be open minded about AMD DLSS, let the benchmarks do the talking in the end.
 
Personally I don't trust this guy, but I'm willing to be open minded about AMD DLSS, let the benchmarks do the talking in the end.

For me, screenshots need to do the talking. FPS benchmarks are nice but I really don't care if this can double the FPS if it destroys the image quality at the same time.

This needs to look as good as DLSS 2.0, at the very least, imo
 
Are educated guesses exclusive now?
Well it is his guess so it is technically exclusive to his thought process. The slim is not exactly that hard to guess at it will be once 5nm is mature enough for an apu of the ps5 size. The sooner the slim is out better for Sony to drop production costs and bring the system price down.
 
lol this is the guy who said the ps5 had zen 3 and rdna 3 features lmao.

the guy needs to be banned and needs to start washing his hair with head & shoulders.

ahah you made me laugh ""needs to start washing his hair with head & shoulders"".

He is not a bad guy, but sometimes he goes a bit too far, to say the least.
 
Watched the video, this guy seems to know his stuff. Lots of words I never heard before and in the background you can see a ton of technical equiqment, which, as an educated guess, could very well be meant for gaming. But I gotta say...not really intrigued by the t-shirt he wears.
 
1440p is 2.2 less pixels than 4k.
It naturally doubles performance.

Now, the tricky part is bribing enough people to add the "and it's better than native!" bovine feces to it.
 
1440p is 2.2 less pixels than 4k.
It naturally doubles performance.

Now, the tricky part is bribing enough people to add the "and it's better than native!" bovine feces to it.

It doesn't have to be better than native.

It needs to be better than whatever resolution gets double the performance. Which DLSS absuletely does, and hopefully AMD can match it.
 
DLSS is software, getting copied was inevitable. the advantage was temporary.

if SR is good we can expect 4K PSVR2 games. probably at 120FPS also.
 
the tricky part is making it still look native res. DLSS2.0 does this very convincingly but other upsampling methods in the past really don't. Checkerboard rendering comes closest to looking native if properly handled like in Dark Souls... the question is, will 2160p created through AMD Super Resolution look better than a well implemented 2160cb? 2160cb also should use about as much GPU power as FidelityFX Super Resolution does, so it has to be better than it or it will be pretty much useless.
 
Last edited:
the tricky part is making it still look native res. DLSS2.0 does this very convincingly but other upsampling methods in the past really don't. Checkerboard rendering comes closest to looking native if properly handled like in Dark Souls... the question is, will 2160p created through AMD Super Resolution look better than a well implemented 2160cb? 2160cb also should use about as much GPU power as FidelityFX Super Resolution does, so it has to be better than it or it will be pretty much useless.

Is it possible to get even more performance by stacking the upscaling? like use super resolution to upscale from 900p to 1440p, then checkerboard to 4K. could be used to get more 120fps games. AFAIK increased framerate helps reduce artifacts in checkerboard image.

another way is perhaps super resolution can be used to more accurately fill in the blank pixels in the checkerboard image.
 
Is it possible to get even more performance by stacking the upscaling? like use super resolution to upscale from 900p to 1440p, then checkerboard to 4K. could be used to get more 120fps games. AFAIK increased framerate helps reduce artifacts in checkerboard image.

another way is perhaps super resolution can be used to more accurately fill in the blank pixels in the checkerboard image.
Yes, but the data loss would hurt the overall quality much more.

The key to a solution like this is going to be balancing quality and performance. Lowering the initial resolution would improve performance at the cost of image quality.
 
Is it possible to get even more performance by stacking the upscaling? like use super resolution to upscale from 900p to 1440p, then checkerboard to 4K. could be used to get more 120fps games. AFAIK increased framerate helps reduce artifacts in checkerboard image.

another way is perhaps super resolution can be used to more accurately fill in the blank pixels in the checkerboard image.

that's not how checkerboarding works. checkerboarding uses 50% of the horizontal resolution, arranged in a checkerboard pattern and then alternates after each rendered frame to fill in the missing information

4K checkerboarded is 1920x2160 per frame

but yeah you could do that but the end result would most likely be not that great
 
Yes, but the data loss would hurt the overall quality much more.

The key to a solution like this is going to be balancing quality and performance. Lowering the initial resolution would improve performance at the cost of image quality.
could be used to help run more demanding games. to avoid issues like CP2077 on last gen. in the future games can take more advantage of a PS5pro then reduce native resolution to fit it on the PS5.
 
he probably knows someone in marketing or does the PR slides or simply stealing off Chinese boards chatter

other than that, another clown living off clickbait guesses on YT, like Mlid. hopefully GAF censor his channel like the old days of spong.com
 
he probably knows someone in marketing or does the PR slides or simply stealing off Chinese boards chatter

other than that, another clown living off clickbait guesses on YT, like Mlid. hopefully GAF censor his channel like the old days of spong.com

he has a good track record so far.
 
could be used to help run more demanding games. to avoid issues like CP2077 on last gen. in the future games can take more advantage of a PS5pro then reduce native resolution to fit it on the PS5.
In that case they'd just start with a lower resolution rather than bumping it up twice.
 
This guy lol, zen 3 in consoles, infinity cache in consoles, rdna2 rt better than nvidia. He's starting to understand at least that good fortune tellers are vague so their predictions can be twisted to fit a narrative. Like this one, double performance but doesn't mention how it will effect image quality. He can turn around and say see I was right even if it's as basic as a resolution slider.
 
making it still look native res.
DLSS2.0 does this very convincingly
Some reviewers seem to lie very convincingly.

We have checked changes to images on this very forum.
There were no issues identifying which of the images was 4k and which was DLSS upscale from 1440p (in totally not confusing way called 4k DLSS quality)
Guess which of them had added blur and loss of details...

This crop is from the site that reviewed DLSS 2 in DS and was "better than native" orgasmic about it. Yet htis is from the pics they've presented to support it:

Was:
JT7ryel.png


DLSS-ed (2.0, the TAA(U) derivative):
CEN8wpL.png


Now, does it NOT do good things to a 1440p render? It does indeed, everything "lines" is notably improved. It is by mile the best TAA(U) derivative we have. But stop pretending it does things it does not.
 
Everyone can get double the performance, it means nothing. It matters how much visual sacrifices have to be made while doing so.

I'm looking forward to the comparisons and hope that this will lead to AMD and Nvidia improving their tech further, so we can get even better results.
 
Some reviewers seem to lie very convincingly.

We have checked changes to images on this very forum.
There were no issues identifying which of the images was 4k and which was DLSS upscale from 1440p (in totally not confusing way called 4k DLSS quality)
Guess which of them had added blur and loss of details...

This crop is from the site that reviewed DLSS 2 in DS and was "better than native" orgasmic about it. Yet htis is from the pics they've presented to support it:

Was:
JT7ryel.png


DLSS-ed (2.0, the TAA(U) derivative):
CEN8wpL.png


Now, does it NOT do good things to a 1440p render? It does indeed, everything "lines" is notably improved. It is by mile the best TAA(U) derivative we have. But stop pretending it does things it does not.

Here you go, someone created a comparison based on math and facts. I don't think it helps you out though.
 
Would be nice to render everything in 1080p and spit out a crisp 4k picture. So many resources could be deployed into rendering the scene.

Imagine how well fast asset streaming would pair with that.
 
Last edited:
Here you go, someone created a comparison based on math and facts. I don't think it helps you out though.
That's interesting, thanks.
Choice of the image is unfortunate and I have doubts about the logic behind using bi-linear for downsampling.

I don't think you got what the comparison in that post was about though.
(I'm referring to Oleg's post)
 
That's interesting, thanks.
Choice of the image is unfortunate and I have doubts about the logic behind using bi-linear for downsampling.

I don't think you got what the comparison in that post was about though.
(I'm referring to Oleg's post)
What was bilinear filtering used for?
 
This sounds great but seems like it'll be done without any dedicated HW like nVidia, so eats GPU resources in order to spare more?

3nm seems to be next gen, but 5nm Pro is pretty easy with a slight RAM boost or probably PCIe 5.0 SSD would be enough upgrade.
 
Top Bottom