AMD FSR 3.0 to be Hardware Accelerated on RDNA 3 GPUs!

WCCFTECH Article:

Hello Dear Viewers :D

Today, with a very small video about the new possibility of FSR 3.0 being hardware accelerated on RDNA3 cards!

00:00 - FSR 3.0 on RDNA 3

05:22 - Conclusion
06:24 - Channel Members
06:45 - More Videos

Hope you enjoy the video. Comment with your thoughts about it. Don't forget to hit like, subscribe and share the video if you can.

If you want, support me on:

Social Networks:

AMD PC USED for CPU testing:

MB: MSI B450 Tomahawk MAX II
RAM: 2x8Gb Patriot Viper 4400Mhz CL19 (3200MHz CL14)
GPU: AMD RADEON RX 6800 @2400/2100 MHz
HDD: Seagate Barracuda 2Tb 7200 rpm
SSD: 500Gb Kingston Nvme
PSU: Aerocool X-Strike 800W 80+ Silver Semi Modular
CASE: None

INTEL 10/11th Gen PC USED for CPU testing:

MB: MSI Z490 Tomahawk (Latest Bios)
RAM: 2x8Gb Patriot Viper 4400Mhz CL19 (3200 MHz CL14)
GPU: RX 6800 @2400MHz / 2100MHz
SSD: Kingston V300 120Gb
HDD: Seagate Barracuda 2Tb 7200 rpm
PSU: Seasonic SIII 650W Bronze
CASE: Aerocool One Frost Edition

INTEL 12th Gen PC USED for CPU testing:

MB: Gigabyte Z690 Aorus ELITE (DDR4)
RAM: 2x8Gb Patriot Viper 4400Mhz CL19 (3200 MHz CL14)
GPU: RX 6800 @2400MHz / 2100MHz
SSD: Kingston V300 240Gb
HDD: Seagate Barracuda 2Tb 7200 rpm
PSU: Aerocool X-Strike 800W 80+ Silver Semi Modular
CASE: None

My Main PC (used for GPU testing):

CPU: AMD Ryzen 5 5600X (4.75GHz Curve Optimizer)
CPU Cooler: Noctua NH-D15S
MB: Asus X570-F Strix
RAM: 2x8Gb Patriot Viper 4400Mhz CL19 (3800 MHz, CL16 / Tightened Sub-timings)
GPU: AMD RADEON RX 6800 @2400/2100 MHz
HDD: 1x2TB + 1x4TB
SSD: 240Gb Kingston Nvme + 500Gb Toshiba Nvme
PSU: Corsair RM750X (White) 750W 80+ Gold
CASE: Thermaltake View 51 TG

#AMDFSR #FSR3 #rdna3 #AncientGameplays #RX6800 #RX5700XT #R53600RX5700XT #GPUOverclocking #AMDWattman #AMDRadeon #Ryzen52600 #I510400F #I511400F #I512400F #R52400G #R55600G #RX6600XT #BestGamingPC2021 #GamingPC2021 #Tech #Computer #Cheap #PCGaming #PCBuild #Build #Gaming #Hardware #AMD #Nvidia #Intel #PC #Ram #Budget #Ryzen #BestBudgetPC #BestBudgetPC2021 #BestCheapPC2021 #BestAMDCheapPC #Ryzen55600g #Ryzen32200G #Ryzen52400Ggameplay #R52400G #FX83702021 #FX8370overclock #R53600XT #R52600 #Ryzen55600X #IsRyzen52600good? #Ryzen52600 #Ryzen51600 #Legendasemportuguês

[AncientGameplays]

Well, in my opinion, I think AMD will just keep supporting FSR 2.0 for older cards and consoles, and one of the things that will keep it alive is the fact that it is open source, so modding community can work with it! As for FSR 3.0 it will most likely come some time after all RDNA 3 cards are released.

[robertkincannon]

I'm glad to hear that AMD isn't taking their foot off the pedal. I think this is great news, it shows how much emphasis they've been putting into the software side of things. I think their driver team is much improved and this will become more important as the processes and design between them and the competition approach parody.

[kishaloyb.]

AMD handled this upscaling conundrum better than anyone. instead of leaving their old customers dry and disappointed, they gave them FSR 1/2. Even GTX owners got the taste of it. For new customers, they'll be getting FSR 3.0. As a RX 5700 owner, I have no regrets and my next GPU definitely will also be from AMD. AMD staggered realize of FSR was much better and appreciated than the sudden release of DLSS which left many old Nvidia customers disappointed.

[scribeofsolace]

Let’s get this dude to 100k by the end of the year dammit 😔✌️

[PeterPauls]

I have an nVidia card (RTX 3080 12GB) and DLSS 2.4.3 is out now. It is significantly better than FSR 2.0 and I'm really glad for FSR 3.0 and I HOPE that AMD will use A.I. to accelerate their image reconstruction. FSR 2.0 will remain an open source solution and FSR 3.0 will be a closed A.I. accelerated API... win-win.

[smifffies]

Always make cost effective advances in GPU technology. In future/current games enable FSR 3.0 for RDNA 3 cards as default and drop down to FSR 2.0 for older/competitors non DLSS capable cards.

[ellypsis]

This will become the new standard, just like Freesync, thanks to AMD ;)

[Focaz]

Niceeee…but i will keep my 6900xt for the next 4 to 5 years

[Trunks]

Its very interesting news and how AMD utilizes this could be awesome. BTW would love to see your take on the DLSS to FSR2 mod for Cyberpunk 2077. Allowing the use of FSR 2 on Cyberpunk 2077.

[sportler]

Love the memes in the begining of the videos ty

[DukeAbbaddon]

3D & 2D Audio & Haptic Wavelet Alpha Blend (c)RS

Constitutes a method of pixel free codecs & audio distribution without glitches or distortions;
(due to aliasing); Combining both blending & Dynamic Sharpening of a light sort ..
Both has few issues with performance & also looks better..

*

Furthermore utilizing RGB + A or A:C1 to A:C53 (Alpha) B1:D to B:D53
(Constituting smoothing & then sharpening gradients)
Intending to improve fidelity though alpha blend channel & wave-shaping

*

Codec Observations & thoughts RS : I think that AA combined bytes on this operation are 4 cycles per frame on MUL/+- fraction floats, on an AVX 256Bit array (optimally) Load 2, Run 3 Cycles & 2 Save

*

enable_vp_scaling
R+G+B +-* (Backplane +- A) = ((Result {1, 2, 3, 4}) ADD) DIV4 + Remainder
R+G+B +-* (Backplane +- A) = ((Result {5, 6, 7, 8}) ADD) DIV4 + Remainder

1 : 2 : 3 : 4 > Nn
1b : 2b : 3b : 4b > Nn

Ri
Ri 5 Ri
1 : 2
Ri 6 R 8 Ri
3 : 4
Ri 7 Ri
Ri

*

FMA Accumulators : Comment on denoising

Blend +- Addition Scaling & Super Sample

Pn assumed Base Value 32/48 (for all contained data(100%)) : Centre Pixel surrounded by 8 Virtual Pixel Pv,

(Super Sampling & upscaling)

Pn(48) - 8 : mov 8 ways 1 each (25%)

Blend Pv (Addition)

Blend with upscaling separation points

*

Application of +- & * to resampling on remerge with central Pixel of method One to three:

Counter operatively we move a fraction between point,

R <> {5 : 6 : 7 : 8} : {1 : 2 : 3 : 4}
Ri <> {1 : 2 : 3 : 4}
Ri <> {5 : 6 : 7 : 8}

Ri
Ri 5 Ri
1 : 2
Ri 6 R 8 Ri
3 : 4
Ri 7 Ri
Ri



Blend * MulDiv Scaling & Super Sample

Pn assumed Base Value 32/48 (for all contained data(100%)) : Centre Pixel surrounded by 8 Virtual Pixel Pv,

(Super Sampling & upscaling)

(Pn(48) * 0.15) * 8{Pv : 1 : 2 : 3 : 4 : 5: 6: 7: 8} = Pv * 8

+ Pn{Centre Pixel}

Pn(48) - 8 : mov 8 ways 1 each (25%)

Blend Pv (Addition)

Blend with upscaling separation points

*

In the case of Audio & data the application of data path interleaving the principle constitutes,

Data Channels A to X & RGB & A constitute channels of B to X & A is alpha & XOR,

Channel interpolating with alpha blending constitutes one of audio & haptics' greatest improvements..

A Key weaknesses of both Audio & 3D Data is clipping & also blending..

We need this to happen fast though! & on very efficient hardware; Such as Bluetooth headsets.

R+G+B +-* (Backplane +- A) = Result |

Channels A1 to A53 +-*(Variable Alpha Level, Constituting Smoothing)*-+ B1 to B53 (Left ear surround)

Channels C1 to C53 +-*(Variable Alpha Level, Constituting Smoothing)*-+ C1 to C53 (Right ear surround)


The MUL/+- Method should work fine on AMD, I am sure from papers that they do have MUL/-+ for machine Learning & Scaling, & At a minimum Supersampling Scaler Architecture (Works on all GPU) Applied Workarounds: disable_vp_scaling


RGBA/X Image format & Display Colour Bit/rate Enhancer & Compressor (c)Rupert S

*
All Display Connector types include RGB+BW; VGA, DVI, HDMI, Displayport; Image formats include RGB+ BW &or Alpha

Connector Type : Red + Green + Blue + BW Channel in other words 10:10:10+10 with the BW Channel 30Bit to 40Bit,
If alpha blended with image of BW Channel; Both BW or from Black to Red:Green:Blue,

The Backplane "BW Channel" Alpha can be more complex:

Black & White adds both light & shadow & by combination multiplies the colour palette
Rainbow Red:Green:Blue add an immense diverse colour palette

CSS Pattern Spectrum Vector(tm)
CSS Colour plane adds Pattern Spectrum Vector(tm) & is very small in file size but surprisingly good!
*

Produces results & with compression on the BW Channel set to specify Colour precision &or Image Precision & advantaging ourselves of the backplane colour of the display or composing ART/Image/CSS,
Higher Depth perception is created & upgrading bit-depth; Compression takes care of Bandwidth/Storage Usage issues.

One Colour (Alpha) Subtracts from Data Rate used by image while multiplying the image colour depth & detail

Most particular to texture, textures, raytracing & displays for the ultimate blend

Also reduces total bandwidth used for new detail by utilizing Alpha paint to blend the colour pallet of RGB into the underlaying Colour, Colours, Pattern or Frame..

The merge is additive & renders the full composed colour spectrum with light & shadow,

RGB -+ or * Backplane colour producing the results of both reducing data Send, Save/Reproduced

R+G+B +-* (Backplane +- A) = Result

Achievable with MulDiv Mul+- & XOR so Computation can be relatively affordable & Quick; Computation & Power wise

CSS Pattern Spectrum Vector(tm)
CSS Colour plane adds Pattern Spectrum Vector(tm) & is very small in file size but surprisingly good!

HSL CSS Colour tables work in 3 modes:

RGB with A Alpha on a CSS Defined colour or
RGB + Defined colour gradient +&OR simple layer Image
Or

2 Plainer such as Single Layer backing WebP : R G B + Alpha level & combinatory back layer,
Simple = High Vector Compression,
Pattern = Maths & average dither
Image (Simple Blocky)

Background can be pure 2 tone : Black To White & Red, Green, Blue to > Black or To White

Priority of back plane is higher compression rates & the aim is to turn 8 bit per channel in to +8 (example)

8, 8, 8 = 8, 8, 8 x 8 Image (8 of one colour to black or to white)
10, 10, 10 = 10, 10, 10 x 10 Image (8 of one colour to black or to white)
24 = 36
30 = 40

On a screen the natural shade is Black; So on a screen you additive to black or Divide total colour by black..

Depending on transparency levels...

Research topic RS

[reviewforthetube]

Makes sense tbh and we'll let's see the competition start! Doesn't bother me one bit tbh but I know many many will freak out about it haha!

[johnpaulbacon]

Great video - Every thing sounds great, but I am reserving right to make any opinion until everything has been released and see how it all performs. I am very happy with my Power Color Red Devil RX 6900xt for now.

[filds]

I wish the modding community somehow able to implement RSR for older cards :(,

[chickenpasta]

I don't care if RDNA 2 and below cards don't get the full FSR 3 but I really hope we get some of the improvements of it if not all of them

[gunzorkgaming]

Hope FSR 3.0 will work on 6000 series GPU.

[zhila]

this is amazing. now i want RSR 2.0 and 3.0

[TheTaurus]

In the FSR 2.0 Open Source Code is an Option implemented to run FSR over the Raytracing Accelerators in the Shaders.
So it gives Hardware Accelerated FSR.

[serphvarna]

Itd be so cool that 2.0 is the standard but if the game detects 3.0 compatible gpu itll use that instead

[nikita]

I hope they will get FSR 3.0 on RDNA 2 Architecture GPUs... I mean FSR 2.0, runns on GPUs under RDNA architecture ^^