Double the frame rate of the old graphics card without spending a penny!In the future, all players owe AMD a thank you – AMD, graphics card, external graphics card – Fast Technology (the media of Drive Home) – Technology changes the future

It doesn’t matter what brand of graphics card you’re using now, what kind of games you’re playing.

There is a high probability that in another month, you don’t need to change the graphics card, overclock, or flash the driver.

The game frame rate can be doubled in vain.

At that time, I don’t know if those profiteers will cry to death.

In fact, it’s all thanks to a new technology AMD is about to release – FSR 2.0.

Seeing this, some friends may say:

I know that AMD released an FSR 1.0 technology last year.

A chicken rib, scattered and scattered~

Hey~ wait a minute! Last year’s FSR 1.0 was indeed a useless and useless technology, I admit it.

But this time, although the version number of FSR 2.0 is only one digit, it is completely different from its predecessors.

However, it is such a king-level FSR 2.0. I have searched the whole network and have not seen a few people who have an in-depth chat with you.

Then, it’s up to me to criticize you to tell everyone about it~

(The front shake is slightly longer, and those who are familiar with this story can skip directly to the next section)

>/ AI oversampling to change framerate gameplay

Before we can talk about what FSR 2.0 is, we need to dial back the timeline.

Take a look at another technology released by Nvidia next door in 2018-2019: DLSS.

This is the case. Since the end of 2018, NVIDIA has successively released a series of new RTX 20 series game graphics cards.

In this series, in addition to the traditional graphics computing unit, the game graphics card also integrates an independent unit dedicated to AI computing for the first time.

With this independent AI computing unit, that is, “Tensor Core”, NVIDIA has made an operation that everyone has never seen before:

In some specific games, the graphics card only renders the 720p resolution game screen, but what we see on the monitor is 1080p resolution.

Likewise, the graphics card renders the game at 1080p resolution, but ends up rendering it at 4K resolution.

There is a gap of three-quarters of the pixel in the middle, which is abruptly filled in by the AI ​​computing unit in the graphics card!

This technology, named by NVIDIA as Deep Learning ( Learning ) Super ( Super ) Sampling ( Sampling ).

That is, the DLSS we mentioned earlier.

Since three of the four pixels depend on guessing, there are many cases where DLSS guessed wrong and rolled over at the beginning. (For example, the blur in the left frame of the picture above)

However, with the technical iteration, the picture output by the current DLSS 2.4 technology is no different from the traditional computing rendering.

And the efficiency of this thing is terrifying – in those games that support DLSS, the number of frames of AI supersampling-assisted rendering can be directly doubled compared to traditional methods.

It is precisely because of DLSS that the RTX 2060 gaming laptop at my house two years ago has been carried until now.

>/ The big pattern of the red camp counterattack

But everything has two sides. Although DLSS is powerful, it also has two sad limitations:

First, its operations must be handled by an independent AI computing unit, which means that only NVIDIA’s RTX series graphics cards can enjoy this technology.

The low-end entry-level MX series, the older GTX series graphics cards, and all the graphics cards of the rival AMD are all out of touch with DLSS.

Moreover, DLSS technology requires the integration and adaptation of games. Some old but configuration-intensive games such as Graphics Card Crisis, GTA 5, and Horizon 4 will not enjoy the joy of doubling the frame rate when playing them.

But just last year, AMD released another technology called FSR, which opened the pattern.

The full name of FSR is FidelityFX Super Resolution, which is very long and hard to remember.

But it doesn’t matter.

It is actually a picture supersampling technology similar to DLSS.

but! FSR does not need a dedicated AI computing unit for support, and the traditional graphics card rendering unit can execute the sampling algorithm required by FSR.

So FSR does not limit the brand and model of graphics cards: AMD graphics cards can run, NVIDIA and Intel graphics cards can also run, and even the garbage core graphics integrated in the CPU can run.

Most importantly, FSR does not require game adaptation, and most games can enjoy the frame rate bonus it brings

(If you talk about this, then I’m not sleepy.jpg)

Um. . . In fact, what I said above is ambiguous:

The details are as follows – although AMD officially announced that FSR needs game support, but because AMD has open sourced the entire function part, there are a lot of community gurus to help the bad game developers to adapt.

For example, the graphics card crisis and GTA 5 mentioned above, you only need to replace the display rendering library in the game directory with the version changed by the gods and demons, and you can easily enable FSR in the game.

The frame rate can be increased by almost 60%.

And later, this magic modification scheme also derived a “painless universal version”:

Just download a program less than 2 MB in size, it can automatically identify the currently running game window, and apply FSR oversampling scaling to it without modifying the game file.

For example, “Lossless Scaling” that can be downloaded and used for free ▼

Under these painless general solutions, almost all games cannot escape the “magic claws” of FSR.

I even used FSR for Genshin Impact.

I have to say that once the native rendering resolution drops, the game becomes much smoother, and the computer fan that was buzzing before immediately stopped moving.

>/ Spatial scaling

The fatal flaw of FSR1.0

According to the situation, should FSR 1.0 kill the Quartet, and the NVIDIA DLSS has been defeated?

Not really. . . Although FSR 1.0 is more applicable, its picture quality is also worse.

Let’s take a look at the following set of comparison pictures. The details of the pictures obtained by FSR oversampling are a lot worse than the natively rendered pictures:

Elements are missing, smeared, serrated, three cauldrons a lot.

On the other hand, when comparing the pictures calculated by DLSS with the native rendering, there is really not much difference.

The key behind this is that DLSS is a time-based supersampling algorithm that incorporates AI operations. And AMD’s FSR 1.0 is a spatial dimension-based supersampling algorithm.

Don’t worry, everyone, I will repeat the above sentence in human words. . .

Let’s take this example: if I gave you a picture with a thin code and asked you to restore it to its original appearance based on this “heavy code rate” photo, what would you do?

Yes, I know everyone has seen it, the person in this picture is Obama.

But what if you don’t know? It’s just a guesswork.

Well, the following is what the computer completes by guessing through algorithms.

Not Obama at all, the race has changed, right?

AMD’s FSR 1.0, the logic implemented is actually a bit similar to “guess mosaic”:

After the graphics card renders a frame of 720p resolution, the algorithm just stares at this frame and racks its brains to complete the pixels to 1080p.

As a result, it is possible to make Obama into Obama cattle.

And DLSS on NVIDIA’s side is smarter, it’s not just staring at this low-resolution original picture.

Instead, the 1080p high-definition frame rendered in the previous frame is also pulled out and placed next to it for reference.

Then, by analyzing the changed parts, accurate completion is made.

It’s like two people doing sketches, one person only relies on the brain to supplement YY, and the other person puts a reference object in front of him.

You say, who can draw better.

>/ Learn from each other and welcome new students

However. . . Here it is, it’s not over.

No, the pressure has come to our main leader today: FSR 2.0.

I didn’t expect AMD to be so powerful. In less than a year, the new version of FSR directly abandoned the previous space-based oversampling algorithm.

It uses the same time-based oversampling algorithm as NVIDIA DLSS.

And from the Demo of AMD’s official demonstration, FSR 2.0’s restoration of scene details and improved frame rate are almost comparable to Lao Huang’s DLSS.

Click to see the high-definition picture ▼

If this is over, I can at most say that AMD is really shameless and brave, and finally catches up with the level of the next door two years ago.

No, the most outrageous thing about this is that although the temporal oversampling algorithm is used, FSR 2.0 still does not require additional AI operations to achieve!

And still open source!

That said, it still runs on most graphics cards you can think of, and most games.

Thanks in advance for a wave of big cows who will do folk adaptation in the future. . .

The AMD official also teased Nvidia about this in the blog: “Although AI is a useful wheel, for a good oversampling algorithm, AI is just a sufficient and unnecessary condition.”

But then again, after all, FSR 2.0 uses a better oversampling algorithm, so the requirements for the graphics card itself will definitely increase.

AMD officials said that graphics cards with performance above the GTX 1070 line can get the most ideal FSR 2.0 frame rate improvement experience.

From this point of view, it is estimated that the situation of “nuclear display can also be used” in FSR 1.0 will not appear.

But I think if your graphics card is GTX 1050 or GTX 1060, you can still give it a shot.

“I can mention it, but only a little bit” ▼

>/ Thank you, Su Ma

I believe that the vast majority of our game-loving friends buy their own graphics cards are N cards.

Because most of the time, NVIDIA graphics cards represent more advanced technology, more complete functions, and higher stability.

Maybe some A-players don’t see me saying this – I understand, after all, I’m also a little A-player myself.

But think about it, who first supported the ray tracing technology, who first supported the complete DX12U technical specifications, and who first adapted the WDDM 3.0 driver model?

Not to mention the cuDNN syntactic sugar, NVIDIA Broadcast, RTX Voice and other practical components that the N card has. Every time I think of these, I still have a RX 6800XT in my office desktop and I will drool. . .

However, after the release of FSR 2.0. I found that I don’t envy N-card users anymore.

Because the function of the N card is cool, it is only cool for its own users.

But what AMD has introduced, like Mental, Vulkan, OCAT, and this time FSR 2.0.

They benefit all gamers.

Hashtags: AMD graphics card external graphics card