AMD Radeon RX 6600 XT Linux Gaming Performance

[reba]

The results you posted were the outliers in that review.




I thought you didn't have any preference yet your seem to go shill for Nvidia a lot. Some of the previous posts feel like PR coming from you.

So I admit I have been wrong.

While theoretically PCIe 4.0 x8 is twice as fast ast PCIe 3.0 x8, in practice, for the mentioned card, it makes a difference of... 5%, as shown by birdie's benchmarks.

[artivision]

1. DLSS is NOT a Lanczos upscaler. DLSS can actually restore fine details as seen in multiple reviews, while FSR by definition cannot do it.

https://www.anandtech.com/show/15648...motion-vectors



(open in full screen)

2. WHAT EFFECTS ARE AVAILABLE IN NVIDIA BROADCAST?
The NVIDIA Broadcast app includes:

• Noise Removal: use AI to remove background noise from your microphone feed – be it a loud mechanical keyboard or doorbell ringing. The AI network can even be used on incoming audio feeds to mute that one friend who won’t turn on push-to-talk.
• Room Echo Removal: Use AI to filter out the room echo - or reverb - that typically occurs in small rooms, for clearer audio output.
• Virtual Background: use AI to remove the background of your webcam feed and replace it with game footage, a replacement image, or even a subtle blur.
• Auto Frame: use AI to track your head movement, automatically cropping and zooming so you’re always in the frame, even as you move around.
• Video Noise Removal: Enhances the quality of a webcam, removing the visual noise that is present in low light situations to get a cleaner picture.

3. I don't understand what you're trying to say.

Come on man get your mind straight, there is no parthenogenesis. There is only one high end upscaler available on the market "Lanczos". This includes software, TVs, Gpus and others. Also DLSS doesn't work when you strafe and have a lot of temporal artifacts.

https://docs.unrealengine.com/4.26/e...eenPercentage/

[Myownfriend]

Come on man get your mind straight, there is no parthenogenesis. There is only one high end upscaler available on the market "Lanczos". This includes software, TVs, Gpus and others. Also DLSS doesn't work when you strafe and have a lot of temporal artifacts.

https://docs.unrealengine.com/4.26/e...eenPercentage/

Unless your goal is to counter-troll then you gotta drop this. It's not Lanczos or TAA. DLSS is AI upscaling and Nvidia aren't the only company to do this, they're just the only one I'm aware of that use a real time version of it for games.

[artivision]

Unless your goal is to counter-troll then you gotta drop this. It's not Lanczos or TAA. DLSS is AI upscaling and Nvidia aren't the only company to do this, they're just the only one I'm aware of that use a real time version of it for games.

No AI, that v1 to v1.9 was dropped. Today with v2 and above is just temporal upsampling that another company created and have nothing to do with TAA, only use's the game's motion vector information that recorded to support TAA. In the future Sampler Feedback will also replace those information.

[Myownfriend]

No AI, that v1 to v1.9 was dropped. Today with v2 and above is just temporal upsampling that another company created and have nothing to do with TAA, only use's the game's motion vector information that recorded to support TAA. In the future Sampler Feedback will also replace those information.

Where are you getting that v2.0 doesn't use Deep Learning? The big change between it and 1.9 is that 2.0 uses tensor cores instead of shader cores. That's it. All versions of DLSS used neural networks trained with supersampled frames from the game. That training data is then stored in the driver and used to try to redraw the high resolution images from lower resolution input data using motion vectors as hints. It's actually a much more resource intensive upscale than the temporal upsampling that you linked to.

[artivision]

Where are you getting that v2.0 doesn't use Deep Learning? The big change between it and 1.9 is that 2.0 uses tensor cores instead of shader cores. That's it. All versions of DLSS used neural networks trained with supersampled frames from the game. That training data is then stored in the driver and used to try to redraw the high resolution images from lower resolution input data using motion vectors as hints. It's actually a much more resource intensive upscale than the temporal upsampling that you linked to.

Nope v1 is AI upscaling that failed because of blurriness and v2 is Temporal Upsampling. Yes Lanczos and other upscalers can be programmed like fp32 or fp16 or matrix math (tensor). When NV said that v3 will be used on all TAA games but with per game activation, just meant that upscaling should be done before UI, blur or grain and nothing more. No per game training, that is old news. NV sells hot air and its not the first time, they must say AI or train a few insignificant things in order to not get sued for fraud.

[Myownfriend]

Nope v1 is AI upscaling that failed because of blurriness and v2 is Temporal Upsampling.

Yes, temporal upsampling... using AI.

You're mixing DLSS with FSR.

https://www.extremetech.com/gaming/325512-amds-fidelity-fx-super-resolution-based-on-lanczos-filter-which-is-pretty-fine

[smitty3268]

Could AMD release a card with a full 16x wide PCI-E bus? They could. Would it cost a tad more? It would. By releasing a card with a limited bandwidth, did they increase their margins while screwing their customers? Yes, they did.

Isn't that the same thing as saying NVidia "cheated" by releasing a 3070 with 8GB of memory?

Could NVidia release a card with a full 12GB of memory? They could. Would it cost a tad more? It would. By releasing a card with limited memory, did they increase their margins while screwing their customers? Yes, they did.
/end sarcasm

Of course, at a certain point you just have to admit that the 6600xt is meant to be a cheaper mid-level card, and that inherently has certain performance/cost tradeoffs vs a more expensive card, and you have to evaluate whether that's worth it to you or not. The 3070 is arguably worse since it is meant to be a more expensive tier card.

[drakonas777]

AMD/FOSS related thread almost every single time:

birdie: Statement1, statement2, statement3 why AMD/FOSS is bad/inferior;
<SOMEBODY> statement1, statement2 are wrong, and here is reason/source/evidence why;
birdie: But statement3 is correct, so I'm right and you are AMD/FOSS fanboy. Also you personally attack me now, and your sources are biased towards AMD (in my opinion).

[reba]

Yeah, it's not worth the hassle.

It's like seeing a pupil writing "12 + 11 = 24" and you try to explain why and where something went wrong with the mathematics and then a struggle arises with tears, foot stomping, screaming and all the drama and you so "wtf, I just wanted to explain and help" and they so "I don't like you" and you so "wtf, I just tried to..." and they so "bruh, what is wrong with you, also I'm already 3rd grade and I do mathemuticks more longer as then you" and you so... yeah, you let them stand in their puddle and turn away as it's probably the best for both of you.

Edit: no disrespect meant, it's just me getting clean with myself and aligning myself how to deal with further so called discussions