TSMC Completes Its Latest 3 nm Factory, Mass Production in 2022

[Sovsefanden]

It's 2.0 quality mode, my friend.
And you can see images are blurred even on those sites that are hyping it.


My condolesces. Did you undervolt it yet?


Oh, so kind of Jensen Huang to hide the uncomfortable pics, ain't it?
But I'll go to, let me generous, overclock3d review, hyping the hell out of it, oh, look at this, from the page 2 of the review:


Look what happens to the bush

View attachment 177530

When magic was applied:

View attachment 177529

Now, to be clear here:

It is expected from temporal anti-aliasing.
It does improve the lines (again, grass, eyebrows, hair) you could see it on the right of the same picture.

But the point is: it is NOWHERE close to what it is hyped to be. Not even freaking remotely.
It is just the best TAA derivative we have, not more, not less.

DLSS 2.0 in Death Stranding - Nvidia 4K Performance Trump Card | DLSS 2.0 Image Quality Compared | Software

DLSS 2.0 Image Quality Compared

www.overclock3d.net

You are clueless, so funny when AMD fanboys are raging

I have tried DLSS 2.0 myself, because I can actually afford new and relevant hardware

 

[medi01]

DirectML will never bring what DLSS does

Brought to you by... "RDNA2 will only compete with 2080Ti" folks, known as "spot off team"...
You sound like you don't know what DirectML does, little one.
(if you wanted upscaling tech from AMD, it's FidelityFX CAS, Checkboard rendering)
But let me elaborate:

DirectML is Microsoft's API that reduces API overhead when doing neural network inference (i.e. "processing" input through the neural network that already has been trained).

The demo has brought up "NN upscaled" vs billinear upscaling (not the best upscaling we have mind you) visible on this pic:

Looks pretty good, although, one needs to note that:

• this was done using NV's NN along with weights
• the said stuff is available on the respective github
  
• too slow even on 2080Ti, sad

It wipes the floor with TAA derivative, but is too computationally expensive at the moment.

Gaming with Windows ML - DirectX Developer Blog

Neural Networks Will Revolutionize Gaming Earlier this month, Microsoft announced the availability of Windows Machine Learning. We mentioned the wide-ranging applications of WinML on areas as diverse as security, productivity, and the internet of things. We even showed how WinML can be used to...

devblogs.microsoft.com

[it is too painfull to read, whaaat, that picture cannot be real, whaaaa whaaaa.... but I can still write something insulting!]

No problems, and sorry if it offended your feelings as an owner of 3080.
I hope you didn't overpay scalpers.

I can actually afford new and relevant hardware

Oh, you poor thing, could not afford 3090? So sad.

 

[Sovsefanden]

Brought to you by... "RDNA2 will only compete with 2080Ti" folks, known as "spot off team"...
You sound like you don't know what DirectML does, little one.
(if you wanted upscaling tech from AMD, it's FidelityFX CAS, Checkboard rendering)
But let me elaborate:

DirectML is Microsoft's API that reduces API overhead when doing neural network inference (i.e. "processing" input through the neural network that already has been trained).

View attachment 177532

The dome has brought up "NN upscaled" vs billinear (not the best upscaling we have mind you) visible on this pic:

View attachment 177531

Looks pretty good, although, one needs to note that:

• this was done using NV's NN along with weights
• the said stuff is available on the respective github
  
• too slow even on 2080Ti, sad

It wipes the floor with TAA derivative, but is too computationally expensive at the moment.

Gaming with Windows ML - DirectX Developer Blog

Neural Networks Will Revolutionize Gaming Earlier this month, Microsoft announced the availability of Windows Machine Learning. We mentioned the wide-ranging applications of WinML on areas as diverse as security, productivity, and the internet of things. We even showed how WinML can be used to...

devblogs.microsoft.com

No problems, and sorry if it offended your feelings as an owner of 3080.
I hoped you didn't overpay scalpers.


Oh, you poor thing, could not afford 3090? So sad.

Thanks for a good laugh

 

[medi01]

Thanks for a good laugh

NP, I hope you've learned something.

 

[Sovsefanden]

NP, I hope you've learned something.

An AMD fanboy with an old ass system acting like you have experience with new tech, you are not in a position to learn anyone about hardware

 

[user112]

they may call it 3nm but in reality it's probably more like 4.5nm given how they brand their nodes.

 

[TheoneandonlyMrK]

This solution (3nm) is based on a 12 inch wafers but in the future the industry will move to 18 inch wafers, which means higher utilization of the fab and better pricing per wafer and eventually better prices to the end user. Basically much more dies per wafer. This die per wafer calculator show the various options per wafer size: https://anysilicon.com/die-per-wafer-formula-free-calculators/

It's a old thread but so is that 18" promise, the simple fact is you build the fab to do 18" , or you don't.

And few have or are.