• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

NVIDIA 4000 Series

Lightnix said:
Yeah because I think the shaders have to be able to communicate with them as the scene is being computed.

My guess is they will want basically break it down by the existing architecture block diagrams (I know we're in a green team thread but bear with). For AMD at least it would be similar to how MCDs have been created from the periphery of those diagrams between RDNA2->RDNA3 to include the memory controllers and infinity cache.

The next step probably means splitting out whole shader engines (which include groups of shaders, TMUs, RAs, etc.) or groups of them into their own separate dies as these are supposed to operate relatively independently of each other. Then the GCD would become a much smaller coordinator with the geometry engine or whatever.
Click to expand...
I guess SLI/CF would be easier to have for starters.
 
Murphy said:
Personally i don't think we'll ever switch over to full path tracing, the computation requirements and by extension the amount of silicon needed far exceed what can reasonably be considered a cost effective die.

The 4090 has a 600 mm² die already and it's only capable of path tracing in limited amounts, all be it less limited than lesser cards, even while being assisted with upscaling.

/Hot take :D
Click to expand...

We've seen pretty big leaps from the Pascal series only being able to do 10FPS at low resolution in Quake 2 RTX to stuff like the 4090 doing 100 FPS at 4K - next 1-2 generations are going to see huge increases in what is possible with path tracing.
 
Kavalier said:
What exactly do you mean by this, are there still more 4090 cards to be released?
Click to expand...
I mean they haven't released all the "standard" 4000 models yet like the 4070/4060/4050 etc.

Nvidia don't tend to start releasing "super" variants until the standard release of all the lineup is completed. That's assuming they even will release "super" variants. That's only happened with the 2000 series. It didn't happen with the 3000 series, so they might not bother with the current 4000 series.
 
If the 4060/ti struggles at 1080p high (not even ultra), that is gonna be bad. I wonder if they double the memory like they did for the 3060. Although at 128bit (4 chips), dense enough chips may not exist.
 
Last edited:
muon said:
If the 4060/ti struggles at 1080p high (not even ultra), that is gonna be bad. I wonder if they double the memory like they did for the 3060. Although at 128bit (4 chips), dense enough chips may not exist.
Click to expand...

They will have to otherwise reviews won't look good if future ports are equally as bad as forgotten.

I'm firmly holding out for the 50xx series now given even rdna 3/ada is ******** the bed now especially if path tracing becomes more common :D
 
You must log in or register to reply here.
Back
Top Bottom