RTX 5080 could be faster than RTX 4090 in ray tracing, top 3 Blackwell GPUs tipped to...

[Disiw]

And where is the news in this? on the contrary, if it was not faster it would be real news

it has been always that the top end previous generation be beat by the less top end newer generation model: 4080 was faster than the 3090 Ti in nearly all games, and the 3080 was faster than the 2080 Ti

4080 is not nearly faster than 3090 Ti. It is always faster.
While using much less power and has support for DLSS3 + Frame Gen on top.

 

[HardReset]

MCM with higher power draw than Monolithic is not normal. Just look at Ryzen.

I know MCM is about scalability. Everyone knows that.

Monolithic Ryzen consumes much less than MCM Ryzen.

 

[Disiw]

5070 TI = 999$
5070 = 799$
5060 TI = 599$
5060 = 499$

After 4/5nm has been confirmed, instead of 3nm, I doubt this is going to be the case now.

5060 to 5070 is what AMD will compete with (on performance) - 5090, 5080, including 5070 Ti if it launches at all, will be faster than anything AMD has. 5080 will beat 4090 and 5070 Ti should deliver 4090 perf or close.

I could easily see 5070 Ti at 799-899 dollars if this will be the case.

5080 at 1199 dollars.
5090 at 1799-1999 dollars.

Nvidia can price them however they want. AMD is sleeping.

 

[ScottSoapbox]

Nvidia easily dominates using monolithic.
AMDs MCM approach for GPUs kinda failed. Uses more power than Nvidia. MCM should deliver better performance per watt not worse than monolithic approach.

Lets see if AMD is going back to monolithic with 8000 series.

You seem to be missing the point.
Monolithic chips will keep getting bigger because node shrinks are running into physics. Bigger chips are more expensive and more likely to include defects because they are bigger.
A Multi-chip approach solves this size issue, but the trade-off is connecting the chips fast enough, which requires additional R&D to redesign everything.
The question isn't IF but WHEN Nvidia will switch approaches. The deciding factors will be how much they can charge for their largest chip and how many more cores they need to add to keep their performance lead.
(It is also worth mentioning that Nvidia already switched to multichip for their AI cards for the second reason)

 

[Disiw]

You seem to be missing the point.
Monolithic chips will keep getting bigger because node shrinks are running into physics. Bigger chips are more expensive and more likely to include defects because they are bigger.
A Multi-chip approach solves this size issue, but the trade-off is connecting the chips fast enough, which requires additional R&D to redesign everything.
The question isn't IF but WHEN Nvidia will switch approaches. The deciding factors will be how much they can charge for their largest chip and how many more cores they need to add to keep their performance lead.
(It is also worth mentioning that Nvidia already switched to multichip for their AI cards for the second reason)

Well AMD is going back to monolithic for RDNA4 according to several leakers.

I guess AMDs first gen MCM GPU approach was not really a succes.