[SOLVED] RIP 3090?

[thepersonwithaface45]

After the AMD press conference, it's $1000 for a 6900xt which rivals the 3090 at less power and smaller form factor.
It's $650 for a card that rivals the 3080.

You think nvidia is gonna release TI versions or super versions at an acceptable price that can compete with this?
I did not expect this at all.
The thing that worries me is going for an AMD card with an intel processor after they mentioned how having an AMD processor will benefit you with their smart tech.

 

[digitalgriffin]

Here is the quote from the leak or rumor:
"
Kopite7kimi, a leaker better known for revealing details on GeForce RTX Ampere series months in advance is now reporting on a new SKU that would fill the (already narrow) gap between GeForce RTX 3080 and RTX 3090 SKUs.
The graphics card would allegedly feature 9984 CUDA cores, 1280 cores more than RTX 3080. Kopite reports that the GPU variant on this SKU is likely the GA102-250. The memory configuration would also change to a 384-bit bus (compared to 320-bit on RTX 3080). A wider bus means that the card will likely feature a capacity of either 12 GB or 24 GB (same as RTX 3090).
.........According to his information, NVIDIA is also allegedly planning to release a GA102-150 model featuring 7424 CUDA cores, possibly codenamed RTX 3070 Ti."

I think we are in sync.

 

[renz496]

i think it depends in what nvidia intend to do. so far we haven't seen any kind of respond from nvidia yet. and i forgot to mention this in my previous post but i think when nvidia price 3090 at such massive price compared to 3080 despite very small difference in performance (not even 20% faster) they probably aim 3090 to get it's own niche. but if we only look it from gaming perspective such pricing indeed does not make sense at all. more than double the 3080 price for less 10 to 15 percent better performance and double the VRAM? it is beyond pure ripoff.

so will nvidia release a new GPU that is faster than 3090? that is entirely possible. hence the talk ampere refresh on TSMC 7nm process next year. some said nvidia can't do it because that will going to piss of 3090 buyer but if you look what nvidia has done with the pass titan then you know they will do still going to do it. but funny thing is the one that most often talk about this "people being pissed" is the one that not buying the hardware itself. right now GA102 sits at 628mm2 using samsung 8nm process. with TSMC we know nvidia can push the die size towards 800mm2 (A100 is 826mm2). so if nvidia want something that is faster than GA102 they just need to create new chip with more resource (more Cuda cores and everything) at TSMC 7nm node. with further optimization on the design itself like they did with second gen fermi the power consumption probably will not going to significantly increase than what we see on 3090 despite being bigger. for nvidia having that performance crown is everything.