New GeForce G102 to be a 630mm2 8nm GPU,RTX coming to 3050,CUDA and RT/Tensor power decoupled

[cucker tarlson]

Chip size and power supply details leaked - NVIDIA Ampere GA102 in Samsung’s 8 nm increasingly likely | igor´sLAB

Before I act as a (future) painter and create a possible package including the dimensions known so far, we should review the rumors and leaks of the last days. Let’s take the usually well-informed…

www.igorslab.de

NVIDIA's GeForce RTX 3060 Mainstream $300-$400 Ampere Gaming Graphics Cards Rumored To Feature GA106 GPUs, GA107 Entry-Level Ampere GPU Taped Out Too

Alleged specifications of NVIDIA's next-gen GeForce RTX 3060 & RTX 3050 graphics cards with Ampere Gaming GA106/GA107 GPUs have been detailed.

wccftech.com

NO TEU. GA102 has taped out. It was based on SAMSUNG 10nm. And the new tegra will use the same process node

 

[Vayra86]

Thats a whole lot of conjecture and speculation. As always, I am patiently in a holding pattern until an official release.

 

[John Naylor]

I don't understand the fascination with die size and # cores, or anything else frankly. It's not as if I get to by a smaller case and finish tasks faster because someone shrunk their die size,. Does 16 cores necessitate that my video editing will complete inn less time ? Obviously not, so why should anyone have reason to care ?

 

[mtcn77]

I don't understand the fascination with die size and # cores, or anything else frankly. It's not as if I get to by a smaller case and finish tasks faster because someone shrunk their die size,. Does 16 cores necessitate that my video editing will complete inn less time ? Obviously not, so why should anyone have reason to care ?

The same reason we like chonkers. Size matters.

 

[Fouquin]

I don't understand the fascination with die size and # cores, or anything else frankly. It's not as if I get to by a smaller case and finish tasks faster because someone shrunk their die size

For a lot of professionals more cores does mean their work gets done faster. Anyone working with CAD or 3D design software that has the proper multithreaded optimization (read: all of them at this point) will benefit greatly from the extra cores. More importantly, die size is directly correlated with process node and transistor budget. Reducing the node affords more transistors in the same die space. This is how we got to incorporating local caches, controller logic, purpose built fixed function logic, and much much more. You have reaped the benefits of these improvements without even realizing it.

 

[TheoneandonlyMrK]

I don't understand the fascination with die size and # cores, or anything else frankly. It's not as if I get to by a smaller case and finish tasks faster because someone shrunk their die size,. Does 16 cores necessitate that my video editing will complete inn less time ? Obviously not, so why should anyone have reason to care ?

Kind of a strange standpoint to have and be into tech, for me the tech came from the fact I wanted to try new gaming technologies and over time , almost all the time, in fact, there are always games and now to a lesser extent applications admitedly, that could run better ,and i have not got 8K 144hz and the power to run at least crysis on it yet, let alone the cinematic life-like games i want, oh n vr.

Nothing matters more than the features and performance they can offer for a price, though it would help if they realised the majority of us are not freaking millionairs.


Yall better get some cores ouut, MOOOAR cores ,ya hear me Nvidia, yeh you n the other two.

 

[ppn]

So GA104 ~~330mm2
8nm ~~ 66% better density than 12/14/16nm.
Fake 7nm found in Navi10, ~~~66% better density than 12/14/16nm.

 

[cucker tarlson]

So GA104 ~~330mm2

bigger probably
+400mm2

 

[xkm1948]

Would be nice if GA102 use A100's TSMC 7nm.