ZEN+ Delidded

Interesting that the die-size is the exact same.

We all knew that the '12nm' was really just a slightly improved 14nm, but apparently the actual transistor size isn't what's improved.
 
  • Like
Reactions: erek
like this
Interesting that the die-size is the exact same.

We all knew that the '12nm' was really just a slightly improved 14nm, but apparently the actual transistor size isn't what's improved.

https%3A%2F%2Fblogs-images.forbes.com%2Fpatrickmoorhead%2Ffiles%2F2017%2F09%2F12lp-1200x731.png


Up to 15% higher density than 16nm, aka TSMC 16nm. The node properties are

TSMC "16nm":
48-nm Fin Pitch
90-nm Gate Pitch
64-nm 2D Interconnect Pitch
0.07µm² HD SRAM

GloFo/Samsung "14nm":
48-nm Fin Pitch
78-nm Gate Pitch
64-nm 2D Interconnect Pitch
0.064μm² HD SRAM

TSMC 16nm already was up to 15% less dense than 14LPP. So 12LP brings nearly zero density improvements over 14LPP, which is not strange because despite its marketing name 12LP is just an extension/tweaking of 14LPP.
 
Interesting that the die-size is the exact same.

We all knew that the '12nm' was really just a slightly improved 14nm, but apparently the actual transistor size isn't what's improved.

Well the chip is not fully 12nm or fiully 14nm its a blend. Just as INtel will not use a super small process for the uncore, AMD is only going to use the smallest process on the parts that matter most. The die size is probably the same size because of this or that economical reason. I bet ZEN 2, the real 2nd gen Zen is going to be smaller overall due to the process shrink across the board.
 
what is the point of taking of the ihs just to put a difernt substance in the middle. when the ihs is removed just plop a cooler right ontop of the die
 
Interesting that the die-size is the exact same.

We all knew that the '12nm' was really just a slightly improved 14nm, but apparently the actual transistor size isn't what's improved.

That empty space can later be filled with stuff. AMD hinted at a possible 2800x down the line.
 
That empty space can later be filled with stuff. AMD hinted at a possible 2800x down the line.

I don't imagine the 2800X being a different tapeout versus the 2700/X. It would be a LOT of extra effort for AMD to make a different silicon design for just one chip.
 
I imagine they'll reserve 2800 and 2900 names for TR4 chips using Zen+ dies. So a 2800X may be the next generation's equivalent to the 1900X
Respin take 6 months or so? They might stumbled upon some things might be a different stepping for 2700X by that time with different name?
 
View attachment 68771

Up to 15% higher density than 16nm, aka TSMC 16nm. The node properties are



TSMC 16nm already was up to 15% less dense than 14LPP. So 12LP brings nearly zero density improvements over 14LPP, which is not strange because despite its marketing name 12LP is just an extension/tweaking of 14LPP.

If they had taken the time to re-layout the chip, they could have shrunk it a little, but that would have taken more time and probably resulted in the same or lower speeds due to higher circuit density and the resulting hot spots. By shrinking the features and leaving the layout the same, they make it much faster to market and get a little more speed due to lower heat density. Not a bad plan given the minor shrink, but they lose probably 10 extra chips per wafer by not shrinking the die.
 
I don't imagine the 2800X being a different tapeout versus the 2700/X. It would be a LOT of extra effort for AMD to make a different silicon design for just one chip.

My guess is once they have enough binned chips for TR you'll see the 2800x release on am4 around the same time.
 
I imagine they'll reserve 2800 and 2900 names for TR4 chips using Zen+ dies. So a 2800X may be the next generation's equivalent to the 1900X

Or the 2800X will just be binned (but still mainstream parts), or will be released when yields are a little better. Maybe an extra 100MHz boost on the 2700X for same TDP or something like that.
 
Back
Top