Question Zen2, making a 16/12 core CPU process like an 8/6 core to get higher clocks on the fly?

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,654
136
Kedas it won't matter because the 6 and 8 core versions have the IO die as well. Meaning that in the 2 chiplet vs 1 chiplet, performance per clock is going to match. So while we don't know what the IO die means for performance the only difference we will see between the two versions is power usuage and maybe a little bit of clockspeed.
 

Kedas

Senior member
Dec 6, 2018
355
339
136
Kedas it won't matter because the 6 and 8 core versions have the IO die as well. Meaning that in the 2 chiplet vs 1 chiplet, performance per clock is going to match. So while we don't know what the IO die means for performance the only difference we will see between the two versions is power usage and maybe a little bit of clockspeed.

The performance hit will not be because it has a separate I/O die but because it has two 8 core dies that shares data over IF (like Threadripper), obviously very application dependent, Cinebench for example couldn't care less if it's 2 dies or 1. Games.... well that's why they introduced gamer mode.
 

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,654
136
The performance hit will not be because it has a separate I/O die but because it has two 8 core dies that shares data over IF (like Threadripper), obviously very application dependent, Cinebench for example couldn't care less if it's 2 dies or 1. Games.... well that's why they introduced gamer mode.

But that hit will be the same across the board. All data will be handled through the IO chip and then handed of to the needed core. We don't know what IF is going to be like in this chip, we don't know what AMD did with the IO chip in regards to keeping latency low, or if they did anything now. The IO die will make them all one node. they will all have the same access to the same memory at the same speed. Which is one of the bigger issues with TR.
 

TheELF

Diamond Member
Dec 22, 2012
3,967
720
126
The IO die will make them all one node. they will all have the same access to the same memory at the same speed. Which is one of the bigger issues with TR.
The question is if this node will be the slow one or the fast one...if both nodes need to connect to the outside I/O chip my bet is on both going to be slow,on TR at least one of them is fast.

Edit:typos
 

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,654
136
The question is if this node will be the slow one or the fast one...if both nodes need to connect to the outside I/O chip my bet is on both going to be slow,on TR at least one of them is fast.

Edit:typos
That is what we are going to wait out. I mean I doubt AMD makes this move for Consumer CPU's if they didn't have an answer for this. But I agree with you there.