• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

[pcgamesn] AMD is giving Threadripper 2 moar cores and a top TDP of 250W

scannall

Golden Member
Jan 1, 2012
1,751
1,177
136
it probably is meant to be read "threadripper 2" + "more cores".... Not "2 more cores"...
Possibly. Adding a CCX to the die would likely be easier. For a 24 core monster.
We'll see. Exciting times ahead. ;-)
 

CatMerc

Golden Member
Jul 16, 2016
1,114
1,146
136
More cores would require a new motherboard due to the memory channels.
 

NeoLuxembourg

Senior member
Oct 10, 2013
752
206
116
Same here, I think TR 2xxx will "only" be Zen+ dies on the actual Platform.

Without having new dies, a >16 cores TR would require 4 Zepline dies. This would require a new substrate (Epyc can't be used because 8ch MC) and would limit the Zepline dies to a single MC, what would certainly have an impact on performance.

That said, because AMD will not release Epyc with Zen+ dies, could it be that they plan a X499 , 8ch MC platform for those Epyc Zen+ derivates?
 
  • Like
Reactions: Organik

scannall

Golden Member
Jan 1, 2012
1,751
1,177
136
More cores would require a new motherboard due to the memory channels.
Only if they added a third die to the package. Adding a CCX to each die would keep the same memory channels. But, as always with rumors, take them with a large grain of salt.
 

CatMerc

Golden Member
Jul 16, 2016
1,114
1,146
136
Only if they added a third die to the package. Adding a CCX to each die would keep the same memory channels. But, as always with rumors, take them with a large grain of salt.
Adding a CCX would require a new die, something I 100% doubt they'd do for Threadripper. It will most likely reuse the dies from Zen+.
 

wahdangun

Golden Member
Feb 3, 2011
1,004
139
106
Same here, I think TR 2xxx will "only" be Zen+ dies on the actual Platform.

Without having new dies, a >16 cores TR would require 4 Zepline dies. This would require a new substrate (Epyc can't be used because 8ch MC) and would limit the Zepline dies to a single MC, what would certainly have an impact on performance.

That said, because AMD will not release Epyc with Zen+ dies, could it be that they plan a X499 , 8ch MC platform for those Epyc Zen+ derivates?


There is no 12 nm EPYC+ they straight to 7nm.

So maybe in q4 this year they can release TR with EPYC 7nm based die, and increase core count and performance substantially.
 
  • Like
Reactions: maddie

CatMerc

Golden Member
Jul 16, 2016
1,114
1,146
136
The only way that might happen is if TR2 is actually Zen 2, which is unlikely. To do 24 cores, the only logical way to do it would be the Epyc configuration, and going that route more than halves the bandwidth of the inter-die data fabric due it it being split among four dies instead of two.
Or if it's a new motherboard. Don't need Zen 2 to achieve this config.

As for the bandwidth, it was a "might as well" move from AMD. The bandwidth provided to EPYC is already described as "over provisioned". The extra Threadripper links likely don't do much in reality.
 

tamz_msc

Platinum Member
Jan 5, 2017
2,787
2,470
136
Or if it's a new motherboard. Don't need Zen 2 to achieve this config.

As for the bandwidth, it was a "might as well" move from AMD. The bandwidth provided to EPYC is already described as "over provisioned". The extra Threadripper links likely don't do much in reality.
Hopefully we'll get to know in a few hours time.
 

LightningZ71

Senior member
Mar 10, 2017
632
566
106
I think the biggest likelihood is for TR2 to be just like TR1, but with binned and power pushed zen+ dies with a Max boost in the 4.4-4.6ghz range. Those speeds will generate a lot of heat and power demand, thus the higher tdp.

Less likely, TR2 is zen+ based, but is available in a Max cores configuration that has the two dark dies enabled and connected to the other two dies, but with no local ram channels. It's a 4ch package, each die is configured for 6 active cores at maximum achievable clocks.

Least likely, TR2 uses a new mcm. The package is essentially EPYC with zen+ cores, 4 enabled dies, each with only one active channel, and only 64 pcie lanes. They can make any core combination they want. The new mcm remains compatible with existing TR sockets. Those memory channels would have some interesting routing in the mcm.
 

eek2121

Senior member
Aug 2, 2005
729
783
136
You actually read wccftech?

Threadripper 2 will feature 16 cores. Any increase in cores would have to be done in the CCX due to socket compatibility. However, adding cores would increase memory and L3 latency and lower clock speeds. PCGamesN and WCCTech are both not reliable sites. Links to them should be banned IMO.

EDIT: AMD is not going to spring for a new platform/die configuration either. They are looking to increase margins. Whether or not we will get the 2800X is up for debate, but the 2900X, 2920X, and 2950X are not. Same configuration as last year....only faster.
 

mattiasnyc

Senior member
Mar 30, 2017
356
337
106
You actually read wccftech?
Occasionally.

The thing with sources is that when they're questionable you can still figure out if there's anything there worth 'ingesting'. When a source like that reports from an event it's far less likely to be bogus compared to them just quoting an unnamed source.

In addition to AnandTech I also found TechReport's report on it.

Threadripper 2 will feature 16 cores.
It fascinates me how some can be so certain of these things, yet clearly without having the knowledge to be that certain.

Any increase in cores would have to be done in the CCX due to socket compatibility.
I guess not.

PCGamesN and WCCTech are both not reliable sites. Links to them should be banned IMO.
Or, you know, you could have said "Thanks, guess my certain proclamations of what AMD would bring was wrong. Now I know." Followed by a thumbs-up emoticon.

EDIT: AMD is not going to spring for a new platform/die configuration either. They are looking to increase margins. Whether or not we will get the 2800X is up for debate, but the 2900X, 2920X, and 2950X are not. Same configuration as last year....only faster.
You should tell AMD it's not up for debate. I honestly don't understand how you can be so certain about these things....

But seriously, it's not unthinkable that the 2900x, 2920x and 2950x are "merely" updated Zen+ Threadrippers, whereas there are added to the TR lineup, with 24 and 32 cores respectively.
 

Shamrock

Golden Member
Oct 11, 1999
1,158
150
106
You actually read wccftech?

Threadripper 2 will feature 16 cores. Any increase in cores would have to be done in the CCX due to socket compatibility. However, adding cores would increase memory and L3 latency and lower clock speeds. PCGamesN and WCCTech are both not reliable sites. Links to them should be banned IMO.

EDIT: AMD is not going to spring for a new platform/die configuration either. They are looking to increase margins. Whether or not we will get the 2800X is up for debate, but the 2900X, 2920X, and 2950X are not. Same configuration as last year....only faster.
Well, if you don't like wccftech, then how about pcgamer?
https://www.pcgamer.com/let-the-core-wars-begin-amd-one-ups-intel-with-32-core-ryzen-threadripper-cpu/
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
Hot off the press: Threadripper 2 to feature up to 32 cores:
https://www.anandtech.com/show/12906/amd-reveals-threadripper-2-up-to-32-cores-250w-x399-refresh

4 active dies
2 dies with dual memory channels each
2 dies without direct memory access (+latency)
There has to be 1 DPC 8CH TR4 boards incoming for these CPUs, as otherwise they'll be the most futile consumer HEDT CPUs ever released.
A 4P CPU with half of the cores having no local DRAM available, meaning the memory accesses will be routed through GMI interface (> 120ns on top of the normal latency) to the cores with the memory present.
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
13,951
3,637
136
Crazy. The issue with the design though is the latency is not addressed at all. In fact, it may be worse.
If you're buying 32 cores and they're not branded Epyc, you probably don't care much about the memory latency - you care about core throughput. Which a 250W TDP platform with 4x binned 2700X dies would bring in spades. No one buys a 32 core CPU for gaming.
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
13,951
3,637
136
There has to be 1 DPC 8CH TR4 boards incoming for these CPUs, as otherwise they'll be the most futile consumer HEDT CPUs ever released.
A 4P CPU with half of the cores having no local DRAM available, meaning the memory accesses will be routed through GMI interface (> 120ns on top of the normal latency) to the cores with the memory present.
I suspect you are correct. We will see.
 

ASK THE COMMUNITY