Zen 2 rumors

Guys.... it is a rumors thread which includes hype, speculation and real leaks. I mean seriously, why pic nits with rumors? We will all find out the facts when they are ready actually released. When that happens we can all come back here and find out who was closest to reality or full of shit with their rumors, speculation, hype and leaks.
 
juanrga said:
No one negated that AMD is progressing. Just mentioning that those leaks about Zen2 are so suspicious as former leaks about Zen.



They were leaks. The 5GHz on air come from an easter egg publication. The faster than Kabylake on gaming from AMD slides. The IPC from leaks about CPU-Z. the HBM on Zen APUs from bitsandchips,...

All fake leaks.
Click to expand...
Actually the 5ghz leak came from a source you been quoting for a good while Canard PC.
cyberguyz said:
Guys.... it is a rumors thread which includes hype, speculation and real leaks. I mean seriously, why pic nits with rumors? We will all find out the facts when they are ready actually released. When that happens we can all come back here and find out who was closest to reality or full of shit with their rumors, speculation, hype and leaks.
Click to expand...

Some of it is quite evil raising expectations making anything less look like a flop.
Some of the speculation can be within reason but if you look at the differences between Zen and Zen+ the biggest step will be with Zen 2 but mostly because of the process used.

Clocks and IPC can be calculated guesses. So some leaks can be more plausible then others.
 
Last edited:
cyberguyz said:
Guys.... it is a rumors thread which includes hype, speculation and real leaks. I mean seriously, why pic nits with rumors? We will all find out the facts when they are ready actually released. When that happens we can all come back here and find out who was closest to reality or full of shit with their rumors, speculation, hype and leaks.
Click to expand...

Exactly. This is something I don't understand. "Your predictions aren't realistic, but mine are! LOLOL!" Come on, this is speculation and bullshitting. If this conversation were in meatspace, it'd be Ford vs. GM kind of argument over a few beers, speculating on what each company is gonna come up with next. It's harmless. If anyone is taking it SERIOUSLY, they are the problem, not the conversation itself.
 
DuronBurgerMan said:
Exactly. This is something I don't understand. "Your predictions aren't realistic, but mine are! LOLOL!" Come on, this is speculation and bullshitting. If this conversation were in meatspace, it'd be Ford vs. GM kind of argument over a few beers, speculating on what each company is gonna come up with next. It's harmless. If anyone is taking it SERIOUSLY, they are the problem, not the conversation itself.
Click to expand...

Speculation is fun! And getting feedback on predictions is even more fun!

What sense has to post your speculations/prediction on a forum if you don't want hear what others have to say?

There are realistic predictions and there is nonsense/hype. E.g. Expecting doubling the number of cores per die plus 15% IPC gains plus 4.8GHz stock cores for Zen2 is not a realistic prediction. What is wrong with mentioning it?
 
juanrga said:
Speculation is fun! And getting feedback on predictions is even more fun!

What sense has to post your speculations/prediction on a forum if you don't want hear what others have to say?

There are realistic predictions and there is nonsense/hype. E.g. Expecting doubling the number of cores per die plus 15% IPC gains plus 4.8GHz stock cores for Zen2 is not a realistic prediction. What is wrong with mentioning it?
Click to expand...

Very few of the predictions put here are terribly far off the mark. It is way too early for anyone to judge the what is realistic and what is bullshit. It simply remains to be seen whose crystal balls are cracked and whose gave more more accurate predictions. We won't know until the details are actually released by AMD. Anything else is nothing more than speculation unless the origin is actually breaking AMD's NDA. And from what I heard of NDAs from major companies like AMD, breaking them could be very, very expensive if you get caught.

But each of us throwing our thoughts into the hat is fun. Let's see who wins this race!
 
DuronBurgerMan said:
Exactly. This is something I don't understand. "Your predictions aren't realistic, but mine are! LOLOL!" Come on, this is speculation and bullshitting. If this conversation were in meatspace, it'd be Ford vs. GM kind of argument over a few beers, speculating on what each company is gonna come up with next. It's harmless. If anyone is taking it SERIOUSLY, they are the problem, not the conversation itself.
Click to expand...
Yeah you can speculate away each day or week you just have another one ready until Zen 2 is in retail
 
Has some of you considered the possibility that Zen2 is built around 4-core CCX and 8-core dies again? That is, that Ryzen 3000 for AM4 is up to 8-core again
 
juanrga said:
Has some of you considered the possibility that Zen2 is built around 4-core CCX and 8-core dies again? That is, that Ryzen 3000 for AM4 is up to 8-core again
Click to expand...

I actually still expect it to be a 4 core CCX, but I wouldn't be surprised if it was a 6 core CCX and I would be quite honestly shocked if Zen2 had an 8 core CCX. I think focusing on IPC & clock speeds are more important than adding more cores to the design of the Zen architecture for the time being.

That being said, it would be cool to see a true 6(or 8) core APU.
 
Stoicz said:
I actually still expect it to be a 4 core CCX, but I wouldn't be surprised if it was a 6 core CCX and I would be quite honestly shocked if Zen2 had an 8 core CCX. I think focusing on IPC & clock speeds are more important than adding more cores to the design of the Zen architecture for the time being.

That being said, it would be cool to see a true 6(or 8) core APU.
Click to expand...

I think a 6 core CCX to be the most likely, they may do an 8 core CCX with Zen 3.
 
gigaxtreme1 said:
You know, I have thought about the triangular interconnect to reduce latency. I'm inclined to agree with you architecturewise..
Click to expand...

that would be akin to a "rotary" engine, seems like it absolutely makes sense "to me" this way here all the connections and stuff are as close as possible to each other AND it allows the cores to also be as best cooled by the heatsink/heatspreader as possible, and a "triangle" is likely easy enough to design from the wafer...granted I am not a chip designer, but, that would have solved some issues they supposedly had with TR Gen 1 using only 2 "cores" under the hood with the other 2 being "dummies" for structural stability, a triangle arrangement would likely have been very good in many ways.

better connections, better structural stability (without requiring dummy units) ease of cooling with much more coolers able to handle the load AND the heat spread out a wee bit more.

Interesting ^.^
 
juanrga said:
Has some of you considered the possibility that Zen2 is built around 4-core CCX and 8-core dies again? That is, that Ryzen 3000 for AM4 is up to 8-core again
Click to expand...

Ryzen 1 is only 213mm^2, and AMD has talked extensively about optimizing Zen for power, as opposed to making it a huge core like Bulldozer.

8C Zen 2 would be a tiny die on 7nm. They could do it, but it seems like a waste.
 
  • Like
Reactions: N4CR
like this
brucethemoose said:
Ryzen 1 is only 213mm^2, and AMD has talked extensively about optimizing Zen for power, as opposed to making it a huge core like Bulldozer.

8C Zen would be a tiny die on 7nm. They could do it, but it seems like a waste.
Click to expand...

Porting Zen to 7nm would make a small die ~100mm2. But if Zen2 core is bigger and there is more L3 cache per CCX then the new die would be bigger.
 
juanrga said:
Porting Zen to 7nm would make a small die ~100mm2. But if Zen2 core is bigger and there is more L3 cache per CCX then the new die would be bigger.
Click to expand...

Well yeah, but that's what I'm saying. Based on what they've stated previously, I don't see AMD ballooning Zen into a core bigger than Intel's.
 
dragonstongue said:
that would be akin to a "rotary" engine, seems like it absolutely makes sense "to me" this way here all the connections and stuff are as close as possible to each other AND it allows the cores to also be as best cooled by the heatsink/heatspreader as possible, and a "triangle" is likely easy enough to design from the wafer...granted I am not a chip designer, but, that would have solved some issues they supposedly had with TR Gen 1 using only 2 "cores" under the hood with the other 2 being "dummies" for structural stability, a triangle arrangement would likely have been very good in many ways.

better connections, better structural stability (without requiring dummy units) ease of cooling with much more coolers able to handle the load AND the heat spread out a wee bit more.

Interesting ^.^
Click to expand...

With credit to Wikipedia, I give you the following:

400px-Cray-1_scaling.png


https://en.wikipedia.org/wiki/File:Cray-1_scaling.png
 
If I had to guess, it would say that the performance increase would be similar to Nehalem ==> Sandy Bridge

1.) First major revision to a brand new microarchitecture

2.) Die shrink + clock speed increase
 
_mockingbird said:
If I had to guess, it would say that the performance increase would be similar to Nehalem ==> Sandy Bridge

1.) First major revision to a brand new microarchitecture

2.) Die shrink + clock speed increase
Click to expand...

Zen2 is a revision of Zen. Sandy wasn't a revision of Nehalem, but a new microarchitecture with radical changes such as microp cache, new register file, partitioned L3...

The Pentium 4 was a highly innovative microarchitecture and a radical departure from the previous generation – even more aggres

[...]

But Westmere, the 32nm shrink of Nehalem, will be the last P6 derivative from Intel. After 15 years, the P6 is finally being replaced by a new microarchitecture: Sandy Bridge.

The Sandy Bridge CPU cores can truly be described as a brand new microarchitecture that is a synthesis of the P6 and some elements of the P4. Although Sandy Bridge most strongly resembles the P6 line, it is an utterly different microarchitecture. Nearly every aspect of the core has been substantially improved over the previous generation Nehalem.
Click to expand...

Sandy Bridge is one of the most ambitious and aggressive microprocessors designed at Intel.

[...]

Sandy Bridge is a fundamentally new microarchitecture for Intel. While it outwardly resembles Nehalem and the P6, it is internally far different.
Click to expand...


Taken from RWT. Emphasis mine.
 
juanrga said:
Zen2 is a revision of Zen. Sandy wasn't a revision of Nehalem, but a new microarchitecture with radical changes such as microp cache, new register file, partitioned L3...
Click to expand...

You sure seem to know a lot about Zen 2.

How about you share the details with us, because we love to know too.
 
  • Like
Reactions: N4CR
like this
dragonstongue said:
that would be akin to a "rotary" engine, seems like it absolutely makes sense "to me" this way here all the connections and stuff are as close as possible to each other AND it allows the cores to also be as best cooled by the heatsink/heatspreader as possible, and a "triangle" is likely easy enough to design from the wafer...granted I am not a chip designer, but, that would have solved some issues they supposedly had with TR Gen 1 using only 2 "cores" under the hood with the other 2 being "dummies" for structural stability, a triangle arrangement would likely have been very good in many ways.

better connections, better structural stability (without requiring dummy units) ease of cooling with much more coolers able to handle the load AND the heat spread out a wee bit more.

Interesting ^.^
Click to expand...
The triangle i referred to is Cores arranged as like a brick staggering layment allowing 3 way cache coherency along any 3 cores in a CCX. That led me to believe in 6 core CCXs. Could do it with 8 core per CCX also and have 3 way cache coherency.
 
_mockingbird said:
You sure seem to know a lot about Zen 2.

How about you share the details with us, because we love to know too.
Click to expand...

Zen2 is a revision of Zen, therein the "2" and the "Zen".

Sandy Bridge wasn't Nehalem2 but a "fundamentally new microarchitecture"
 
juanrga said:
Zen2 is a revision of Zen, therein the "2" and the "Zen".

Sandy Bridge wasn't Nehalem2 but a "fundamentally new microarchitecture"
Click to expand...

Where intel moved most of their chipset to on die. Where did that come from?
 
juanrga said:
Zen2 is a revision of Zen, therein the "2" and the "Zen".

Sandy Bridge wasn't Nehalem2 but a "fundamentally new microarchitecture"
Click to expand...

I think mockingbird's point (in his second reply) is that we don't know how different Zen2 really is when compared to Zen, not that Sandy Bridge wasn't considerably different than Nehalem.
 
DuronBurgerMan said:
I think mockingbird's point (in his second reply) is that we don't know how different Zen2 really is when compared to Zen, not that Sandy Bridge wasn't considerably different than Nehalem.
Click to expand...

And I will say again that Zen2 is an evolution of Zen, not an "utterly different microarchitecture". There is no radical changes as adding a uop cache. or redesigning the RF.
 
juanrga said:
And I will say again that Zen2 is an evolution of Zen, not an "utterly different microarchitecture". There is no radical changes as adding a uop cache. or redesigning the RF.
Click to expand...

Let us help you understand the fact we know you dont know that for sure since you dont work for AMD. We can only speculate it's minor changes but we have no evidence one way or the other. It's one thing to say I believe this is the way it will happen and another to just state as you just did without fact. Also why it's called a rumor thread since we know so little about it.
 
juanrga said:
And I will say again that Zen2 is an evolution of Zen, not an "utterly different microarchitecture". There is no radical changes as adding a uop cache. or redesigning the RF.
Click to expand...

Do you realize you are stating things as fact in a thread on conjecture? We do not know the scope of zen vs zen2 changes.

IMO penryn -> nehalem was the as you say "utterly different microarchitecture" compared to westmere -> sandy bridge.
 
juanrga said:
And I will say again that Zen2 is an evolution of Zen, not an "utterly different microarchitecture". There is no radical changes as adding a uop cache. or redesigning the RF.
Click to expand...

I don't expect anything radical either. My expectations are somewhat more conservative than yours. Nonetheless, you do not know this as fact. Neither do I. You do this sleight of hand thing quite often. You say that "Zen2 is". Not 'probably is', not 'might be', not 'likely to be', but is. That's a definitive claim in a thread about a thing we can only speculate about.

Also, let's discuss the other side of the argument... note that the article you cite refers to SB as fundamentally new; that it is getting away from the P6 design lineage. But then the same article tells us that it's a synthesis of the P6 and the P4. SB is definitely different than Nehalem in many respects, and we can do a long semantic dance about that. But the fundamental claim is whether or not SB was more distant from Nehalem than Zen2 will be from Zen, and this cannot be definitively known until Zen2 is released and we can inspect the changes. It probably will be closer to Zen than SB was to Nehalem. That I agree on. But that is only known for certain at this moment by AMD and those working with them directly. It isn't known by YOU.
 
With the launch of 32 core Threadripper, my excitement for Ryzen's future is moving towards Ryzen Mobile.

The 2200U, 2300U, 2500U, and 2700U have slowly, but surely begun to infiltrate the greater mainstream laptop market.

I feel like there is a lot AMD can still do to make the most of their hardware.

I'm really curious what AMD could do in this segment, especially at 7nm.

-6 core 25W-45W part?
-Single core sustained at 4ghz??
-higher CU GPU?

From OEMs, I would love to see some thicker designs that favor faster performance. At least single core performance if they wish to conserve battery life.
 
Neapolitan6th said:
With the launch of 32 core Threadripper, my excitement for Ryzen's future is moving towards Ryzen Mobile.

The 2200U, 2300U, 2500U, and 2700U have slowly, but surely begun to infiltrate the greater mainstream laptop market.

I feel like there is a lot AMD can still do to make the most of their hardware.

I'm really curious what AMD could do in this segment, especially at 7nm.

-6 core 25W-45W part?
-Single core sustained at 4ghz??
-higher CU GPU?

From OEMs, I would love to see some thicker designs that favor faster performance. At least single core performance if they wish to conserve battery life.
Click to expand...

I wouldn't be surprised to see a 6 core come out and improvements on core speeds. GPU side I am just not sure if will see a improvement there or not.
 
if Zen 2 introduces an 8-core CCX due to 7nm power and space savings, I can bet we'll see 8/16 APUs in the 25-40w range and 6/12 in the 15-25w areas. Which means a small 4/8 APU with a 6-8 vega cores could be had in a sub-10w package.

I originally thought that AMD would move to a 6-core CCX, but rumors are suggesting an 8-core CCX. An 8-core CCX on 7nm would still be bigger than our 4-core CCX on 12nm due to architectural additions, so an 8-core CCX would suggest that either A: AMD are going to make bigger pieces of silicon, decreasing their profit margins OR B:) there wouldn't be many improvements to IPC, and all improvements to perf would be from pure clockspeed and core count improvements.

I hope we see IPC improvements, but if clockspeeds get boosted by 10%, we'll see a decent bump regardless. Remember that clockspeed improvements are not guaranteed from a node switch.
 
The older rumor was 6 core CCX and 12 core die for Zen2 on 7nm. So Rome would be 48 core and we would expect a future 6 core APU. Then the core count would increase to 64 core for Milan on 7nm+ (aka Zen3).

Recent rumor claims that roadmaps changed radically. Zen2 is back to 4 core CCX and 8-core die (*). So APU will be 4 core. However, Rome is now 64 core. Milan has same core count but twice more threads (aka SMT4). That is the recent rumor and it is coming from several sources.

(*) I asked why Zen2 would be scaled back to 4-core CCX. Problems with scaling up the CCX? Unknown.

It seems this

86e222de7601c82ff3572dd51e416fb7b54ac820ff13c9abd1859b6df4bd1b3c.gif


is not coming neither for Zen2 nor for Zen3.
 
You must log in or register to reply here.
Back
Top