Question Zen 6 Speculation Thread

[inquiss]

Double post

 

[Kaffeekenan]

"Please give me non public information because then people will take you seriously" what if someone doesn't care if you think they're credible or not?

Your argument here relies on this someone wanting to prove themselves in your eyes, to leak stuff for ego or clout.. maybe people on random tech forums post for fun rather than because they want the other random forums members to like them.

Sure, some people may not feel the need to proof that they are credible. But then why post these "beating around the bush things"? I find it annoying to read stuff like "noooot how it works" or "hihi you are sooooo close hihi" and things of that nature. Like seriously, lay out some facts or don't say anything at all. Maybe I should start to write things like "hehe if only you guys knew, Zen 6 has a trick up its sleeve, like dayum." and see if that is something forum members enjoy...

 

[poke01]

Adroc did say the Zen6 will be on N2 before AMD said anything.

Most here and on other sites believed it be on N3P/X

 

[Kaffeekenan]

Adroc did say the Zen6 will be on N2 before AMD said anything.

Most here and on other sites believed it be on N3P/X

Pretty good if true!

 

[LightningZ71]

Personally, Adroc has been "close enough" about enough things for me not to care if he's going to post sources. Some of his biggest misses were misunderstanding the relevance of the data he saw or a change of plans by the manufacturers. He's providing the information for free while protecting his source(s) and he shouldn't react to any of you trying to gleen who/what those sources are lest he out them to the people that can cause them problems, maybe even through one of you.

 

[reaperrr3]

Sure, some people may not feel the need to proof that they are credible. But then why post these "beating around the bush things"? I find it annoying to read stuff like "noooot how it works" or "hihi you are sooooo close hihi" and things of that nature. Like seriously, lay out some facts or don't say anything at all. Maybe I should start to write things like "hehe if only you guys knew, Zen 6 has a trick up its sleeve, like dayum." and see if that is something forum members enjoy...

- Some stuff can't really be proven until the info becomes public, either from AMD themselves, some AIB, OEM or TSMC.
- Someone with access to real sources will want to protect those sources by not giving out too much exact insider info, since that could potentially be traced back to the source by AMD.
- Adroc likes to tease. If you find that annoying, just ignore him, nobody forces you to take him seriously or respond to his teasing.

 

[OneEng2]

I'm not here to spoonfeed anyone. It's a petting zoo, not a place for serious discussions.

With the notable exception of you, it seems to me that most here provide useful and insightful comments many times posting links to information that I would otherwise not have known existed.

You don't provide proof, and you don't even show reasoning for your theories.

Since there are NDAs, anything that is "credible" and points further out to the future (not publicly announced by the company) cannot be shared.

But since nothing is completely air tight, some info makes its way out, 2nd or 3rd hand. it is up to you to decide how credible it is.

That is true; however, it seems to me that there have been PLENTY of credible leaks over the last 30 years that I have been following the tech. There is also plenty of publicly available information to make assertions about the future on. Neither of these are what is happening here. One word answers and snide remarks are not useful. It screams "look at me!".

This, so much this. I get that people here want reliable "proof" for things that are top secret but asking for a link is, well, kinda missing the point.

You'll either get a rumour website which ends up in a debate about credibility or you get an official website which likely has its info based on public information which likely came from the company itself.

If it's sufficiently in the future and non-public asking for proof is backwards. It's up to you to work out which of the whispers and hearsay, which can never be publicly confirmed, can be trusted, or at least considered.

Since this is a "Zen 6 speculation" thread, don't you think that rumor website links are relevant? Your post (in a speculation thread btw) seems to imply that only real "proof" should be considered. I would agree that there is precious "proof" at this time other than SOME version of Zen 6 DC CPU's will be on N2.

Adroc did say the Zen6 will be on N2 before AMD said anything.

Most here and on other sites believed it be on N3P/X

Considering the fact that Zen 5 Turin D was on N3E when all other processors for Zen 5 were on N4P, it doesn't seem like you need much of a crystal ball to determine that Venice D will VERY likely be on N2.

It is the rest of the Zen 5 lineup that is in doubt ..... and the focus of this thread's recent comments.

 

[Thibsie]

[rant]
Frankly, lots of people here are reacting as they were kids.

Of course he can't prove a thing or burn his source. What would expect ?

Of course he won't speculate based on this or that track record because... he's been told (good or bad) how things are (planned) *at the moment the news was told*. What would you expect ?

Expecting anything else is childish to say it politely.

Now, about the formal pov, I agree the way he states thing might upset some of us.
If it annoys you, just put him on ignore and get on with your life, really...
[/rant]

 

[adroc_thurston]

You don't provide proof, and you don't even show reasoning for your theories.

No one ever has to.

Considering the fact that Zen 5 Turin D was on N3E when all other processors for Zen 5 were on N4P

They were all N3.
Then N3(b) crashed.

 

[LightningZ71]

As I said a while back, it appeared to me that a lot of the decisions on what we got from Zen5 were based around a part that was intended for a denser node that turned out to be unavailable.

 

[adroc_thurston]

As I said a while back, it appeared to me that a lot of the decisions on what we got from Zen5 were based around a part that was intended for a denser node that turned out to be unavailable.

It wasn't really unavailable, just expensive and wasn't yielding that well for a while.

 

[OneEng2]

No one ever has to.

They were all N3.
Then N3(b) crashed.

Link?

As I said a while back, it appeared to me that a lot of the decisions on what we got from Zen5 were based around a part that was intended for a denser node that turned out to be unavailable.

Possibly; however, Zen 5 ended up besting anything Intel could bring to the table. It also makes sense not to use the highest cost, lowest yield (highest risk) node for your highest volume and lowest margin parts. In other words, I am saying that it would have been a logical business decision vs. being driven by a technical issue.

If anyone has evidence in either direction, it would be relevant as to how we might expect AMD to behave on Zen 6 as well.

It wasn't really unavailable, just expensive and wasn't yielding that well for a while.

The expensive part would not have changed from the early planning stage. Zen 5 was produced on N3E for DC and N3B was ready for Arrow and Lunar Lake, so I don't see it as being "unavailable". If it was yielding poorly, then it would also be yielding poorly (in all likelihood) on Intel's parts which were released 1 month later (Lunar Lake), so I don't see any evidence that your statement is based on fact.

Again, If I am off base here, please post a link.

[rant]
Frankly, lots of people here are reacting as they were kids.

Of course he can't prove a thing or burn his source. What would expect ?

Of course he won't speculate based on this or that track record because... he's been told (good or bad) how things are (planned) *at the moment the news was told*. What would you expect ?

Expecting anything else is childish to say it politely.

Now, about the formal pov, I agree the way he states thing might upset some of us.
If it annoys you, just put him on ignore and get on with your life, really...
[/rant]

You don't have to burn a source in order to make a reasonable speculation or reply to a comment with something more than a one-word-answer.

If he has "been told" something in confidence then I would expect that his information would have been accurate in the past.

FWIW, I am not even saying his speculation is wrong, only that there is plenty of circumstantial evidence suggesting that his POV is wrong. He could easily respond with his own circumstantial evidence to counter anything others say without violating any supposed NDA.

As an example, AMD is currently using both N4P and N3E for Zen 5. Why would we expect them to change their philosophy for Zen 6? Do these reasons make sense? This is the text book definition of "Speculation".

We know the Zen 5 die sizes for both desktop and DC. We know (leaked info and AMD statements) that the IOD is changing a great deal for Zen 6. It appears to be a modular design based on leaks. We know that SP7 supports no more than 16 channels of DDR memory. We know about how much bandwidth Zen 5 used for its high core count DC processors. We also know the strategy AMD ended up using for the nodes it used for different markets on Zen 5.

All of these things suggest:

1) Zen 6 is likely to have similar CCD die sizes as Zen 5.
2) Zen 6 is likely to be fabricated on more than one node for different markets.
3) Zen 6 will likely require (and get) more bandwidth per core in high core count products.
4) Intel has been behind in DC for several years. AMD may not feel the need to "pull out all the stops" at any cost (and profit) in order to continue the massive landslide of market share in DC.

[/SPECULATION]

 

[reaperrr3]

Link?

Why would there be a link?
That would require AMD disclosing these things, which they rarely do, and such decisions aren't exactly put on PPT slides, either.

If it was yielding poorly, then it would also be yielding poorly (in all likelihood) on Intel's parts which were released 1 month later (Lunar Lake), so I don't see any evidence that your statement is based on fact.

Intel might have booked so much N3b capacity so far in advance that the booked wafers were more than enough to satisfy the weak demand for their products even at a "meh" yield rate.
Yield can also mean clock speed variation, plus compared to mature processes, even a 65% yield rate can be considered poor.
Pat himself also hinted at Lunar Lake being a lossy product for them, so there's that.

 

[adroc_thurston]

Pat himself also hinted at Lunar Lake being a lossy product for them, so there's that.

Oh yeah it's definitely not margin accretive, but that's primarily due to pdn and MoP being sizeable cost adders versus a standard platform.

 

[Thunder 57]

Since there are NDAs, anything that is "credible" and points further out to the future (not publicly announced by the company) cannot be shared.

But since nothing is completely air tight, some info makes its way out, 2nd or 3rd hand. it is up to you to decide how credible it is.

By no means would I expect anyone to post something that might out their source(s). But the way I read it was that the info was on an AMD ppt and was in the wild because Adroc said to "try asking people". I read that as ask people to find the link for OneEng2 and just being difficult by not providing a link to publicly abailable information.

Otherwise he is free to speculate just like anyone else. I (and others) may not like they way he posts a lot with short curt responses or never being able to admit when he was wrong but he's free to do so.

 

[Fjodor2001]

Zen6 10950X
8+24+4+4CU, 100+ TOPS
N2
2026Q3
$799

 

[Thunder 57]

Zen6 10950X
8+24+4+4CU, 100+ TOPS
N2
2026Q3
$799

I hope not. For starters wouldn't it be 12 Zen 6 cores not 8? So maybe 12+24+4+4CU.

I think i'd rather see 24+4+4CU though.

 

[LightningZ71]

It wasn't really unavailable, just expensive and wasn't yielding that well for a while.

So fiscally unavailable. If it yields so low and has costs too high, there is no money to be made, so it is effectively unavailable, even if it's possible to use it to create a chip. I'm sure that Intel can make chips using 18a-p, but only in the test lab. It might be available, but your accountants would have you drug out and executed.

 

[LightningZ71]

I hope not. For starters wouldn't it be 12 Zen 6 cores not 8? So maybe 12+24+4+4CU.

I think i'd rather see 24+4+4CU though.

I would absolutely want 12+24 over 12+12. All day, any day. I think we only get 12 core CCDs on desktop/mobile, so I think the discussion is moot.

 

[Thunder 57]

I would absolutely want 12+24 over 12+12. All day, any day. I think we only get 12 core CCDs on desktop/mobile, so I think the discussion is moot.

I was kind of on the fence when I posted that, but I think you are right. 12 + 24 would probably be best in most cases. Especially if Zen 6/7 is moving to a 2MB L2.

 

[adroc_thurston]

Especially if Zen 6/7 is moving to a 2MB L2.

Z6 is 1M L2.

 

[Thunder 57]

Z6 is 1M L2.

I figured as much but some early "leaks" or "rumors" whatever you want to call them seem to suggest Zen 7 is moving to 2MB L2. I don't see it happening but it's too early to really speculate IMHO.

 

[adroc_thurston]

suggest Zen 7 is moving to 2MB L2.

It's configurable, but overall yea.
Zen4-6 all share the same 1M L2 config.

 

[DrMrLordX]

This, so much this. I get that people here want reliable "proof" for things that are top secret but asking for a link is, well, kinda missing the point.

You'll either get a rumour website which ends up in a debate about credibility or you get an official website which likely has its info based on public information which likely came from the company itself.

If it's sufficiently in the future and non-public asking for proof is backwards. It's up to you to work out which of the whispers and hearsay, which can never be publicly confirmed, can be trusted, or at least considered.

Personally, Adroc has been "close enough" about enough things for me not to care if he's going to post sources. Some of his biggest misses were misunderstanding the relevance of the data he saw or a change of plans by the manufacturers. He's providing the information for free while protecting his source(s) and he shouldn't react to any of you trying to gleen who/what those sources are lest he out them to the people that can cause them problems, maybe even through one of you.

Leak or don't leak. If someone is walking a tightrope between revealing partial information without revealing sources, there are more succinct (and less combative) ways to do it than what we're seeing here.

 

[Fjodor2001]

I hope not. For starters wouldn't it be 12 Zen 6 cores not 8? So maybe 12+24+4+4CU.

I think i'd rather see 24+4+4CU though.

The info is from presentation slides. Will not spoonfeed anyone with source, but just trust me on it...