Amd apu zen 2 everything we know so far

So many months ago that AMD started the deployment of the second generation of its APU ZEN 2 processors that we almost reached the end of the cycle, which takes a whole year, and there is an absence that draws attention within the entire portfolio.

This is an OPINION article ! Please, consider its reading if you only want official confirmed information since what is exposed here is not based in any way on it. Thank you.

The new desktop APUs

Unlike its hitherto notebook sisters, who did receive their corresponding updated revision at 14+ (aka 12nm GF) by the end of the year, we have no news of the hypothetical 3xxxG (Zen + 12nm) or 4xxxG (7nm zen2).

The new notebooks with 3xxxU have slightly improved the frequency for both the cpu and the iGPU, which maintains the other characteristics in the passage from Raven to Picasso. A full-fledged Toc.

Will AMD use chiplets for the new APUs or are specific 7nm designs to be unveiled where everything is in a single SoC?

A first explanation for the absence of desktop APUs could be that it is the segment where AMD finds less competition from Intel and, therefore, where it can further extend the existing product and continue selling it to those who are interested in this type of configuration.. Would this involve skipping the second generation of the APU Desktop - 3xxxG?

Another explanation could be related to the need to wait for the 7nm to be ready both in terms of the CPU part (Zen 2) and the iGPU (Navi) and they are released, surely coinciding with the 50th anniversary of the company. Would this imply bringing forward the 3rd generation of the APU Desktop - 4xxxG?

It should be noted that the internal design of the APU has nothing to do with what a CPU would be.

But taking into account that in AMD there was already an important variety when it came to building each chip with different variations in the CCX configurations, with and without MCM's, going one step further towards even more modularity and the use of chiplets should resonate with strength inside our heads when thinking about the evolution of Raven Ridge.

What can we expect internally from upcoming AMD APUs based on the use of TSMC's 7nm cores and chiplets?

In principle, there will be an extensive assorted assortment. To the custom designs of the next video consoles from Microsoft and Sony (sorry but I leave out the one from Atari) we must add the PC ones.

We also have the 'perpetual' fact that ZEN and NAVI have been developed from the beginning with the obligation to meet the needs of video game consoles when the time came for them to effect their generational change (perhaps the last?). Arguably they are a 'commission' from the manufacturers of game consoles.

It would not be unreasonable to suppose that AMD cannot launch / unveil a product 'similar' to the one that will include the game consoles if they have not been previously launched unless you want to anger your partners enough.

With the recent news of the emergence of Google in the world of video games (Google Stadia), it would not be out of place to also give it a thought on whether or not it can make sense that PS5 and XBOX '4' advance their calendars and leave 'now'. to the circuit instead of waiting for its due date… when Stadia was not something to pay attention to.

Could the supposed 'toc' for the desktop versions of Picasso have been 'suspended' after it was learned that Google will launch Stadia this year, which would precipitate Sony and Microsoft adjusting their calendars (producing the necessary official launch of the technology by those who have largely defrayed it) and at the same time allowing 3rd generation APUs to bring their originally scheduled launch (ZEN's third year end) ahead by 6 to 9 months if none of this had happened?

Perhaps AMD has had information on the matter far in advance of us (it is evident that they have) and they have used it conveniently in their favor (which if that were the case, it would also be ours for saving us a continuous 'toc', being able to hit a generational jump (tic).

But well, what we do:

Taking into account the size offered by the AM4 socket and what we know about the 7nm chiplets as far as the CPU is concerned, and the respective of the graphic cores, I have many questions - rhetorical yes… -

It would not make any sense to use more than one CPU chiplet in the APU at least in a really powerful model (remark: from APU - powerful - POWERFUL APU).

Using just one chiplet and leaving the rest of the space to occupy it with iGPU would allow creating an authentic cucumber in AM4 format for solutions that do not want to use discrete graphics. A worthy evolution to what Raven supposed at the time, still having to divide the space by 50/50. Something with which to level up forcefully.

If it is feasible that some APUs carry only 1 chiplet, will they also use I / O DIE or what they are hiding from us is precisely the way in which the Zen 2 could be manufactured in addition to a chiplet + i / o die variant that we all understand applied to Rome and a very high number of computing chiplets vs. a single i / o die maximized and that makes all the sense in the world?

The design of Raven and Picasso with only a CCX and Infinity Fabric modified in terms of interconnection with the addition of the graphic part of the chip, obviously (since they are APUS), could give us an idea of ​​how it would have supposedly been. the logical evolution of the APU when the 7nm were ready… especially since AMD had said absolutely nothing about chiplets.

As the day came when he spoke about the matter (I think I remember it was in the Next Horizon)… well, either the above went down the drain, or we have at least 2 hypothetical variants of how they could mount the new 7nm APUs. With or without chiplets in between.

But the particularities of the APUS and the use of the space available in the DIE make us wonder internally… using an i / o die in the event that there was only one chiplet will not be an easy and above all 'cheap' route occupying a wonderful space that could be filled with more ships if i / o die is not used and it is about interconnecting as it is currently?

Could justifying the use of an I / O die 'special' interconnect the processor CCX (25% of the space) with at least twice the space that houses iGPUs? Or maybe this is either totally impossible (economically unfeasible) or that AMD has simply not decided to develop it with its limited resources and time?

The reality is that for effects there is a screen that hides everything related to APU and 7 nm. The focus of attention is on the impending Zen 2-based cpus and 7nm graphics, with the Radeon VII enduring these months as the opening act of the function that should start with Navi… and it seems that no one notices the APUS chusquerillas, always the seconds.

Also very real is the problem that it would be for AMD itself that too good APUS 'complicate life' for low / mid-range graphics products based on NAVI.

And all this without paying attention to seeing 'what Nvidia says' or does for the part that touches him. Nothing more and nothing less, because with the final concession and launch of those desired by many GTX Turing (RTX for those who want them and also think it is good to pay them), we have some beastly cards with prices, in my opinion, really good. Navi is waiting for you there.

Surely for AMD the best thing is that right now the number of people with concerns about the new Zen desktop APUs is as low as possible and most attend to the previously listed as 'officially' the focus of attention.

And in the meantime: keep selling Ravens… but it never hurts to dig a bit and we'll finally see what ends up happening.

Intel is going to put the batteries in this type of product, it is expected to improve exponentially compared to what they have offered the last decade. Another thing is the time it takes to put the product in the channel.

AMD, in this case, should be recognized as the benchmark to beat.

So, again, the tinkling bells make me think that the next step they are going to take should continue to position them at least one step away from where Intel could 'advance' or they will have lost the battle.

If they allow the competition to get closer to them and offer a product 'similar in performance' to the one we know from Raven and they do not move, or when they do, they present a continuous product (3xxxG) they will show weakness.

Although they are the ugly duckling for gamers, the reality is that these types of chips have gigantic potential in the following years for users who are satisfied with good or very good performance without reaching 'important' resolutions.

Increasing the maximum frequency supported by the memory controller in the zen 2 will be key too, being able to shine what the APUs offer by themselves, or the opposite, if there are no improvements or they are poor.

We recommend reading the best graphics cards on the market

Will they clear up our doubts at upcoming conferences and fairs? Let's hope so.