Question Zen 6 Speculation Thread

[Jan Olšan]

Mobile is to be expected, desktop would be amazing but it is a slower market to adapt. DIMMs will last yet another generation before finally dying an overdue death, hopefully along with ATX.
Maybe some NUCs might adopt CAMM early, could be neat.

No! Anything trying to replace ATX is going to be worse, just look at the mess that cases trying to reinvent the ATX layout do, or the mess of proprietary hardware a la Apple.

ATX tower concept is generally the most space efficient approach, it's a pity large gaming ATX boxes waste all that completely with galons of cable routing space that adds more cable mess than there was before this idea and so on.

The real improvement would be agreeing on a size-reduces tower-like factor (doesn't need to stand vertically of course).

+ SFX PSUs above CPU position - can have a separation metal sheet if the hot air bothers you, but it's going to be sucked out by exhaust fan behind CPU cooler anyway
+ The position is standard so that cables don't need to be long and can go through case - perhaps via some shaping pieces in case next to the mobo's right edge
+ 4-slot mATX boards as standard instead of fullATX - fullATX cards and boards should be specialty.
+ the bottom case floor under the PCIe slots is always perforated so that GPUs can efficiently draw air from outside. Foam spacers to lead the air could easily be added between the GPU and the case floor.
Components like add-in-cards or 2.5/2.5 drives can easily be kept compatible with legacy ATX cases.

This setup can be smaller than lots of modern "ITX" builds with their mess of PCIe risers and cable management space.

 

[adroc_thurston]

Anything trying to replace ATX is going to be worse

no.

or the mess of proprietary hardware a la Apple.

You can copy the Series X windcatcher as the baseline design.

 

[marees]

What are the proposed zen 6 configurations ?

& What is the difference between CCD & CCX ?

 

[adroc_thurston]

What are the proposed zen 6 configurations ?

there are a few.

What is the difference between CCD & CCX ?

CCD is the die, CCX is the core cluster.

 

[Fjodor2001]

Not much is known about Zen6 at this point.

Rumors so far is mostly this:

Source: @InstLatX64:



Source: @Kepler_L2:



Overall source:

AMD Zen 6 To Feature Three CCD Configurations: 8, 16, & Up To 32 Cores, Zen 5C Packs 16 Cores In Single CCX

AMD's next-gen Zen 5 and Zen 6 core configurations have allegedly been revealed with the latter featuring up to 32 cores per CCD.

wccftech.com

 

[Fjodor2001]

If the Intel ArrowLake Refresh rumor is true, it'll be 8P+32E cores.

So based on the info in the previous post, we could have a direct showdown where AMD Zen6 on DT would provide:
1x8C Zen6 + 1x32C Zen6C

I.e. same amount of P & E cores for both Intel and AMD. Optionally AMD would have to go for e.g. 2x16C Zen6 instead. Either way AMD will have to bump the core count on DT, or they'll lose the MT perf battle on next gen DT CPUs.

 

[adroc_thurston]

So based on the info in the previous post, we could have a direct showdown where AMD Zen6 on DT would provide:
1x8C Zen6 + 1x32C Zen6C

I really don't think anyone needs 40 firestorms on a 2ch DDR5 dingus.

same amount of P & E cores for both Intel and AMD

Dense isn't a LITTLE, it's just the same thing slower for gods sake.

 

[Fjodor2001]

I really don't think anyone needs 40 firestorms on a 2ch DDR5 dingus.

Firestorms? Also, might be DDR6. Need moar cores to gobble up all that juicy bandwidth. And DDR6 is not strictly needed anyway, 8+32 will do fine on high spec DDR5.

Dense isn't a LITTLE, it's just the same thing slower for gods sake.

Call it what you like. It's 8 max perf + 32 lower perf cores. Dense/little/whatever doesn't matter. Point is that it'll mean better MT perf, better perf/W, and lower cost per MT perf. AMD will have to counter it somehow, or give up the MT on DT segment to Intel.

 

[CouncilorIrissa]

Firestorms? Also, might be DDR6. Need moar cores to gobble up all that juicy bandwidth. And DDR6 is not strictly needed anyway, 8+32 will do fine on high spec DDR5.

DDR6 isn't making it to client IIRC.

Also, Intel's 8+32 setup lacks SMT, so you don't really need 40 be competitive in nT.

 

[Fjodor2001]

DDR6 isn't making it to client IIRC.

Ok, is there any source confirming this? What timelines are we looking at?

And anyway, DDR5 max perf will have improved by the time when Zen6 / ArrowLakeRefresh is released anyway.

Also, Intel's 8+32 setup lacks SMT, so you don't really need 40 be competitive in nT.

Fair enough. But still, AMD Zen6 16C will not be enough to counter Intel ArrowLakeRefresh 8P+32E in MT perf anyway.

So if max AMD Zen6 DT SKU will not be 8CxZen6 + 32CxZen6C, what will AMD use instead? E.g. 2x16C Zen6?

 

[adroc_thurston]

Firestorms

Yea.

Need moar cores to gobble up all that juicy bandwidth.

You really don't.
You can always go on twitter and ask the y-cruncher guy (Alex) about that.

8+32 will do fine on high spec DDR5.

You need help.

Point is that it'll mean better MT perf, better perf/W, and lower cost per MT perf

It'll be higher cost to win cinememe race. Useless. Dismissed.

 

[Fjodor2001]

Yea.

?

You really don't.
You can always go on twitter and ask the y-cruncher guy (Alex) about that.

So what will you use all that DDR6 bandwidth on then, if you think it'll only be 16C Zen6 for AMD? Will Zen6 consume 2x memory bandwidth compared to Zen5 per core?

Intel apparently thinks differently, and thinks 8+32 ArrowLakeRefresh will be well fed with DDR6, or even speed bumped DDR5 if that's what will be available at time of release.

You need help.

So what about Intel Arrow Lake Refresh 8+32 then?

It'll be higher cost to win cinememe race. Useless. Dismissed.

So why will Intel be releasing 8+32 ArrowLakeRefresh then?

And regardless, those that need max MT perf on DT at lowest cost and highest perf/W will go Intel. Unless AMD counters. Are you suggesting AMD will give up this market segment?

 

[adroc_thurston]

?

It's fire and storm. Something.

So what will you use all that DDR6 bandwidth on then,

Client isn't getting DDR6.

Will Zen6 consume 2x memory bandwidth compared to Zen5 per core?

Zen5 already consumes that.

Intel apparently thinks differently

They don't think at all.

and thinks 8+32 ArrowLakeRefresh will be well fed with DDR6

It's the exact same platform. What DDR6?

So what about Intel Arrow Lake Refresh 8+32 then?

AMD doesn't care about Intel not one bit.
They no longer exist.

So why will Intel be releasing 8+32 ArrowLakeRefresh then?

Because they're desperate to win at least a single benchmark.
pretty simple?

 

[StefanR5R]

Dense isn't a LITTLE, it's just the same thing slower for gods sake.

Call it what you like. It's 8 max perf + 32 lower perf cores.

– Zen 4 and Zen 4c have the same iso-clock performance.
– For certain power budgets, Zen 4c have a little higher iso-power performance than Zen 4.
(If the workload is not last-level cache size constrained.)

(Remains to be seen if Zen 5(c) and/or Zen 6(c) stick with this scheme, but so far there was no hint that they wouldn't.)

 

[Kepler_L2]

So why will Intel be releasing 8+32 ArrowLakeRefresh then?

They aren't.

 

[Mahboi]

Client isn't getting DDR6.

We're all off to LPDDR6?

 

[Fjodor2001]

It's fire and storm. Something.

?

Client isn't getting DDR6.

Timelines. Proofs.

Zen5 already consumes that.

You mean it's saturated, and CPU perf cannot be improved over 16C Zen5 DT unless memory bandwidth is improved beyond speed bumped DDR5? So 16C Zen6 will be futile on DDR5. Source needed in that case.

They don't think at all.

So Intel is stupid, and you are smarter than them? Intel is not capable of judging the market for ArrowLakeRefresh 8P+32E?

It's the exact same platform. What DDR6?

Source needed.

AMD doesn't care about Intel not one bit.
They no longer exist.

You checked AMD vs Intel CPU DT market share lately?

Because they're desperate to win at least a single benchmark.
pretty simple?

So you are admitting AMD will lose the MT perf crown to Intel?

Meaning for those looking for max MT perf on DT, at lowest cost, and highest perf/W, Intel will be the best option from ArrowLakeRefresh / Zen6 onwards?

 

[Fjodor2001]

They aren't.

Source? And if not, what will Intel be releasing on DT after regular ArrowLake instead (and at what core count specs)?

 

[adroc_thurston]

We're all off to LPDDR6?

Obviously.

?

Yeah, pretty hot.

Timelines. Proofs.

the what.
DDR6 is optimistically H2'26.

You mean it's saturated, and CPU perf cannot be improved over 16C Zen5 DT unless memory bandwidth is improved beyond speed bumped DDR5?

Sort of.

Source needed in that case.

I've already told you whom to ask.

So Intel is stupid, and you are smarter than them?

Pretty much everyone is smarter than them.
It's a company that failed to reconcile ISA deltas between Core and Atom in any sane timeframe.

Source needed.

you can find the docs online.

You checked AMD vs Intel CPU DT market share lately?

Yeah, AMD dominates DIY.

So you are admitting AMD will lose the MT perf crown to Intel?

Cinebench? hopefully.

Meaning for those looking for max MT perf on DT, at lowest cost, and highest perf/W, Intel will be the best option from ArrowLakeRefresh / Zen6 onwards?

I don't mean anything.

Source? And if not, what will Intel be releasing on DT after regular ArrowLake instead (and at what core count specs)?

Begging for sources does nothing.
You're a beggar. You can't choose.

 

[uzzi38]

Source? And if not, what will Intel be releasing on DT after regular ArrowLake instead (and at what core count specs)?

Arrow Lake again with the same core count. After that will be something new.

8+32 ARL went the way of my childhood hopes and dreams!

(Dead in a ditch somewhere)

 

[Mahboi]

the what.
DDR6 is optimistically H2'26.

Sort of.

Hold the phone, that means that Z6 is going to be memory bottlenecked from the start???
At least on client?

 

[adroc_thurston]

Hold the phone, that means that Z6 is going to be memory bottlenecked from the start???

In y-cruncher? yea.

 

[Mahboi]

Huh.
So they don't move off AM5 just because DDR6/LPDDR6 is gonna come too late?

 

[uzzi38]

On a serious note though in a lot of workloads Zen 5 is going to be memory bandwidth bound. Zen 4 already is in some of the most rigourous workloads, and even in some games etc you can see benefits to more memory bandwidth (see: launch day Starfield).

Even just based off of information in the public domain Zen 5 is a much bigger core but likely to be a very similar memory setup. Bandwidth bottlenecks are practically ensured. Even if Zen 6 were to switch to a new platform with much more robust I/O, it seems unlikely it would provide enough bandwidth to feed 40 cores, even if 8+32. That's over double the existing core count. Even if you want to break them down by clocks and assume 8 Zen 6 cores running at 4.5GHz and 32 Zen 6 cores running at 3.2GHz (which are likely conservative values for both), that's still going to need vastly more memory bandwidth than the current 16 cores running at ~5GHz (which is likely an overestimate if anything).

 

[Fjodor2001]

Yeah, pretty hot.

?

the what.
DDR6 is optimistically H2'26.

Some say 2025/2026. And Zen6 and ArrowLakeRefresh is when?

Anyway, speed bumped DDR5 will be sufficient for above 16C.

Sort of.

Clarification and proof needed, for why you think CPU perf cannot be improved beyond Zen5 16C without DDR6.

Essentially you're saying 16C+ Zen6 will be pointless over Zen5 at 16C, since 16C Zen5 is already saturated on DDR5.

you can find the docs online.

Link?

Yeah, AMD dominates DIY.

DIY is not representative of DT market share.

Cinebench? hopefully.

No, generally. AMD will lose MT crown to Intel on DT?

I don't mean anything.

Well, then I'll repeat:

"for those looking for max MT perf on DT, at lowest cost, and highest perf/W, Intel will be the best option from ArrowLakeRefresh / Zen6 onwards?"

Yes/No? And why?

Begging for sources does nothing.
You're a beggar. You can't choose.

So you have no source, meaning your claim is moot.