Discussion Beyond zen 6

[jpiniero]

There's also just slotting in the APUs too.

 

[marees]

There's also just slotting in the APUs too.

I don't think the process nodes have been set in stone

grimlock point 4 (15watt mendocino replacement) could be samsung 2nm too, I think

 

[Thunder 57]

Client stomached only one cost per xtor bump so far (N5).
Now it'll have to stomach 3 in a row.
You will pay more and you will like it.

Are you really trying to bait me? I'll just wait for you to be wrong again. Too bad it's a long wait.

 

[jpiniero]

Are you really trying to bait me? I'll just wait for you to be wrong again. Too bad it's a long wait.

From AMD's own financial statements, the Client ASP increase for the entire year was 31%. That was this year... who knows what further increases it'll be when Zen 7 comes out.

 

[adroc_thurston]

From AMD's own financial statements, the Client ASP increase for the entire year was 31%. That was this year... who knows what further increases it'll be when Zen 7 comes out.

no trust me they're not gonna do ASP hikes because...
...because I said so,

 

[Thunder 57]

From AMD's own financial statements, the Client ASP increase for the entire year was 31%. That was this year... who knows what further increases it'll be when Zen 7 comes out.

Not sure where it's coming from. Prices have been steady over the past year.

no trust me they're not gonna do ASP hikes because...
...because I said so,

That's rich coming from you. Everything you say is "trust me bro" thrown in with a bit of "pay me" every now and then.

But I'm out before this gets any worse. I'll just leave it with I expect prices will go up due to it being 3+ years away and the larger CCD. But not to the levels you are claiming (and claimed about Zen 5).

 

[Fjodor2001]

From AMD's own financial statements, the Client ASP increase for the entire year was 31%.

But that does not say how the price for the different SKUs has changed. It's not like the 9800X3D has gone up in price for example.

Could be that people are just buying higher performing SKUs which cost more. E.g. going for 9950X3D instead of 9800X3D.

 

[Thunder 57]

But that does not say how the price for the different SKUs has changed. It's not like the 9800X3D has gone up in price for example.

Could be that people are just buying higher performing SKUs which cost more. E.g. going for 9950X3D instead of 9800X3D.

I actually looked up that model when I saw the ASP thing.

AMD RYZEN 7 9800X3D 8-Core, 16-Thread Desktop Processor

Amazon price tracking & price history for AMD RYZEN 7 9800X3D 8-Core, 16-Thread Desktop Processor

camelcamelcamel.com

Pretty much flat over the past year.

 

[jpiniero]

But that does not say how the price for the different SKUs has changed. It's not like the 9800X3D has gone up in price for example.

Could be that people are just buying higher performing SKUs which cost more. E.g. going for 9950X3D instead of 9800X3D.

It's all of the Client CPUs combined, desktop and laptop.

 

[Fjodor2001]

It's all of the Client CPUs combined, desktop and laptop.

So...?

 

[ETI4711]

Not sure where it's coming from. Prices have been steady over the past year.

The ASP rises even if the proportion of expensive products in sales figures increases.

I don't understand all the discussion about Zen 7 prices. AMD can set whatever price it wants. However, that doesn't mean that many people will buy at the price AMD wants, and if the sales figures aren't right... AMD offered discounts on Zen 4 pretty quickly.

And that's why the price isn't determined by the costs, but by what buyers are willing to pay for it. It's not rocket science, but rather the basics that every salesperson knows.

 

[yuri69]

SKU prices are sometimes set like weeks prior launch and Zen 7 is years away... oh

 

[adroc_thurston]

SKU prices are sometimes set like weeks prior launch and Zen 7 is years away... oh

This has nothing to do with individual SKU pricing and everything with semis economics being dung.
Intel's the same camp (worse, really, bLLC is a 2D slab of N2) anyway.

 

[511]

This has nothing to do with individual SKU pricing and everything with semis economics being dung.
Intel's the same camp (worse, really, bLLC is a 2D slab of N2) anyway.

They are going to have the margin for packing and the rest of the die but that 2D Slab is worrying if they mess it up also on this cost per xtor is going up as process shrinks cause you need to do more for less gains

 

[Joe NYC]

Yeah man, whatever you say.
How's that 5.1 on PTL doing? Why are the margins 35% on 18A ramp?

N2 isn't "mega expensive", it just costs more.
2.5D substrate also costs more.
New cIOD (on a newer node) also costs more. It all adds up.
Moreso for Zen7 which is A14.

Is V-Cache die staying of N6 or also moving to newer node?

 

[Joe NYC]

The 32c will exist?

16c for full core Florence. 8 * 16 = 128c

High end desktop (with V-Cache) will also get it, and apparently there will be an 8 core die for the low end.

 

[eek2121]

ehhh.
Zen6 alone is 2 nodes of money + packaging cost addr + new moremoney cIOD.
Zen7 compounds on that with A14 so prices will just go up and up.

It's the reason most laptop segments are becoming increasingly hard to live in. DIY can tolerate mass ASP hikes. Laptop can't; OEM opmargins are razor thin as is.

Binning will help keep the price under control. So will competitive pressure.

AMD will charge as much as they think they can get away with, however.

If Intel Nova Lake is unexpectedly competitive, we won’t see much change.

What some folks need to prepare for is the top dawg chip 24c/48t) creating a new price tier if Intel is uncompetitive or they don’t charge premium for the rumored 16+32 part.

…and yes, people are willing to pay big bucks for a top tier chip. Just like they do for GPUs. I imagine anything under $1,500 would sell, provided the performance is there.

Past history has shown that most folks buying top tier aren’t price sensitive.

 

[adroc_thurston]

Is V-Cache die staying of N6 or also moving to newer node?

idk.

Binning will help keep the price under control

nope, TSM is the king of yield.

So will competitive pressure.

if only it existed.

 

[Fjodor2001]

High end desktop (with V-Cache) will also get it, and apparently there will be an 8 core die for the low end.

Apparently based on what?

 

[Fjodor2001]

This has nothing to do with individual SKU pricing and everything with semis economics being dung.
Intel's the same camp (worse, really, bLLC is a 2D slab of N2) anyway.

The semis can invent whatever price, but if too high consumers will not pay. They will stay on much cheaper previous gen instead. Most will not pay 100% more (based on your prediction) to get approx 10% more perf.

Things may be different for the AI Datacenter bubble companies with unlimited wallets, until the bubble pops.

 

[Joe NYC]

Apparently based on what?

MLID had some slides that showed existence of 8c chiplet. Once it exists, it will be used.

As a comparison, Zen 6 does not have 8c standalone chiplet.

 

[adroc_thurston]

The semis can invent whatever price, but if too high consumers will not pay

Of course they will pay.
GPU market already survived a few rounds of ASP hikes and it's doing better than ever.

Most will not pay 100% more (based on your prediction) to get approx 10% more perf.

Yes they do hence why 9800X3D owns the DIY space.

 

[Abwx]

From AMD's own financial statements, the Client ASP increase for the entire year was 31%. That was this year... who knows what further increases it'll be when Zen 7 comes out.

That s mainly due to the 9800X3D, since it launched late in 2024 it had low financial footprint for 2024 but a full year of sales in 2025 was enough to increase the global ASP y/y.

 

[eek2121]

The semis can invent whatever price, but if too high consumers will not pay. They will stay on much cheaper previous gen instead. Most will not pay 100% more (based on your prediction) to get approx 10% more perf.

Things may be different for the AI Datacenter bubble companies with unlimited wallets, until the bubble pops.

Do not interpret your own financial comfort zones to be universal. Many of us will happily pay as much as needed for high end performance.

 

[Doug S]

Is V-Cache die staying of N6 or also moving to newer node?

There's not really that much benefit to going smaller in terms of SRAM cell density. Might help latency a bit, but shaving a few cycles of latency off an L3 isn't that big of a win.

I suppose the price difference between N6 and N4 wafers isn't all that large though. So they might but it isn't going to make all that much difference either way.