Question Zen 6 Speculation Thread

[Joe NYC]

The only thing getting N3P is the Mobile SKU

Or more precisely, monolithic mobile Zen 6 CPUs.

Which, unit-wise, will be higher number of CPUs than chiplet based mobile, using N2. + chiplet based desktop N2 parts.

 

[511]

I wonder what node they would use for Medusa halo GPU if i had to guess it should be N3P due to the fact that they will use it for dGPU as well.

 

[Joe NYC]

I can’t wait for Zen6, it will shut up the ARM bros for a bit.

I am 100% hopeful of Zen 6 achieving >6GHz. It’s only 400Mhz away. Doable via NanoFlex

Whatever advantages (real or perceived) there were for Arm for hyperscalers, Venice Dense is likely going to erase all of those advantages.

 

[Joe NYC]

Link please.

Page 172 - Question - Zen 6 Speculation Thread

Page 172 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[Joe NYC]

I'd expect $50 ASP bumps for sure.

That sounds reasonable, and it would be a great deal for the users:

8 Zen 5 cores + 96 MB L3
vs.
12 Zen 6 cores + 144MB of L3

 

[Joe NYC]

His secret sponsor is AMD( cause his AMD leaks are pretty accurate but his Intel leaks are hit or miss)

It really would not surprise me if, as part of psychological warfare, AMD leaked its roadmap to MLID, in order to send Intel to a state of panic.

 

[Doug S]

It really would not surprise me if, as part of psychological warfare, AMD leaked its roadmap to MLID, in order to send Intel to a state of panic.

If you wanted your competitor to panic, wouldn't you leak FAKE roadmaps that scare them more than the real plans would?

 

[gdansk]

If you wanted your competitor to panic, wouldn't you leak FAKE roadmaps that scare them more than the real plans would?

Maybe. But it might inspire them to over-deliver to fight a fake roadmap.
I'd focus on making up rumors about competition such that people are always disappointed by them and they always seem to be under-delivering.

But I'm not sure hyping up your competitor is a good idea. I suspect all these misinformation campaigns have too much chance to backfire for them to be considered.

 

[511]

Whatever advantages (real or perceived) there were for Arm for hyperscalers, Venice Dense is likely going to erase all of those advantages.

Along with DMR 🤣🤣.

 

[Io Magnesso]

Amazon US:
9800x3d: $472
7900x3d: $369
14900k: $438
285K:$559

Intel is already charging more for weaker gaming performance. Thinking that Intel will charge less for a CPU with 2 large N2 dies is delusional.

And it is not like Intel will not NEED to charge even higher prices. Intel is still in the middle of a financial downfall, the turnaround is still in the distant future (if ever).

This NVL with big cache will probably well > $700

When did I say Intel would lower the price?
Is that okay?
Are you under an illusion?
Are you crazy?

To begin with, the two Intel CPUs you're using as examples haven't had their prices raised at all. They've always been priced like that. If you said, "They're more expensive than these AMD CPUs, so they need to be lowered in price," I'd understand, but why do you use the word "I'm raising the price"?
Are you living in a fantasy world?

Are you crazy? I'm interested. You're making things up as if I said something I didn't say.

I was just commenting on the price of the 9800X3D in my area. I have not heard the price on Amazon America.

Please stop making strange accusations...
Please...

 

[511]

Amazon US:
9800x3d: $472
7900x3d: $369
14900k: $438
285K:$559

Intel is already charging more for weaker gaming performance. Thinking that Intel will charge less for a CPU with 2 large N2 dies is delusional.

Hello 285K is more than a gaming CPU gaming is not everything

And it is not like Intel will not NEED to charge even higher prices. Intel is still in the middle of a financial downfall, the turnaround is still in the distant future (if ever).

This NVL with big cache will probably well > $700

I can bet you the 24C Zen 6 and 52C NVL will be at least 40-50% more expensive cause of the upgrades

 

[igor_kavinski]

You have to wonder if Nova Lake is to be another "leapfrog" concocted by Gelsinger that will fizzle out just like like all his other "leapfrogs" have fizzled out so far.

Leapfrog. Concocted. Gelsinger. Fizzled out.

Yep. All of that checks out!

 

[Io Magnesso]

The capacity is limited only by what you place orders for. There was, maybe a 1+ year long capacity shortfall for CoWoS type packaging, but that is probably easing too. So there is nothing in the way of AMD placing orders for greater capacity, and TSMC completing that capacity.

If there is a 9600x3d that is universally available (not limited to only certain retailers like 5600x3d was) and AMD prices it between $250 and $300, then AMD will double the shipments of x3d parts.

Which indicates that AMD probably already ordered that capacity increase...

However, I don't know if AMD will use its capacity for consumer products.
TSMC is also working hard to expand the post-process line, but it still puts a lot of load on it.
There are many strong rivals competing for the production volume…

I think that AMD will also use that capacity mainly for servers/AI products…

 

[Io Magnesso]

I think this time around, AMD will be definitely in top 3 users of the most advanced node (N2). Products that will use N2:
- all Venice generations server CPUs
- Mi400 generation of datacenter GPUs
- premium client CPUs

Those numbers will start to add up, since AMD will be close to 50% server CPU volume, and most likely, AMD datecenter GPU will be on a strong upswing with Mi400 generation (from low single digit market share).

And there will likely be a big shift in the premium client CPUs in AMD direction...

Mi400 is still in the future, right?
because of its nature, the die size is large.
It will not be made from the first year of production of N2, probably from the second year when the volume has increased.

 

[Joe NYC]

When did I say Intel doesn't need to Mark-up? Are you under an illusion?
Are you crazy?

To begin with, the two Intel CPUs you're using as examples haven't had their prices raised at all. They've always been priced like that. If you said, "They're more expensive than these AMD CPUs, so they need to be lowered in price," I'd understand, but why do you use the word "I'm raising the price"?
Are you living in a fantasy world?

Are you crazy? I'm interested. You're making things up as if I said something I didn't say.

I was just commenting on the price of the 9800X3D in my area. I have not heard the price on Amazon America.

Please stop making strange accusations...
Please...

If you follow the thread, @igor_kavinski raised a possibility of greater performance and lower prices of gaming CPUs in the future, when Nova Lake with large L3 ships.

My point was that Intel is now charging more, for lower gaming performance, why would Intel charge less for better performance in the future? It does not make sense to have that expectation.

 

[Io Magnesso]

When did I say Intel would lower the price?
Is that okay?
Are you under an illusion?
Are you crazy?

To begin with, the two Intel CPUs you're using as examples haven't had their prices raised at all. They've always been priced like that. If you said, "They're more expensive than these AMD CPUs, so they need to be lowered in price," I'd understand, but why do you use the word "I'm raising the price"?
Are you living in a fantasy world?

Are you crazy? I'm interested. You're making things up as if I said something I didn't say.

I was just commenting on the price of the 9800X3D in my area. I have not heard the price on Amazon America.

Please stop making strange accusations...
Please...

Please give up, really...

 

[igor_kavinski]

It does not make sense to have that expectation.

Yeah, unless Intel decides to go kamikaze and put a massive dent in AMD's gaming marketshare by pricing their bLLC chips too aggressively.

 

[Kepler_L2]

Along with DMR 🤣🤣.

Why would DMR be anything special if Panther/Coyote Cove in NVL looks like ass?

 

[Io Magnesso]

If you follow the thread, @igor_kavinski raised a possibility of greater performance and lower prices of gaming CPUs in the future, when Nova Lake with large L3 ships.

My point was that Intel is now charging more, for lower gaming performance, why would Intel charge less for better performance in the future? It does not make sense to have that expectation.

You can't read…?
So why don't you understand what I'm saying…？

The comments about Nova Lake were made by @igor_kavinski and I was talking about the 9800X3D.
Seriously, why did you reply to my comment instead of replying to @igor_kavinski's comment?
Stop making it seem like I said what he said.
Should I have replied to his comments

 

[Io Magnesso]

Why would DMR be anything special if Panther/Coyote Cove in NVL looks like ass?

Wouldn't it be better if we could keep the ARMs in check for the time being?
If that's the case, even if it's not particularly good, it's necessary to hit it
At the very least, Clearwater Forest is likely to have a good fight with the ARMs if all goes well.

 

[Joe NYC]

Hello 285K is more than a gaming CPU gaming is not everything

I can bet you the 24C Zen 6 and 52C NVL will be at least 40-50% more expensive cause of the upgrades

I wonder what kind of use case will there be for Zen 6 with 24 cores. Because the 12 core Zen 6 is going to eat a lot of the > 8 core market by offering 12 cores.

If there is not some new use case, I don't see the volumes being high, and if the volume is low, who cares about the pricing? It is just symbolic.

 

[igor_kavinski]

I wonder what kind of use case will there be for Zen 6 with 24 cores.

Heavy multitaskers. Those who like to browse and encode or do other CPU intensive background tasks. More cores can easily be "felt" in such cases.

 

[Joe NYC]

However, I don't know if AMD will use its capacity for consumer products.
TSMC is also working hard to expand the post-process line, but it still puts a lot of load on it.
There are many strong rivals competing for the production volume…

I think that AMD will also use that capacity mainly for servers/AI products…

So this would be advanced packaging we are discussing. Even more specifically, 3D hybrid bond packaging.

As of now, AMD is not offering Turin with V-Cache, so the demand from there is zero. There is demand for Mi300/325/355, but since AMD missed some of its expectations in H2 2024, for shipments of these products, these products are not crowding out the client.

Going to Zen 6, the Zen 6 Dense CCD, with 32 cores already has a generous amount of L3 (2x of Genoa Dense and Turin Dense cores). So I doubt AMD will be offering V-Cache for these.

From MLID leaks, the 12 core CCD of full Zen 6 cores will be all V-Cache capable, so AMD can offer it in server, desktop and mobile. Whether there is in fact Venice with 12 core CCDs with V-Cache - remains to be seen.

After this process of elimination, the only product that can eat up the 3D packaging capacity is Mi355. It's shipment will likely be up, but not by insane margin...

In other words, if AMD increased its orders, for greater 3D packaging capacity, it is unlikely that in the short term, that server division would steal it from client division.

 

[Markfw]

In other words, if AMD increased its orders, for greater 3D packaging capacity, it is unlikely that in the short term, that server division would steal it from client division.

I am pretty sure that the server division will always be a number one goal for AMD. It is what has taken them where they are today. And the client CPUs use the server chiplets, the CPU ones. The IOD I am sure is quite different.

 

[Joe NYC]

Mi400 is still in the future, right?
because of its nature, the die size is large.
It will not be made from the first year of production of N2, probably from the second year when the volume has increased.

Mi400 is for the future, but, interestingly, in a leaked TSMC document, there will be N2 Mi400 wafers being processed in late 2025. Meaning, among the first N2 wafers being processed.

The production cycle of AMD "Mi" products is long. From 1st wafers to final product. I forgot the exact time but somewhere between 6 months and 1 year. So, to sell Mi400 in 2026, its wafers have to go into the oven long time before the product shipments.