Question Zen 6 Speculation Thread

[adroc_thurston]

We should set an over/under on clock speed. If it was set at say 6.7GHz I would take the under.

Just stick it at 7G cuz it's a funny number.

 

[511]

Just stick it at 7G cuz it's a funny number.

6.7/6.9 are the funnier numbers I would rather have 6.9 than 7

 

[Thunder 57]

Just stick it at 7G cuz it's a funny number.

At 7GHz anyone with a brain would take the under. At 6.5 I think because of the hype there would be more over than under. So I tried to split the difference. I'm open to suggestions. I think 6.7/6.8 seems reasonable.

 

[StefanR5R]

AMD Zen 6 info and speculations that I found so far. I'm not an expert and probably missed a few.

• Zen 6 developed from scratch [...]

I propose you make a mental note of the site where you found this, and going forward ignore anything which this site claims about unreleased hardware. (It's even safe to ignore what this site says about released hardware.) Ditto for several of the other bullet points.

 

[adroc_thurston]

6.7/6.9 are the funnier numbers I would rather have 6.9 than 7

zoomer meme. gitmo.

At 7GHz anyone with a brain would take the under

lmao.

I think 6.7/6.8 seems reasonable.

6.7 you can get on N3X.
They have something better instead.

My bet is a hair under 7, so probably 6.95 or something.

 

[OneEng2]

I wouldn’t call them optimists

Prior to the current insanity, most people were guessing small IPC increases for Zen 6 over Zen 5 (10-15%). I'm still guessing this is correct since the number of transistors dedicated to raising IPC per core isn't going up that much (the additional transistors are MOSTLY going into more cores per CCD).

AMD will not likely NEED 7Ghz to more than best NVL in MOST everything. As a result, I seriously doubt AMD would release a 7Ghz Zen 6. They would simply clock everything down and get better yields (ie raising their margins).

Can someone explain how AMD could have miraculously pulled out 30% plus IPC improvements without adding a serious number of transistors per core? I am just asking since it appears unlikely on the surface.

 

[Jan Olšan]

Personally, I'm more interested in WHEN we'll get some info rather than bound-to-be off speculating.

 

[Saylick]

My bet is a hair under 7, so probably 6.95 or something.

C'mon, that's a bad bet. You don't think if the engineers told Lisa that they could only must 6.95 GHz that Lisa wouldn't immediately ask, "What? No, get to a clean 7 GHz for marketing purposes"?

 

[adroc_thurston]

C'mon, that's a bad bet. You don't think if the engineers told Lisa that they could only must 6.95 GHz that Lisa wouldn't immediately ask, "What? No, get to a clean 7 GHz for marketing purposes"?

Those are the same people that shipped 5950X at 4.9 so yes.
They're silly goobers.

 

[Saylick]

Those are the same people that shipped 5950X at 4.9 so yes.
They're silly goobers.

The difference between 6.95 GHz and 7.0 GHz is smaller than 4.9 GHz to 5.0 GHz.

 

[adroc_thurston]

(the additional transistors are MOSTLY going into more cores per CCD).

Not really, no.
Dense CCD is mostly L3 now.

AMD will not likely NEED 7Ghz to more than best NVL in MOST everything. As a result, I seriously doubt AMD would release a 7Ghz Zen 6. They would simply clock everything down and get better yields (ie raising their margins).

NVL is a joke for lolcows.
AMD has bigger and meaner beasts to compete against on 1t.

 

[adroc_thurston]

The difference between 6.95 GHz and 7.0 GHz is smaller than 4.9 GHz to 5.0 GHz.

The difference is exactly one bin of speed, so nope!

 

[LightningZ71]

Personally, I'm inclined to believe that, if they can achieve 7Ghz with a decent non-zero percentage of chips, they will release a product that features it. Overall, since they aren't going to have a core count even close to NVL's top SKUs, they're going to push Fmax as hard as they can to wring the most they can out of SMT. I'm FAR more interested in what they can achieve with all-core boost. I want them to get up to6.3- 6.4Ghz all core as I don't think that Intel will be able to get anywhere near that in all-core.

 

[adroc_thurston]

Personally, I'm inclined to believe that, if they can achieve 7Ghz with a decent non-zero percentage of chips, they will release a product that features it. Overall, since they aren't going to have a core count even close to NVL's top SKUs, they're going to push Fmax as hard as they can to wring the most they can out of SMT. I'm FAR more interested in what they can achieve with all-core boost. I want them to get up to6.3- 6.4Ghz all core as I don't think that Intel will be able to get anywhere near that in all-core.

NVL doesn't matter at all.
Again, AMD has bigger and better things to go up against.
There are far, far more competent CPU design houses than Intel.

 

[gdansk]

6250mhz max

 

[MerryCherry]

NVL doesn't matter at all.
Again, AMD has bigger and better things to go up against.
There are far, far more competent CPU design houses than Intel.

ARM titans?

 

[adroc_thurston]

ARM titans?

Yeah.
Apple, QCOM and (duh) ARM itself.

 

[adroc_thurston]

6250mhz max

They can get 9% via core design tweaks.

 

[LightningZ71]

NVL doesn't matter at all.
Again, AMD has bigger and better things to go up against.
There are far, far more competent CPU design houses than Intel.

On PC desktop, they are the competition. On portable (laptops), you have arm entrants that are growing market share, most notably Apple. Server is it's own beast, and I think AMD will be fine there for Zen6.

 

[adroc_thurston]

On PC desktop, they are the competition.

They aren't, 30% gaming perf delta is a gap unsurmountable.

Server is it's own beast, and I think AMD will be fine there for Zen6.

Server AMD won't have any comp ever. We're talking segments they haven't won completely 7 years ago.

 

[poke01]

I mean I don’t get why people think NVL will be hard to beat. Yes it adds new stuff to the ISA but the jank Cove architecture is still present. Until that goes away,Intel isn’t beating AMD.

AMD currently mostly beats Intel while being a node behind, with less L2 and a much smaller core.

 

[adroc_thurston]

mean I don’t get why people think NVL will be hard to beat.

People see "48 cores" and go 'scary'.
In relative terms, it's 2x stuff with a tock and a single node shrink.
AMD gets a tock, two node shrinks, -L slab and 50% more stuff.

AMD currently mostly beats Intel while being a node behind, with less L2 and a much smaller core.

Intel also has packaging advantage.

 

[LightningZ71]

My bigger concern for AMD is Intel getting their (explitive deleted) together with respect to their fabric and L3 performance. Arrow lake doesn't have bad single core throughput. Not amazing, but not bad.

 

[adroc_thurston]

Arrow lake doesn't have bad single core throughput.

It ties AMD while being a node ahead and having generally much more stuff.
It's very bad.
N3 comp slams them both ISO power or at like half the power (hello M5).

 

[LightningZ71]

Customers rarely care about the nuts and bolts. It's what can they afford that gets the job done, or how much performance can they afford. While I agree that Cove is essentially a kitchen sink being thrown at the problem, to the majority of consumers, it doesn't matter. Now, you expand to look at ARM, and take Apple for the very real competition that they are now representing, there's your story. AMD has a ways to go there.