Question Zen 6 Speculation Thread

[511]

It is not, the 16C chip is sold about 500$ in the chinese retail market, so OEMs get it for much lower than this.

The halo laptops are quite expensive

 

[Abwx]

The halo laptops are quite expensive

That s not the point, you said that the chip was sold for too much by AMD, the fact that some OEMs are overpricing their products has nothing to do with AMD s pricing.

 

[511]

That s not the point, you said that the chip was sold for too much by AMD, the fact that some OEMs are overpricing their products has nothing to do with AMD s pricing.

It's $550 For the Chip which is fine for this silicon considering it's 9950X+4060-4070 Laptop but for us we have to buy Laptops from OEMs unless you are from China also China always get the good stuff.

 

[adroc_thurston]

The halo laptops are quite expensive

It's in the name methinks.

 

[Abwx]

It's $550 For the Chip which is fine for this silicon considering it's 9950X+4060-4070 Laptop but for us we have to buy Laptops from OEMs unless you are from China also China always get the good stuff.

The chip is mainly used in often overpriced SFFs, barely in laptops for the time, eventualy there will be a few lauched during the CES.

Zotac zur Computex 2025: Handheld mit Strix Point, Mini-PC mit Strix Halo, Grafikkarten und mehr

Zotac wird zur Computex 2025 ein komplettes Portfolio aus den Bereichen Grafikkarten, Handhelds, Mini-PCs sowie Enterprise-Lösungen zeigen.

www.computerbase.de

Beelink GTR9 Pro/AI Mini: AMD Strix Halo zieht für 2.000 USD in weitere Mini-PCs ein

In China wurde der Beelink GTR9 Pro/AI Mini mit AMD Strix Halo enthüllt. Die kleinen großen Mini-PCs kosten umgerechnet ab 1.800 US-Dollar.

www.computerbase.de

GMKtec Evo-X2 mit „Strix-Halo“: Ryzen AI Max+ 395 Mini‑PC ab 1.499 Euro vorbestellbar

Der im März vorgestellte GMKtec Evo-X2 mit Ryzen AI Max+ 395 kann ab sofort in zwei Konfigurationen vorbestellt werden.

www.computerbase.de

Erster „Strix‑Halo“-Mini-PC: GMKtec stellt Evo‑X2 mit Ryzen AI Max+ 395 offiziell vor

GMKtec hat mit dem Evo-X2 den ersten Mini-PC auf Basis von AMDs APU-Speerspitze in Form des Ryzen AI Max+ 395 vorgestellt.

www.computerbase.de

 

[adroc_thurston]

The chip is mainly used in often overpriced SFFs

They're priced appropriately. It's in the name.

 

[marees]

The way I think AMD views this market, regarding GPU is in Medusa Point:
- those who don't care will be ok with 8 CU (probably clocked faster on N3P)
- those who do care but don't care about power - will get it with dGPU
- those who do care and also care about power - Will get the Halo CPU product

How do we know it is not 8 wgp ?

 

[adroc_thurston]

How do we know it is not 8 wgp ?

Cuz it's the value part.

 

[ToTTenTranz]

very interesting

https://twitter.com/x/status/1923255707173871868

https://twitter.com/x/status/1923262909477433563

Did I get this right? R9 is basically a 12core CCD with a full R7 SoC serving as IOD?

As for the 8CU iGPU, I guess AMD concluded it's not really worth it to make it larger unless there's LLC or significantly faster RAM.
I am expecting those CUs to clock as high as RDNA4 chips or even higher, though.

 

[Josh128]

Did I get this right? R9 is basically a 12core CCD with a full R7 SoC serving as IOD?

As for the 8CU iGPU, I guess AMD concluded it's not really worth it to make it larger unless there's LLC or significantly faster RAM.
I am expecting those CUs to clock as high as RDNA4 chips or even higher, though.

That appears to be the case, how bizarre.

For the iGPU, 8 CUs at 2nm is going to be a very small piece of silicon unless if they are of similar design to RDNA 4, and will be able to be and in fact MUST be clocked moon in order to not have less graphics perf than Strix Point 16 CU APUs. There is also the possibility that the CUs are of a completely different "UDNA" architecture and may be architecturally equivalent or superior to the 16 CUs of RDNA 3.5 in Strix Point. I think its pretty clear that they wouldnt push out a regression of graphics in APUs in 2 year newer, full process node advancement R9 parts.

 

[adroc_thurston]

8 CUs at 2nm

It's N3p.

 

[jpiniero]

That appears to be the case, how bizarre.

Need the room for AI (at least until Wall Street loses interest). It's going to be expensive enough as it is.

 

[Josh128]

It's N3p.

Ah, interesting. Is N2 only going to be for the Zen 6C EPYC CCDs?

 

[adroc_thurston]

Is N2 only going to be for the Zen 6C EPYC CCDs?

no it's just for non-poorpeople CPUs.
Also, N2p.

 

[LightningZ71]

It's N3p.

So, 8CU of RDNA4 3.5+ on N3P as opposed to 16cu of RDNA4 3.5 on N4P.

That's going to be a regression on the computer side, but due to memory bandwidth limits and no MALL cache, the real difference may not be that great. Looking at how well Kraken's iGPU does vs. The 12cu of Phoenix point, it is reasonable to conclude that real performance will likely fall between Kraken and Strix Point. Given how much actual performance on iGPUs is dependent on FSR performance, it is concerning that it will keep up well enough.

 

[adroc_thurston]

it is concerning that it will keep up well enough.

It won't be close but that's by design.

 

[LightningZ71]

It won't be close but that's by design.

Yeah, this does strike me as a value part. I would ask if Kraken is going to be kept as a long cycle part, but a quick look at Amazon still shows 3xxx parts and a bunch of new built Zen2 laptop parts in the channels, leading me to believe that they are all essentially forever products already.

 

[jpiniero]

Yeah, this does strike me as a value part. I would ask if Kraken is going to be kept as a long cycle part, but a quick look at Amazon still shows 3xxx parts and a bunch of new built Zen2 laptop parts in the channels, leading me to believe that they are all essentially forever products already.

It can have a long tail apparently. The #4 Top Seller on Amazon right now has the N4020, which Intel stopped producing last year.

 

[Josh128]

no it's just for non-poorpeople CPUs.
Also, N2p.

No idea what that means, but OK.

 

[OneEng2]

Ah, interesting. Is N2 only going to be for the Zen 6C EPYC CCDs?

It's the subject of some debate.

Certainly the Zen 6C EPYC CCD's will be N2. I am guessing ~20 cores while others are guessing 32.

Since N3P is likely only about 8-10% less dense than N2, I am guessing that AMD uses N3P for everything EXCEPT Venice Dense (Zen 6C EPYC).

Time will tell.

 

[adroc_thurston]

while others are guessing 32.

it's not a guess.

I am guessing that AMD uses N3P for everything EXCEPT Venice Dense (Zen 6C EPYC).

They're using N2p everywhere except the poverty parts.

No idea what that means, but OK.

Everything above mainstream notebook swimlane is some flavour of N2p.

 

[Win2012R2]

I guess they've learnt their lesson of cutting transistor budget for Zen5 to fit it in N4.

 

[511]

For Zen 5 they used the short Cell Height and more Metal Tracks vs Zen 4 where they used taller cell to achieve clocks but lower Metal Tracks

 

[Josh128]

no it's just for non-poorpeople CPUs.
Also, N2p.

This says N2, not N2P, and its specifically EPYC CCDs. What evidence do you have that it is used on anything other than that?

 

[Win2012R2]

This says N2

N2 is a class, they obviously don't have fully ready N2P right now and even if they did they would not be officially confirming such details so long from release.