Question Zen 6 Speculation Thread

[OneEng2]

My guess is
$449 12-core R7
$389 10-core R7
$329 8-core R5
$279 8-core R5 Zen5 (9700X refresh)

Basically a small reduction in $/core, nothing crazy.
This pricing pushes people with less than 300 dollars to buy last gen 8-cores, which is what AMD really wants.

I think that it isn't just AMD that is going to have to do this. Once upon a time, a new process node got you huge performance increases by doubling your transistor budget. A new generation processor was SO much better than a last gen one that the old gen simply wasn't worth much any more.

Today, we look forward to .... 10%? 15-20% optimistically?

Perhaps the ONLY way to make this make sense for an OEM is for them to keep selling the product for a much longer time period after the release of the previous gen ..... and to sell multiple generation chips on older (and less expensive) architectures and nodes for the lower cost.

 

[Josh128]

My guess is
$449 12-core R7
$389 10-core R7
$329 8-core R5
$279 8-core R5 Zen5 (9700X refresh)

This is a pipe dream. Thats pre-COVID 2019 Zen 2 pricing. This is N2 + N3 (allegedly). You actually think they will give you 12 Zen 6 cores and potentially 2 LP cores on the GPUCPUIOD for less than the launch price of 12 core Zen 5, which was on a very mature node and had the smallest gen on gen gains of any Zen? Total ST gains are a mystery, but at the very least, Zen 6 is shaping to be the largest MT gains Zen history.

You can easily add $100 to those Zen 6 SKUs, if we're lucky. The only thing potentially holding back insane pricing might be Intel, but in the position they are in, even they will be severely limited in what they can do there.

 

[MS_AT]

The kicker is what IPC uplift will we see for gaming. It could differ significantly from the geomean of 10%.

For youtube content creators the less the better. If it is 0, think how many videos they could create about missing performance, what went wrong, etc

On a more serious note games rarely achive ipc better than 1.5. I guess what you want to see is improvement to overall performance. And you can increase the latter if the cycle time is shorter while ipc stays unchanged

Actually it would be nice if somebody could measure ipc for games on M4 to see if magic frontend can go past ipc 2 there.

 

[Joe NYC]

Instead, AMD leaned in to the gamerz ecosystem when Strix Halo was rolled out. Not the right move when looking for margin.

No, they did not. In fact, AMD did the opposite.

Targeting gaming laptops would have been having an SKU with:
- full GPU
- single CCD
- V-Cache

That's 3 strikes against something people would recognize as a gaming laptop SKU

 

[Z O X]

I think that it isn't just AMD that is going to have to do this. Once upon a time, a new process node got you huge performance increases by doubling your transistor budget. A new generation processor was SO much better than a last gen one that the old gen simply wasn't worth much any more.

Today, we look forward to .... 10%? 15-20% optimistically?

Why so pessimistic?
The shift is now in creativity instead of brute forcing.
BTW, maybe the biggest IPC shift in history was with Sandy over Nehalem and that was like 20 - 25% (excluding Bulldozer for obvious reasons) ...

 

[511]

Why so pessimistic?
The shift is now in creativity instead of brute forcing.
BTW, maybe the biggest IPC shift in history was with Sandy over Nehalem and that was like 20 - 25% (excluding Bulldozer for obvious reasons) ...

Skymont over Crestmont was insane as for Zem 6 I expect 10-15%

 

[Markfw]

AMD "Zen 6" Processors to Use TSMC 2 nm Node for CCDs, 3 nm for IOD

AMD will transition its next-generation "Zen 6" desktop Ryzen and server EPYC processors to a split-node strategy, placing CPU core complex die (CCD) on TSMC N2P 2 nm node variant and building the I/O die (IOD) on the N3P 3 nm node. At TSMC, volume ramp for the 2 nm node is around the third...

www.techpowerup.com

I have not seen this. 2nm for ccd's and 3 nm for IOD ?

 

[Joe NYC]

Price increase

pricing model from now on generally will be

last gen = normal prices
new gen = premium prices

(doesnt include rehashes, different thing. though there won't be rehashes until zen6 and that's a good thing)

I think you are right. It is quite conceivable that for 8 and 6 core, AMD will continue to sell Zen 5. And limit Zen 6 to 12 and 10 core SKUs. It might be too much waste to sell 12 core CCD as 8 core SKU.

 

[MS_AT]

I have not seen this. 2nm for ccd's and 3 nm for IOD ?

Check the source, it links back to this forum

 

[Markfw]

Check the source, it links back to this forum

Sorry, I still don't see the link anywhere there. It doesn't matter, thanks anyway.

 

[marees]

Sorry, I still don't see the link anywhere there. It doesn't matter, thanks anyway.

This one

N2P

N2P

N3P

 

[Markfw]

delete. Found it.

 

[Josh128]

Why so pessimistic?
The shift is now in creativity instead of brute forcing.
BTW, maybe the biggest IPC shift in history was with Sandy over Nehalem and that was like 20 - 25% (excluding Bulldozer for obvious reasons) ...

Sandy over Nehalem was childs play my man. Conroe over Netburst was like +85% IPC gain. By far the biggest jump in the last 25+ years.

 

[511]

Sandy over Nehalem was childs play my man. Conroe over Netburst was like +85% IPC gain. By far the biggest jump in the last 25+ years.

Wtf 85% that's ridiculous

 

[Markfw]

Wtf 85% that's ridiculous

That why I converted to Intel in 2006 (I think that's when it was)

 

[poke01]

That why I converted to Intel in 2006 (I think that's when it was)

also why Apple moved to Intel then. Nobody else even came close.

 

[Thunder 57]

Sandy over Nehalem was childs play my man. Conroe over Netburst was like +85% IPC gain. By far the biggest jump in the last 25+ years.

Thats bunk, 100%. Maybe 50%.

 

[soresu]

Conroe over Netburst was like +85% IPC gain

Even if it were true (it isn't) that's more like comparing Bulldozer -> Zen1.

P6 -> NetBurst was by no means a win for Intel either.

When you compare Conroe to it's real µArch predecessors - the upgraded P6 based Banias and Dothan (Pentium M) - the gains look a lot less exciting.

 

[adroc_thurston]

the gains look a lot less exciting.

Merom clocked a whole lot higher though.
Even compared to Yonah.

 

[Abwx]

Zen 6 is shaping to be the largest MT gains Zen history.

With 50% more cores and 1.7-1.75x expected better MT it s definitly no, it wont match the Zen/Zen+ to Zen 2 transition.

AMD Ryzen 9 3950X im Test: Benchmarks in Anwendungen und Spielen

AMD Ryzen 9 3950X im Test: Benchmarks in Anwendungen und Spielen / Benchmarks in Anwendungen / Benchmarks in Spielen (Full HD und UHD)

www.computerbase.de

 

[Josh128]

That why I converted to Intel in 2006 (I think that's when it was)

I did the exact same, except I held out for the Wolfdale 45nm refresh , Core 2 Duo E8400. It was around 2008ish for that one. Coming from an Athlon 64 3000+, the E8400 more than doubled the fps in the MAME Atari Seattle and Vegas Voodoo Graphics arcade emulators (San Francisco Rush, Mace: The Dark Age, NFL Blitz, Gauntlet, etc). It was the first chip that ran most of them them at 100% speed (just barely). It was a revolution.

I'd even say that 85% IPC over Netburst was conservative. In many cases, it was 100% or more. It was incredible. It did for CPUs what Voodoo Graphics did for GPUs / video cards.

 

[fastandfurious6]

Conroe over Netburst was like +85% IPC gain

Conroe(2006) -> ivy bridge (2012) = intels golden era, banger after banger

Then completely fell off the cliff 🤣

 

[Thibsie]

K6-III@400 to Duron@650 to Athlon1000 (Tbird), AthlonXP(Thoroughbred) Thinkpad(Yonah), Core2Duo7200 ('cos really Bulldozer was, well....) to PentiumG (5060) to Ryzen2600 then 5600+.

I'm really an AMD guy but when the product stinks, good for the other guy.

 

[Josh128]

Thats bunk, 100%. Maybe 50%.

IPC gains in games like these were over 100% unless Im reading this wrong.

 

[511]

Can't believe Intel did these crazy gains lol they just need to do it one more time.