Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

[DisEnchantment]

Speculate at will

 

[poke01]

https://videocardz.com/newz/asus-presents-proart-gopro-edition-with-amd-ryzen-ai-max-395-strix-halo

A new laptop with Strix halo

 

[regen1]

https://www.amd.com/en/products/processors/desktops/ryzen/9000-series/amd-ryzen-7-9850x3d.html

Ryzen 7 9850X3D

 

[adroc_thurston]

9950X3D2 bros...
WAKE ME UP INSIDE

 

[inquiss]

Since it looks like 9950X3D2 now will happen after all, what conclusions can we draw from that? One is that it signals that we can expect games to make use of up to 16C/32T going forward. Because why otherwise include X3D cache for more than 8C/16T.

We can conclude that it's a SKU that doesn't really makes sense

 

[adroc_thurston]

Since it looks like 9950X3D2 now will happen after all, what conclusions can we draw from that? One is that it signals that we can expect games to make use of up to 16C/32T going forward. Because why otherwise include X3D cache for more than 8C/16T.

Bros...
Games are definitely using 16c now

 

[RTX2080]

This is also my thought. Not everything exist in lab could become a real product that you could buy. Same goes for Intel and other manufacturers.

https://twitter.com/x/status/2008443309438480519

Latest rumor is not only 9850X3D but also there would be something like R7-9750X and R5-9650X on the way, I'm not sure. Better than nothing/Nothingburger. Move on, all hopes and bets shall be on Zen6.

 

[lightmanek]

Happy with these announcements on the CPU side ...


My 9950X3D can live another day 😅

 

[Joe NYC]

This is also my thought. Not everything exist in lab could become a real product that you could buy. Same goes for Intel and other manufacturers.

https://twitter.com/x/status/2008443309438480519

View attachment 136293



Latest rumor is not only 9850X3D but also there would be something like R7-9750X and R5-9650X on the way, I'm not sure. Better than nothing/Nothingburger. Move on, all hopes and bets shall be on Zen6.

That would indicate possibly new stepping, if AMD can raise clock speed by ~200 MHz across the board.

Why put all the effort into it if Zen 6 is imminent? Maybe when 12 core Zen 6 is on the market, Zen 5 will remain in AMD line up for lower core count processors.

 

[ToTTenTranz]

This is also my thought. Not everything exist in lab could become a real product that you could buy. Same goes for Intel and other manufacturers.

https://twitter.com/x/status/2008443309438480519

View attachment 136293



Latest rumor is not only 9850X3D but also there would be something like R7-9750X and R5-9650X on the way, I'm not sure. Better than nothing/Nothingburger. Move on, all hopes and bets shall be on Zen6.

MLID says a bunch of people from the press were briefed about the 9950X3D2. Could be cancelled, could be launched at a later date.

Why put all the effort into it if Zen 6 is imminent? Maybe when 12 core Zen 6 is on the market, Zen 5 will remain in AMD line up for lower core count processors.

The rationale behind the 9950X3D2 and the 9850X3D was precisely the suggestion that client Zen6 is getting a delay into 2027.

 

[Joe NYC]

The rationale behind the 9950X3D2 and the 9850X3D was precisely the suggestion that client Zen6 is getting a delay into 2027.

His reason why Zen 6 would be delayed to 2027 is flawed though:

Why would AMD leave on the table 1 year worth of sales of a processor that is going to have > 20% performance increase (on N2P), waiting for another 3-4% over the 20% that N2X can deliver? It does not make sense.

 

[ToTTenTranz]

Why would AMD leave on the table 1 year worth of sales of a processor that is going to have > 20% performance increase (on N2P), waiting for another 3-4% over the 20% that N2X can deliver? It does not make sense.

The delay wouldn't happen because AMD wanted to, but rather because of delays in the production line.

 

[Joe NYC]

The delay wouldn't happen because AMD wanted to, but rather because of delays in the production line.

I don't think N2P is delayed.

The problem that MLID is talking about is AMD not using the process that is available (N2P) and instead delaying release, waiting for process that is not available yet (N2X) for very marginal performance benefit.

 

[Kepler_L2]

The delay wouldn't happen because AMD wanted to, but rather because of delays in the production line.

There are no delays, N2P is literally in production right now

 

[adroc_thurston]

There are no delays, N2P is literally in production right now

well, N2 is.
But N2p bears a magic trick.

 

[Saylick]

But N2p bears a magic trick.

I'm guessing that would be that N2P shares the same design rules as N2, so in essence there would be zero delay for AMD to use N2P even if they started their design targeting N2.

 

[Tigerick]

Yeah, just like X2-Elite is using combination of N3P + N3X before HVM of N3X; now some news said X2-Elite is fabbed at N3X.

If Venice is using N2P library most likely to boost ST performance, the rest still using N2 node.

 

[adroc_thurston]

I'm guessing that would be that N2P shares the same design rules as N2, so in essence there would be zero delay for AMD to use N2P even if they started their design targeting N2.

Yuh, you TO N2 and write the final mask set on N2p.
#winning

 

[marees]

This is also my thought. Not everything exist in lab could become a real product that you could buy. Same goes for Intel and other manufacturers.

https://twitter.com/x/status/2008443309438480519

View attachment 136293



Latest rumor is not only 9850X3D but also there would be something like R7-9750X and R5-9650X on the way, I'm not sure. Better than nothing/Nothingburger. Move on, all hopes and bets shall be on Zen6.

Supposedly, 9950X3D2 leak: https://browser.geekbench.com/v6/cpu/15755183

hold your horses !!!

Extensive details about the CPU have still not been revealed, but when asked on Tuesday morning in a small group of journalists, AMD at least stated: Stay tuned! More information should therefore be available soon. AMD had already uploaded an image to its press area that could show a dual-X3D processor.​

https://www-computerbase-de.transla...665/?_x_tr_sl=auto&_x_tr_tl=en&_x_tr_hl=en-GB

https://twitter.com/x/status/2008700495255462160

https://x.com/ComputerBase/status/2008637546050248841?s=20

 

[jdubs03]

Get prepared to pay a pretty Penny for a marginal Mary.

 

[eek2121]

hold your horses !!!

Extensive details about the CPU have still not been revealed, but when asked on Tuesday morning in a small group of journalists, AMD at least stated: Stay tuned! More information should therefore be available soon. AMD had already uploaded an image to its press area that could show a dual-X3D processor.​

https://www-computerbase-de.transla...665/?_x_tr_sl=auto&_x_tr_tl=en&_x_tr_hl=en-GB

https://twitter.com/x/status/2008700495255462160

https://x.com/ComputerBase/status/2008637546050248841?s=20

Yep, came here to post this:

https://videocardz.com/newz/amd-ryzen-9-9950x32d-confirmed-dual-3d-v-cache-cpu-is-coming

Here we go (again) 🤣


I guess the bean counters won.

 

[MS_AT]

interesting marketing spin

 

[ToTTenTranz]

9950X3D2 bros...
WAKE ME UP INSIDE

WAKE UP

AMD Workstation | Sytronix

www.sytronix.co.uk

 

[ToTTenTranz]

Zen 5 definitely sounds promising, but I think a lot of the hype is getting ahead of the facts. We’ve got some interesting hints about IPC and core changes, and Turin/Granite Ridge should show what the architecture can really do, but until AMD actually publishes numbers it’s all speculation. Strix Point especially will be interesting to watch for efficiency, not just raw performance. Best to wait for real silicon before calling this a huge leap.

Is this a post from 2023? From a LLM with a 2023 data training cutoff?

 

[marees]

Is this a post from 2023? From a LLM with a 2023 data training cutoff?

training/testing probably

 

[Fjodor2001]

We can conclude that it's a SKU that doesn't really makes sense

AMD disagrees with you since 9950X3D2 is about to be launched based on the latest info/leaks.