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

Page 962 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

StefanR5R

Elite Member
Dec 10, 2016
6,670
10,550
136
Copying this here for reference:
Apparently, this is Turin-Dense:
1745043616754-jpeg.122320
 

jpiniero

Lifer
Oct 1, 2010
16,799
7,249
136
I wonder if they are spinning a silicon version from the initial plans that didn't have an NPU, but did have a MALL cache instead? That would be... interesting.

Unlikely. If anything, I could see the AM5 products be where they dump the busted NPU dies.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
27,236
16,106
136
copilot pcs and rebranding..... Waste of everyone's money and time. Dell takes the cake lol
With Dell's recent huge loses most likely due to sticking with Intel exclusively for many years, it's no surprise anything they do.
 
  • Like
Reactions: r.p

gdansk

Diamond Member
Feb 8, 2011
4,568
7,681
136
What a missed opportunity to do dual chiplet 3D cache
There is still plenty of time before Zen 6 launches. They could even launch it after Zen 6. But based on AMD's earlier comments it seems they find the configuration questionable.
 

StefanR5R

Elite Member
Dec 10, 2016
6,670
10,550
136
EPYC 4005 was launched yesterday.
Compared to EPYC 4004,
– there are no 4c, 8c-105W, and 12c-X3D models anymore,
– a 16c-65W model was added,
– TDP of the 16c-X3D top dog was bumped from 120W to 170W.
The 16c-X3D model is still heterogeneous, like its Ryzen counterpart.

https://www.amd.com/en/products/processors/server/epyc/4005-series.html
https://www.amd.com/en/newsroom/pre...ls-epyc-4005-series-processors-deliverin.html

Article with AMD's slide set:
https://www.theregister.com/2025/05/13/amd_epyc_4005/
 

StefanR5R

Elite Member
Dec 10, 2016
6,670
10,550
136

Unsurprisingly the 16 Core models are a lot faster than the top Raptor Lake Xeon... since the Cinebench Accelerators wouldn't really work there.
They are not merely "not really working there", they are altogether disabled by Intel in Xeons. (Always have been, in all small-socket Xeon generations since Intel put heterogeneous cores into the corresponding laptop CPUs.)
 
  • Like
Reactions: lightmanek

adroc_thurston

Diamond Member
Jul 2, 2023
7,070
9,816
106
As long as it s not loaded, that is, if you just use it as a e-reader, otherwise once it s loaded with more than a single core it will drain 30W.
It's not gonna draw this much platform power on average. Boosts are very transient in normal use.
Also LNL 1t power is 12-15W iirc.
 
  • Like
Reactions: marees
Jul 27, 2020
27,978
19,117
146
Even a browser like Firefox is multithreaded since ages, there s no more purely ST real tasks, so even with transients there will be several cores loaded.
Very true.

1747427244032.png

This is idle activity on my main laptop. Just tons of browser tabs in Chrome, Firefox and Brave.
 

adroc_thurston

Diamond Member
Jul 2, 2023
7,070
9,816
106
You re hitting 100% on several cores when just downloading a page with a 100-300Mb internet connection
Not really, no, lmao.
so a transient but at full 30W on such a laptop.
You're not doing full 30W there over the full workload and most LNL SKUs are 17W anyhow.

Crikey, why do *I* have to defend Intel in a Z5 thread of all places.