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

Page 297 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

PJVol

Senior member
May 25, 2020
854
838
136
ISO is the international standards office so iso clocks or iso node or iso power is just short hand for standardising the given parameter.
Not sure about node, but how clocks or power can be "standardised" ?)
Or may be "iso" is kind of e-slang synonym of "the same" in this context?
I just never seen using the term "ISO' in this context in any publications.
 

biostud

Lifer
Feb 27, 2003
19,902
7,008
136
Not sure about node, but how clocks or power can be "standardised" ?)
Or may be "iso" is kind of e-slang synonym of "the same" in this context?
I just never seen using the term "ISO' in this context in any publications.
Neither have I...
 

Timorous

Golden Member
Oct 27, 2008
1,978
3,864
136
Not sure about node, but how clocks or power can be "standardised" ?)
Or may be "iso" is kind of e-slang synonym of "the same" in this context?
I just never seen using the term "ISO' in this context in any publications.

By setting them to be the same on all test subjects.

It is used quite frequently. TSMC have used it to refer to things like power savings at iso clocks. or performance improvements at iso power.
 

kschendel

Senior member
Aug 1, 2018
295
235
116
ISO is the international standards office so iso clocks or iso node or iso power is just short hand for standardising the given parameter.
ISO is in fact just one of many international SDO's (Standards Development Organizations). There are many others, such as the IEC, ITU, and JEDEC just to name three.

It's not really correct to say something like "ISO node" unless there is actually an ISO (or joint ISO/IEC) standard on that particular topic. Better to refer to the specific SDO publishing the relevant standard, assuming that we're talking about an actual standard, and not just some hand-waving that some number of people agree on.
 

gdansk

Diamond Member
Feb 8, 2011
4,567
7,679
136
It's not really correct to say something like "ISO node" unless there is actually an ISO (or joint ISO/IEC) standard on that particular topic. Better to refer to the specific SDO publishing the relevant standard, assuming that we're talking about an actual standard, and not just some hand-waving that some number of people agree on.
What? If someone says iso process it simply means equal process. Iso as in isosceles not International Organization for Standardization.
 
  • Like
Reactions: delta-v

eek2121

Diamond Member
Aug 2, 2005
3,408
5,046
136
i see, thank you for explanation.
Why does the shrink explain the bigger perf increase in MT in the 5950x > 7950x case? I mean, the shrink does not result in more cores for 7950x, that would explain it, and as far as power-headroom goes, is that not reflected in available clocks? Which got increased about the same in both ST (4,9>5,7) as MT (4,4>5,1) case between these generations. Is it cause of the 105W to 170W TDP increase?
Zen 4 had a much higher power limit, and was made on a custom N5 process. Both the process and the high power limit helped Zen 4 shine. AMD also was able to optimize the design to better scale with power (to a point) though from a frequency standpoint they are definitely still behind Intel.
 
  • Like
Reactions: Tlh97 and Timmah!

HurleyBird

Platinum Member
Apr 22, 2003
2,811
1,544
136
I'm guessing the LP cores would be too weak to handle gaming stuff so maybe they are just meant for OS threads. Then the big cores don't get interrupted by having to relinquish control and give their time slices to the OS threads. Should result in fewer hitches and more smoothness in frames.

Main benefit would be idle power.
 

jamescox

Senior member
Nov 11, 2009
644
1,105
136
Yes. Note that Zen 5 and Zen 4 are on similar processes. The process used for Zen 5 is technically better, but the savings are offset by larger cores.

Without a major shrink, Zen 5 cores will need more thermal and power headroom for nT and thus nT gains are usually smaller in those types of cases.

tl;dr
Single core gains are not power limited
multicore gains are and bigger cores means more power needed.
I have not had time to keep up with rumors lately so perhaps this has already been discussed. I am not searching back through ~300 pages of speculation.

I see Turin information saying 128 Zen 5 cores or 192 Zen 5c cores. 192 Zen 5c cores make sense (16 cores * 12 CCDs). How do we get 128 Zen 5 cores though? Is this known? It seems like it needs to be 16 cores * 8 CCDs, implying the existence of a 16 core CCD with Zen 5 cores. If they are significantly larger, then I don't know how reasonable that is. Other option is 16 separate CCDs with 8 cores each. Does the IO die have 16 GMI or only 12?
 

Tigerick

Senior member
Apr 1, 2022
844
797
106
I have not had time to keep up with rumors lately so perhaps this has already been discussed. I am not searching back through ~300 pages of speculation.

I see Turin information saying 128 Zen 5 cores or 192 Zen 5c cores. 192 Zen 5c cores make sense (16 cores * 12 CCDs). How do we get 128 Zen 5 cores though? Is this known? It seems like it needs to be 16 cores * 8 CCDs, implying the existence of a 16 core CCD with Zen 5 cores. If they are significantly larger, then I don't know how reasonable that is. Other option is 16 separate CCDs with 8 cores each. Does the IO die have 16 GMI or only 12?
Turin use the same die as Zen5 thus 16 pcs 8xZen5 = 128 Turin cores.

AMD-ZEN5-EPYC-TURIN-4.jpg

Turin Dense use newer process N3E with 12 pcs 16xZen5c = 192 Turin Dense cores.

AMD-ZEN5-EPYC-TURIN-1.jpg
 
Last edited:

Gideon

Platinum Member
Nov 27, 2007
2,030
5,034
136
Turin use the same die as Zen5 thus 16 pcs 8xZen5 = 128 Turin cores.

This looks awfully dense, particularily for the package routing on the same socket as Genoa (last image and paragraph in this article givea an example for AM4 zen2)

Will thet use the same I/O die layout or use some famcier packgaging?
 

adroc_thurston

Diamond Member
Jul 2, 2023
7,062
9,804
106
Main benefit would be idle power.
Consoles really don't need that, they're not laptops.
It's for always online background downloads stuff without the usual horrible hacks.
This looks awfully dense, particularily for the package routing on the same socket as Genoa
It's a tricky challenge yeah.
Will thet use the same I/O die layout or use some famcier packgaging?
Yea it's the same and very basic still.
Fancypants packaging is all Venice gen parts.
 
  • Like
Reactions: Tlh97

kschendel

Senior member
Aug 1, 2018
295
235
116
What? If someone says iso process it simply means equal process. Iso as in isosceles not International Organization for Standardization.
Unless one is counting keystrokes, why not just say "equal" or "equivalent", without introducing confusion? or indeed "isometric".

I've not run into the notion of "iso" as a standalone modifier; it strikes me as unnecessary argot. Anyway, this is OT, and I'll let it ride.
 
  • Like
Reactions: raghu78 and MadRat

dr1337

Senior member
May 25, 2020
523
806
136
Unless one is counting keystrokes, why not just say "equal" or "equivalent", without introducing confusion? or indeed "isometric".
I mean this is a technically oriented thread, we're talking about the fine details of microarchitecture and how they matter for these products/devices going forward.

I'd really assume the people trying to browse and read these threads would have the ability to parse technical jargon.

Heck throwing "iso node" into the search box gave me results of people saying it as far back as 2014. I think some users just need to exercise their context-clues muscle more instead of trying for semantics.

Edit: and not to say I don't love a discussion on etymology, it would be interesting to know where 'iso' started being popularized as a technical term. But the whole debate around the word/prefix isn't necessary. Its a lot easier to just ask someone what they mean, learn, and then move on.
 
Last edited:

gdansk

Diamond Member
Feb 8, 2011
4,567
7,679
136
I mean this is a technically oriented thread, we're talking about the fine details of microarchitecture and how they matter for these products/devices going forward.

I'd really assume the people trying to browse and read these threads would have the ability to parse technical jargon.

Heck throwing "iso node" into the search box gave me results of people saying it as far back as 2014. I think some users just need to exercise their context-clues muscle more instead of trying for semantics.
To be fair the post in question did erroneously capitalize it (autocorrect gone too far?) which caused some confusion.

Anyway, I think it is now clear to everyone what Boris was saying so I'll move on.
 

delta-v

Junior Member
May 1, 2023
16
20
41
Seems likeliest that iso is just shorthand for isometric which is Greek for 'equal measure'.

This applies cleanly to all uses I have seen in the tech world like iso clock, iso node, iso area etc etc.
Yep. This use of the prefix "iso" to mean "same" is very common in engineering. I typically deal with isothermal (same temperature), isotropic (same physical properties), isocontour (same line or surface on a plot), etc.
 

biostud

Lifer
Feb 27, 2003
19,902
7,008
136
While this semantic discussion can be fun are zen4 and Zen5 build on the same node?

I thought Zen5 was on TSMC 4nm?
 

Hitman928

Diamond Member
Apr 15, 2012
6,695
12,370
136
While this semantic discussion can be fun are zen4 and Zen5 build on the same node?

I thought Zen5 was on TSMC 4nm?

Zen 5 desktops will be on N4 (N4P I think), mobile as well I believe. Zen 4 desktop was N5 and mobile was N4. I believe the 8000 series refresh of Zen 4 mobile is on N4P already.