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

Page 366 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Timorous

Golden Member
Oct 27, 2008
1,723
3,125
136
I do believe quite a few. I find Kepler, Xino etc to credible and I wait and hope for things they say come to pass

But my belief != fact.
I believe that I will successful in my career. This does not mean that I am until things are unfolding before me.

I also have a soft spot for AMD, I bought a lot of AMD shares in 2017 (between 9 and 11 USD each) , little did I know they will grow this big. I have no more AMD shares, already cashed out.
This does not mean I crap on everything else.

I created this account exactly one week before I received my first Zen CPU. I knew there is potential with Zen, I invested in AMD stocks, pre ordered the CPU and got excited about it. Only thing is that if I waited long enough before cashing out, I could be planning my retirement right now. But one thing I dont like is how this forum evolved.
Stilt was active back then and folks like Dresden Boy, retired engineer, etc, putting out rich technical posts.

Now it is an echo chamber for wannabe leakers.

It has always been a mix. Back before Anand moved to Apple we had the likes of Charlie D, Fudzilla, The Inquirer and other tech journals that do broadly what WCCFTech / VideoCardz / MLID / RGT / Twitter leakers do. We also had the likes of IDC with what seemed like inside info or at the very least connections that gave him insight into some of Intels inner workings.

We also had Rollo and the NV marketing posters.

I do think the fact that the front page no longer really attracts the technically minded due to the lack of deep dives that Anand was fantastic at does contribute somewhat but aside from Chips n Cheese and a few people over at the various substack pages nobody really does deep dives into architectures anymore so you end up being left with commits or some patents or whitepapers with which to get into the technicals and without an Anand to take that + speaking to the design teams to present that in an understandable way it is rather hard to get into it.

Maybe we should do as Stephen Fry suggests and leave more room for doubt. I know most of my musings are based on a bunch of unverifiable assumptions so can be wildly wrong or accidentally correct. Still fun though.
 

Mahboi

Senior member
Apr 4, 2024
741
1,316
96
I do think the fact that the front page no longer really attracts the technically minded due to the lack of deep dives that Anand was fantastic at does contribute somewhat but aside from Chips n Cheese and a few people over at the various substack pages nobody really does deep dives into architectures anymore
Space getting too demanding for enthusiasts to write on it in a digestible way, I guess?
I've been paying interest for something like 1-2 years now, feels like I'll need years yet before I get anything else than a fuzzy general impression of how things work.
 
  • Like
Reactions: Kyoka

FlameTail

Diamond Member
Dec 15, 2021
3,076
1,756
106
I do think the fact that the front page no longer really attracts the technically minded due to the lack of deep dives that Anand was fantastic at does contribute somewhat but aside from Chips n Cheese and a few people over at the various substack pages nobody really does deep dives into architectures anymore so you end up being left with commits or some patents or whitepapers with which to get into the technicals and without an Anand to take that + speaking to the design teams to present that in an understandable way it is rather hard to get into
I have noticed this too- that the quality of technical discussion has declined in recent years, especially in places like r/hardware. I don't know about this forum, but I assume it's true here as well. I think a large part of the reason is that indeed, we do not have likes Anand, Andrei or Ian anymore, putting out high quality articles that even the average techie can understand.
 
  • Like
Reactions: SpudLobby

Timmah!

Golden Member
Jul 24, 2010
1,448
695
136
It kinda is. You're just trying to justified your own bias. In what universe is 40% not considered a large jump?
In universe, where most people are not computer enthusiasts like us buying the latest stuff just for the sake of it (within certain budget or not).
I am pretty sure random person not interested in computer hardware, would upon week of using computer with Zen4 and Zen5 for tasks random non-hardware enthusiast people use computers, would be not able to tell the difference. Good luck convincing these people they need to pay significant extra for something they cant even perceive for the most part.
 

Goop_reformed

Senior member
Sep 23, 2023
227
295
96
In universe, where most people are not computer enthusiasts like us buying the latest stuff just for the sake of it (within certain budget or not).
I am pretty sure random person not interested in computer hardware, would upon week of using computer with Zen4 and Zen5 for tasks random non-hardware enthusiast people use computers, would be not able to tell the difference. Good luck convincing these people they need to pay significant extra for something they cant even perceive for the most part.
You're skirting around the question. If you don't consider 40% ipc uplift impressive, then what will ? You're not arguing in good faith.
 
  • Love
Reactions: spursindonesia

Glo.

Diamond Member
Apr 25, 2015
5,755
4,664
136
I do believe quite a few. I find Kepler, Xino etc to credible and I wait and hope for things they say come to pass

But my belief != fact.
I believe that I will successful in my career. This does not mean that I am until things are unfolding before me.

I also have a soft spot for AMD, I bought a lot of AMD shares in 2017 (between 9 and 11 USD each) , little did I know they will grow this big. I have no more AMD shares, already cashed out.
This does not mean I crap on everything else.

I created this account exactly one week before I received my first Zen CPU. I knew there is potential with Zen, I invested in AMD stocks, pre ordered the CPU and got excited about it. Only thing is that if I waited long enough before cashing out, I could be planning my retirement right now. But one thing I dont like is how this forum evolved.
Stilt was active back then and folks like Dresden Boy, retired engineer, etc, putting out rich technical posts.

Now it is an echo chamber for wannabe leakers.
I can only speak for myself. And for the past 5-6 years the enthusiasm for technology is dwindling in me. I am becoming absolutely not excited about anything, apart from truly groundbreaking stuff, like powerful APUs, and efficient tech.

I presume its because tech has become so powerful, that we are not bound by compute, but by memory.

Besides. What do I play these days? World of Warcraft, Hearthstone? That runs on a potato. The games that I would want to play are on consoles, not only PCs. Heck, there are games that are coming to mobile(tablets) that I would happily play on those platforms, instead of PCs.

So what is there for me to be excited about?

I currently see PCs as tools, not objects of worship. I want things to "just work" on PC, the same way they do on iPad/iPhone, consoles.

And that last part: the fact that stuff just works on mobile, and on consoles might be also the reason why the quality of technical discussion on forums is dwindling. People are just not interested in PCs, specs, tinkering, whatever.

P.S. I presume - large part of killing my enthusiasm for PC was my venture to Linux. 1.5 years, and - until stuff works on linux, let alone "just works" - I don't want to have to deal with it.
 

inf64

Diamond Member
Mar 11, 2011
3,757
4,198
136
50% uplift in ST and MT is tremendous.

If it is true, it slays the entire competition.

Intel, Apple, AMD, Qualcomm are all cooked.
They are referencing the old "news" from RGT... Not really reliable, just like MLID. I'm not saying I don't believe the numbers could end up being in the same ballpark, just that these news sources are dubious.
 

yuri69

Senior member
Jul 16, 2013
425
707
136
People are just not interested in PCs, specs, tinkering, whatever.

P.S. I presume - large part of killing my enthusiasm for PC was my venture to Linux. 1.5 years, and - until stuff works on linux, let alone "just works" - I don't want to have to deal with it

This is strange. The open source ecosystem surrounding Linux is a nearly endless source of tinkering, tuning, hacking, and - most importantly - a way to "know how things work under the hood". Discovering new things and understanding how technology really works can be an exciting experience. No other PC platform can provide this.
 

Glo.

Diamond Member
Apr 25, 2015
5,755
4,664
136
This is strange. The open source ecosystem surrounding Linux is a nearly endless source of tinkering, tuning, hacking, and - most importantly - a way to "know how things work under the hood". Discovering new things and understanding how technology really works can be an exciting experience. No other PC platform can provide this.
Yep. And because I HAD TO tinker, troubleshoot, fix, and learn under the hood stuff I don't want to HAVE TO do this anymore.
 

Rigg

Senior member
May 6, 2020
475
1,004
136

More rumours.
The 7950X CB2024 scores in that article are garbage.

Hardware Unboxed (techspot) CB2024
7950X: 2187/126
7950X3D: 2119/124

CPU-Monkey CB2024
7950X: 2142/125
7950X3D: 2108/121
5950X: 1494/95

CGDirector CB2024
7950X: 2185/125
7950X3D: 2117/121
5950X: 1494/98

Average
7950X: 2171/125
7950X3D: 2115/122
5950X: 1494/97

45%/29% gain from 5950X to 7950X
42%/26% gain from 5950X to 7950X3D
38%/36% gain from 7950X to 9950X (rumored)
42%/39% gain from 7950X3D to 9950X (rumored)

If this rumor is accurate, it's an incredibly impressive technical achievement to get this kind of uplift at the same or similar clocks. It's not really a huge overall performance leap over what we got from Zen 3 to Zen 4 in this benchmark though. The extra single thread uplift looks amazing but lets keep in mind that Zen 4 is already significantly behind Apple and Intel in that respect.
 
  • Like
Reactions: Elfear

Saylick

Diamond Member
Sep 10, 2012
3,362
7,061
136
If this rumor is accurate, it's an incredibly impressive technical achievement to get this kind of uplift at the same or similar clocks. It's not really a huge overall performance leap over what we got from Zen 3 to Zen 4 in this benchmark though. The extra single thread uplift looks amazing but lets keep in mind that Zen 4 is already significantly behind Apple and Intel in that respect.
Considering that Zen 4 enjoyed a pretty substantial clock speed gain (+16%) and a decent IPC increase (+13%), if Zen 5 can achieve the same uplift, if not more, on a marginally improved node that's pretty impressive imo.
 

Rigg

Senior member
May 6, 2020
475
1,004
136
Considering that Zen 4 enjoyed a pretty substantial clock speed gain (+16%) and a decent IPC increase (+13%), if Zen 5 can achieve the same uplift, if not more, on a marginally improved node that's pretty impressive imo.
I agree, and said as much in my first sentence. At the end of the day it's the uplift that matters and not how it was achieved though.
 
Jul 27, 2020
17,613
11,369
106
38%/36% gain from 7950X to 9950X (rumored)
The MT uplift indicates much, much improved SMT scaling. That points to a massive overhaul of the structures within the core, much better than anything Intel has ever achieved. I really hope it's true. Zen 5 will breathe much needed life into x86. It actually tracks really well with AMD's way of thinking. They make computing more efficient, rather than brute forcing their way through it like Intel (a pretty moronic approach IMO).

I really wish life weren't such an awful struggle. It's almost unbearable waiting till June :(

And the increased SMT uplift will do wonders for multithreaded game engines as well as multithreaded GPU drivers. Everything will benefit. It's like alleviating bottlenecks in the entire pipeline and making data flow through like a flood!

It's so funny if you think about it. AMD incorporated better and more efficient AVX-512 and Intel killed it in desktop CPUs. Now they purportedly implemented better and more efficient SMT and Intel is allegedly killing it in Arrow Lake. Intel be like, "Hey! No fair that you can do things better than me. I refuse to compete!".
 
Last edited:

Timmah!

Golden Member
Jul 24, 2010
1,448
695
136
The 7950X CB2024 scores in that article are garbage.

Hardware Unboxed (techspot) CB2024
7950X: 2187/126
7950X3D: 2119/124

CPU-Monkey CB2024
7950X: 2142/125
7950X3D: 2108/121
5950X: 1494/95

CGDirector CB2024
7950X: 2185/125
7950X3D: 2117/121
5950X: 1494/98

Average
7950X: 2171/125
7950X3D: 2115/122
5950X: 1494/97

45%/29% gain from 5950X to 7950X
42%/26% gain from 5950X to 7950X3D
38%/36% gain from 7950X to 9950X (rumored)
42%/39% gain from 7950X3D to 9950X (rumored)

If this rumor is accurate, it's an incredibly impressive technical achievement to get this kind of uplift at the same or similar clocks. It's not really a huge overall performance leap over what we got from Zen 3 to Zen 4 in this benchmark though. The extra single thread uplift looks amazing but lets keep in mind that Zen 4 is already significantly behind Apple and Intel in that respect.
To be fair, their 2052 is bit on the low side, but on other hand i have no clue how HW Unboxed scored 2187. Just ran the test myself, 10 minute one, and my score is 2099. When i ran the CPU slightly OC´ed for a while to 5,212 avg all core frequency (but then dropped the OC cause of occasional instability), the score was 2120. Dont know what i could do to get to 2185, except do all-core OC to something like 5,3/5,4.
 
  • Like
Reactions: Joe NYC

Rigg

Senior member
May 6, 2020
475
1,004
136
To be fair, their 2052 is bit on the low side, but on other hand i have no clue how HW Unboxed scored 2187. Just ran the test myself, 10 minute one, and my score is 2099. When i ran the CPU slightly OC´ed for a while to 5,212 avg all core frequency (but then dropped the OC cause of occasional instability), the score was 2120. Dont know what i could do to get to 2185, except do all-core OC to something like 5,3/5,4.
What memory settings/configuration are you running? CB2024 scores are more reflective of memory performance than previous versions of Cinebench. I played around with a 4x8 B-die kit (running 3800 MT/s and Ryzen DRAM calculator fast settings) on a 5800x system I just sold. Even with a manually set 85C temp limit throttling it a bit I was consistently hitting mid 900's multi in CB2024. HUB was using 6000 C30 for that score. CGDirector was at 2185 and CPU monkey not far behind at 2145.

I didn't dig deep to find these scores, but given how consistent they are I'd say your system may be under performing a little in comparison to these test bench setups. Do you have stuff running in the background when running it? I know Armory Crate (gross) significantly hurt my CB2024 scores on my Intel system for example.

Cooling and silicon lottery could also be a factor here. Maybe the test systems are clocking higher at the temp limit.
 

Timmah!

Golden Member
Jul 24, 2010
1,448
695
136
What memory settings/configuration are you running? CB2024 scores are more reflective of memory performance than previous versions of Cinebench. I played around with a 4x8 B-die kit (running 3800 MT/s and Ryzen DRAM calculator fast settings) on a 5800x system I just sold. Even with a manually set 85C temp limit throttling it a bit I was consistently hitting mid 900's multi in CB2024. HUB was using 6000 C30 for that score. CGDirector was at 2185 and CPU monkey not far behind at 2145.

I didn't dig deep to find these scores, but given how consistent they are I'd say your system may be under performing a little in comparison to these test bench setups. Do you have stuff running in the background when running it? I know Armory Crate (gross) significantly hurt my CB2024 scores on my Intel system for example.

Cooling and silicon lottery could also be a factor here. Maybe the test systems are clocking higher at the temp limit.
I have the 64GB G-Skill 6000MHz kit running at that speed, did not tweak it any further. Timings are 30:40:40:96. I guess better/faster RAM is one possibility i did not think about, but you say HUB used the same as mine. Cooling is Arctic Freezer 420, could not really do better than that i think, aside of custom loop.
Would have to check the Armory thing. I have it installed, but i dont see it running on System tray, so not sure if its running in the background or not. I will look into it, thx for headsup.

On topic of the Tweaktown article, seems to be only taking over RTG stuff. Wish it was true, but unless i see the numbers confirmed from some other source, i am not convinced.
 
  • Like
Reactions: Tlh97 and Rigg