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

Page 731 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

exquisitechar

Senior member
Apr 18, 2017
684
942
136
Every time I look more closely at STX results, all I can think of is that I would much rather have a single-CCX 8-core with a larger L3.

It would lose on, like, cinebench, but do better on nearly everything people actually care about.
Absolutely. Zen 5C would have best been kept for Turin Dense, the hybrid config in STX is unnecessary and mostly good for Cinebench bragging rights.
 

CouncilorIrissa

Senior member
Jul 28, 2023
540
2,120
96
Every time I look more closely at STX results, all I can think of is that I would much rather have a single-CCX 8-core with a larger L3.

It would lose on, like, cinebench, but do better on nearly everything people actually care about.
Yeah.
I wonder if Kraken Point will outperform STX in some scenarios as it's a single-CCX design (although it's 4 classic + 4 dense) that won't suffer from horrific cross-CCX latency.
 

inquiss

Senior member
Oct 13, 2010
200
274
136
There are many MT workloads that don’t require that much memory bandwidth.
So you hobble the platform for the hope that people buy it for those workloads. How many people that want that do you think there are? Hardly anyone buys the highest core count CPUs as it is. If you really want higher core counts you get threadripper.
 
  • Like
Reactions: carancho

gdansk

Platinum Member
Feb 8, 2011
2,963
4,493
136
So you hobble the platform for the hope that people buy it for those workloads. How many people that want that do you think there are? Hardly anyone buys the highest core count CPUs as it is. If you really want higher core counts you get threadripper.
You'd be surprised how many people buy based on some irrelevant benchmark.
It's the main reason Strix is 4+8 instead of something more rational. If the competition is spamming cores you better too.

I don't think it would be a good part for many people but it might sell anyway.
 

Mahboi

Golden Member
Apr 4, 2024
1,033
1,897
96
It shreds in PHP, Node, Python, databases and Apache (the last one is seeing especially absurd gains).
The real shocker for server is that Zen 5 is an actual slam dunk.
- DBs
- in memory DBs
- web server
- python/JS/php

All that's left is for someone to test a load balancer and we're set. It's basically THE chip to buy for anything cloud/general web services. Freaking 8 core Zen 5 is nibbling at 13900ks/14700ks, with the odd "destroy everything including 14900Ks" benchmark.

If I were a CP I'd just buy EPYC Zen 5 and nothing else. It may be a bust in games and in so many apps, but for web services it's obliterating any competition.
 

branch_suggestion

Senior member
Aug 4, 2023
392
875
96
I think people are beginning to understand how amazing Strix Halo will be for developers of all sorts of programs.
Just give it day 1 ROCm on both Windows and all major Linux distros and we are good to go. You can pretty much do anything you would usually do on a desktop workstation outside of the really extreme stuff, in which case you just load up a cloud instance or remote desktop.
If AMD marketing has some braincells left, they might steal some customers from the very entrenched Mac dev crowd.
 

inquiss

Senior member
Oct 13, 2010
200
274
136
You'd be surprised how many people buy based on some irrelevant benchmark.
It's the main reason Strix is 4+8 instead of something more rational. If the competition is spamming cores you better too.

I don't think it would be a good part for many people but it might sell anyway.
I'm sure it would sell a couple of units. People here maybe. But this is such a niche of a niche of a niche that's it's clearly not worth the engineers and packaging effort.
 
  • Like
Reactions: Tlh97 and carancho

Mahboi

Golden Member
Apr 4, 2024
1,033
1,897
96
I think people are beginning to understand how amazing Strix Halo will be for developers of all sorts of programs.
Just give it day 1 ROCm on both Windows and all major Linux distros and we are good to go. You can pretty much do anything you would usually do on a desktop workstation outside of the really extreme stuff, in which case you just load up a cloud instance or remote desktop.
If AMD marketing has some braincells left, they might steal some customers from the very entrenched Mac dev crowd.
Hold those horses.
Membw is going to be a MASSIVE issue. I can't even imagine how they intend to do it, unless they're going to go for something like LPDDR5X + some GDDR6 locked in somewhere for the GPU. Something like 16Go GDDR6 for 64Go total RAM may just be a product. Otherwise I really can't fathom how Halo is going to handle 40 CUs when Point is already seeing some games jump 40% from PHX to STX-P and some others barely nudge a 15% because membw is the problem, not compute.
 
  • Like
Reactions: Saylick

CouncilorIrissa

Senior member
Jul 28, 2023
540
2,120
96
Hold those horses.
Membw is going to be a MASSIVE issue. I can't even imagine how they intend to do it, unless they're going to go for something like LPDDR5X + some GDDR6 locked in somewhere for the GPU. Something like 16Go GDDR6 for 64Go total RAM may just be a product. Otherwise I really can't fathom how Halo is going to handle 40 CUs when Point is already seeing some games jump 40% from PHX to STX-P and some others barely nudge a 15% because membw is the problem, not compute.
Halo has a 256-bit memory interface unlike STX1 with 128-bit. It also supports 8533 MT/s memory speed compared to STX's 7500, with 32 MB of MALL on top.
 

branch_suggestion

Senior member
Aug 4, 2023
392
875
96
Hold those horses.
Membw is going to be a MASSIVE issue. I can't even imagine how they intend to do it, unless they're going to go for something like LPDDR5X + some GDDR6 locked in somewhere for the GPU. Something like 16Go GDDR6 for 64Go total RAM may just be a product. Otherwise I really can't fathom how Halo is going to handle 40 CUs when Point is already seeing some games jump 40% from PHX to STX-P and some others barely nudge a 15% because membw is the problem, not compute.
Halo has all the enhancements to uncore, fabrics, membw, mem capacity and even good old MALL.
The CPU is the selling point more than the GPU, at least until GPGPU software gets more competitive.
 

Fjodor2001

Diamond Member
Feb 6, 2010
3,990
440
126
The real shocker for server is that Zen 5 is an actual slam dunk.
- DBs
- in memory DBs
- web server
- python/JS/php
Depends on what you mean by slam dunk. Is it ~20% better than Zen4 in those specific use cases, looking at the perf tests that were linked to (comparing 7700X vs 9700X)?

If they were aiming for 16% general IPC increase with Zen5, then I’d expect more than 20% in some cherry-picked server/DC use cases.
 

Mahboi

Golden Member
Apr 4, 2024
1,033
1,897
96
Halo has a 256-bit memory interface unlike STX1 with 128-bit. It also supports 8533 MT/s memory speed compared to STX's 7500.
STX-P's 16CUs are already feeling the pressure.
Here you double the membw + up it by something like 15%.

16/40 = 0.4
1/2.15 =0.46
It's not even following the CU increase. :eek:
 

Mahboi

Golden Member
Apr 4, 2024
1,033
1,897
96
Halo has all the enhancements to uncore, fabrics, membw, mem capacity and even good old MALL.
The CPU is the selling point more than the GPU, at least until GPGPU software gets more competitive.
Ok the MALL may make a more serious difference. But I'm still skeptical. When I'll see direct bench comparisons between 12CU STX and 16CU STX, I think we'll see just how bad the bottlenecks are.
Already with PHX, I saw some cases where 8 CUs somehow did 90% as good as 12. Not good.
 

marees

Senior member
Apr 28, 2024
455
512
96
Depends on what you mean by slam dunk. Is it ~20% better than Zen4 in those specific use cases, looking at the perf tests that were linked to (comparing 7700X vs 9700X)?

If they were aiming for 16% general IPC increase with Zen5, then I’d expect more than 20% in some cherry-picked server/DC use cases.
Zen (4)5%

(& zen 5c on server is 3nm )

 
  • Like
Reactions: Saylick

poke01

Platinum Member
Mar 8, 2022
2,205
2,803
106
I think people are beginning to understand how amazing Strix Halo will be for developers of all sorts of programs.
Just give it day 1 ROCm on both Windows and all major Linux distros and we are good to go. You can pretty much do anything you would usually do on a desktop workstation outside of the really extreme stuff, in which case you just load up a cloud instance or remote desktop.
If AMD marketing has some braincells left, they might steal some customers from the very entrenched Mac dev crowd.
Been saying this since day1, also for Zen6 (Halo successor) move to on-package memory and that would enable a 512-bit bus on a laptop. If AMD is serious about AI then it’s something they have to do to fully convince that crowd.
 

CouncilorIrissa

Senior member
Jul 28, 2023
540
2,120
96
Depends on what you mean by slam dunk. Is it ~20% better than Zen4 in those specific use cases, looking at the perf tests that were linked to (comparing 7700X vs 9700X)?

If they were aiming for 16% general IPC increase with Zen5, then I’d expect more than 20% in some cherry-picked server/DC use cases.
Except it's much more than 20% in some of those.
1723377912197.png
+25.5% perf @ iso-power is nothing to scoff at. And that's GNR, whereas cloud will buy Turin-D on N3E with even bigger iso-power gains and a non-cope memory subsystem.

And I didn't have to take something like OpenSSL which uses the AVX-512.

Besides, for AMD server/DC is not a "cherry-picked" use case, it's literally what makes them the most money.
 

CouncilorIrissa

Senior member
Jul 28, 2023
540
2,120
96
Ok, seems like Zen5 is primarily for server/DC and not client then. :cry:

Would be nice if AMD went with separate client vs server dies/cores going forward, so client could make other prioritizations. Zen6 perhaps?
Probably? There are some rumours that Zen 6 DT client will be different from server.

Here's a hot take from me: I personally believe that all CPUs post-Alder Lake are a vast freaking overkill for gaming. It's really not that relevant whether CS2 runs at 600 fps or 750.

There are some games that hammer the CPU (Starfield comes to mind), but it's really mostly devs' fault.

What I'd like to see improved is single-core power consumption. It'd be hella nice to see top-end laptop chip top out at 15W @ single-core at most.
 

marees

Senior member
Apr 28, 2024
455
512
96
Ok, seems like Zen5 is primarily for server/DC and not client then. :cry:

Would be nice if AMD went with separate client vs server dies/cores going forward, so client could make other prioritizations. Zen6 perhaps?
An 8x full core (optimized for client) with 3d vcache & 16x c core (optimized for server) for total of 24 cores could help AMD tick all boxes

Not sure if it makes sense for AMD
 
  • Like
Reactions: Tlh97 and Joe NYC

yottabit

Golden Member
Jun 5, 2008
1,496
528
146
An 8x full core (optimized for client) with 3d vcache & 16x c core (optimized for server) for total of 24 cores could help AMD tick all boxes

Not sure if it makes sense for AMD
Yeah so people can complain about how bad the inter-CCX latency is, scheduling issues, and we can hear endlessly “Why didn’t AMD just make a full fat 16 core CPU?”
 
  • Like
Reactions: Tlh97 and marees