AMD Zen 4 Discussion Thread (31% faster than ADL)

matique

Senior Member
Joined
Sep 25, 2017
Messages
797
Reaction score
786
Here's hoping to be better performance uplift than mentioned. 15% ST improvement [including 10% boost in clock + IPC] means IPC improvement isn't much, a little disappointing since it's a whole node change. That said, it'll also be AMD's first foray into DDR5 and AMD has very poor track record of transitioning, initial DDR4 implementation was a complete mess & I have to temper expectations for this round too.

Since AMD has better grasp of the general market now, they don't have to undercut prices anymore and it seems like the new gen would be quite expensive, especially on the motherboard side. While they haven't confirmed OC ability for B650, I hope they allow it. All the x670e/x670 boards looking to be as expensive as their Intel counterparts.

Also it'll be kinda funny if in game, the new processors can't match 5800x3d lmao. May they have non stunted agesa, smooth DDR5 experience and better ST performance and still retain good prices. Also hope there's 1DPC boards being released, 2DPC absolutely sucks for DDR5 OC.
 

Jurchen King

Supremacy Member
Joined
Dec 16, 2007
Messages
6,305
Reaction score
2,305
Like how AMD ironed out agesa issues on current gen that's still ongoing? :p
Yah, if there's one gripe with AMD, it's their constant AGESA updates which sometimes introduces instability and performance issues (for overclockers). Then users have to roll back BIOS or wait & pray for next update to fix existing bugs.

In comparison, I don't see Intel updating their processor microcode as often and only do it to address security vulnerabilities.

Of course, some will use the "AMD ages like fine wine" argument to defend AMD as constantly improving performance through AGESA updates e.g. reduce core latency etc.
 
Last edited:

Mach3.2

Great Supremacy Member
Joined
Apr 8, 2011
Messages
72,405
Reaction score
2,465
Like how AMD ironed out agesa issues on current gen that's still ongoing? :p
The newer BIOSes also have vastly different CPU vcore characteristics as compared to pre-1.2.0.3c BIOSes, limiting the max boost you get if you have EDC set over 140A. You can literally bench better on older BIOSes.

AMD just like to do weird things yeah.
 

matique

Senior Member
Joined
Sep 25, 2017
Messages
797
Reaction score
786
Yah, if there's one gripe with AMD, it's their constant AGESA updates which sometimes introduces instability and performance issues (for overclockers). Then users have to roll back BIOS or wait & pray for next update to fix existing bugs.

In comparison, I don't see Intel updating their processor microcode as often and only do it to address security vulnerabilities.

Of course, some will use the "AMD ages like fine wine" argument to defend AMD as constantly improving performance through AGESA updates e.g. reduce core latency etc.

AMD can't ever fix their core latencies lol. I tried my best and could only bring it down to 54ns. Can only do so much with chiplets, and having to have everything synced. But more worrying is that some people still have USB dropout issues even with stock ram [no WHEA], i thought they would have already fixed that.
 

kimsix

Arch-Supremacy Member
Joined
Jan 6, 2009
Messages
19,012
Reaction score
9,056
not bery impressive leh. ST only 15% faster than a stock 5950X in cinebench23.....

31% faster in blender than Adl but with 35% more cores than Adl and on newer 5nm....

Rpl will add more E-cores....

quite worrying times for amd
 

kimsix

Arch-Supremacy Member
Joined
Jan 6, 2009
Messages
19,012
Reaction score
9,056
The newer BIOSes also have vastly different CPU vcore characteristics as compared to pre-1.2.0.3c BIOSes, limiting the max boost you get if you have EDC set over 140A. You can literally bench better on older BIOSes.

AMD just like to do weird things yeah.

how is your 1933FLCK? :o

my rma'ed b2 got a 1900FCLK hole heartbreak.

OCD me try for 1933FCLK and i find if i keep PBO at auto, near stock, i can avoid whea errors 19.
You try TDC@105 and EDC@140 and PPT@180.
The rest can bios auto even for vddp and vddg :o

of cos this limits allcores results as they get lesser TDC current, so droped from 4.55G to 4.25G under cb23 full load.
 

Mach3.2

Great Supremacy Member
Joined
Apr 8, 2011
Messages
72,405
Reaction score
2,465
how is your 1933FLCK? :o

my rma'ed b2 got a 1900FCLK hole heartbreak.

OCD me try for 1933FCLK and i find if i keep PBO at auto, near stock, i can avoid whea errors 19.
You try TDC@105 and EDC@140 and PPT@180.
The rest can bios auto even for vddp and vddg :o

of cos this limits allcores results as they get lesser TDC current, so droped from 4.55G to 4.25G under cb23 full load.
WHEA city. Even with 1.2V+ vSOC I can't get the WHEAs to slow down, so I gave up and settled with 1866MHz. General consensus over at the OCN Ryzen RAM OC thread is most dual CCD chips likely can't get above 1900MHz IF WHEA free.

My CPU IMC isn't the best, dual rank B dies won't POST at 3733MHz CL15 or lower, doesn't matter what I do. Sold the kit and ate the $50 loss. :spin:


The PBO trade with extra IF frequency boost isn't worth it imo, did you actually test using linpack and other benchmarks to verify your performance didn't actually regressed at 1933MHz IF? My PBO power limits are quite conservative, so that ain't it. My CPU's IMC is just dog water.. :s13:

WHEA 19s are likely a result from infinity fabric instability, there's a whole load of factors that can affect the stability of the infinity fabric; it's a wild goose chase and I'm done play that game. :s13:
 

watzup_ken

High Supremacy Member
Joined
Nov 21, 2003
Messages
25,672
Reaction score
2,123
Yah, if there's one gripe with AMD, it's their constant AGESA updates which sometimes introduces instability and performance issues (for overclockers). Then users have to roll back BIOS or wait & pray for next update to fix existing bugs.

In comparison, I don't see Intel updating their processor microcode as often and only do it to address security vulnerabilities.

Of course, some will use the "AMD ages like fine wine" argument to defend AMD as constantly improving performance through AGESA updates e.g. reduce core latency etc.
I agree. This is my biggest problem with AMD Ryzen processors. I’ve used every generation until I gave up last year. While it is nice that AMD continues to make improvement by means of AGESA updates, the optimisation inevitably will introduce bugs here and there. I think at least the issue is not apparent on the mobile Ryzen processors.
 

TanKianW

Supremacy Member
Joined
Apr 21, 2005
Messages
6,675
Reaction score
3,329
Still having USB driver install/uninstall issues or drop out on a few of my AMD system, including my gaming/workstation PC......Some system just works, some just........

On the production front, I really want to use their consumer CPUs....but then again I gave up.​
 

transparos

Senior Member
Joined
Dec 5, 2012
Messages
1,414
Reaction score
226
hope intel 13th gen also comes out during that period so we can compare and get the best bang for buck
 

matique

Senior Member
Joined
Sep 25, 2017
Messages
797
Reaction score
786
WHEA city. Even with 1.2V+ vSOC I can't get the WHEAs to slow down, so I gave up and settled with 1866MHz. General consensus over at the OCN Ryzen RAM OC thread is most dual CCD chips likely can't get above 1900MHz IF WHEA free.

My CPU IMC isn't the best, dual rank B dies won't POST at 3733MHz CL15 or lower, doesn't matter what I do. Sold the kit and ate the $50 loss. :spin:


The PBO trade with extra IF frequency boost isn't worth it imo, did you actually test using linpack and other benchmarks to verify your performance didn't actually regressed at 1933MHz IF? My PBO power limits are quite conservative, so that ain't it. My CPU's IMC is just dog water.. :s13:

WHEA 19s are likely a result from infinity fabric instability, there's a whole load of factors that can affect the stability of the infinity fabric; it's a wild goose chase and I'm done play that game. :s13:

I experienced USB dropouts, fixing whea 18 then fixing whea 19, weird stuttering at times because it didn't boost properly as well. After ADL came out, was almost a no brainer for me. Intel is expensive but once tuned right and stable, it will be stable. That said, stability was also a hit and miss with DDR5 with certain combinations. Hope AMD has it more stable from launch. Board partners and ram manufacturers probably already learned more about DDR5 through ADL, so it falls on AMD to make sure on their part it's okay.
 

Mach3.2

Great Supremacy Member
Joined
Apr 8, 2011
Messages
72,405
Reaction score
2,465
I experienced USB dropouts, fixing whea 18 then fixing whea 19, weird stuttering at times because it didn't boost properly as well. After ADL came out, was almost a no brainer for me. Intel is expensive but once tuned right and stable, it will be stable. That said, stability was also a hit and miss with DDR5 with certain combinations. Hope AMD has it more stable from launch. Board partners and ram manufacturers probably already learned more about DDR5 through ADL, so it falls on AMD to make sure on their part it's okay.
I personally didn't experience the USB dropouts nor notice any TPM stuttering, but it's just the cumulative "small issues" that you see being reported online that made me cast doubt on AMD's competence. AMD have very good hardware, but some corners just doesn't seem to be polished enough.

Someone on OCN receiving a lemon after trying to RMA their 5950X...
The 5950X he received after he RMA'd the lemon chip from the original RMA have pretty **** silicon too

Can't find the OCN thread regarding launch Zen 3s getting WHEA 18 bone stock, but yeah it was a thing.


One would think silicon binning improves as the process node ages, but I guess not. Maybe the better stuff ended up in Epycs, but at this point who knows.
There's enough **** I've seen on AMD's side that I'm willing to take my chances with Intel for my next upgrade, provided Intel remains competitive of course.
 
Last edited:

matique

Senior Member
Joined
Sep 25, 2017
Messages
797
Reaction score
786
I personally didn't experience the USB dropouts nor notice any TPM stuttering, but it's just the cumulative "small issues" that you see being reported online that made me cast doubt on AMD's competence. AMD have very good hardware, but some corners just doesn't seem to be polished enough.

Someone on OCN receiving a lemon after trying to RMA their 5950X...
The 5950X he received after he RMA'd the lemon chip from the original RMA have pretty **** silicon too

Can't find the OCN thread regarding launch Zen 3s getting WHEA 18 bone stock, but yeah it was a thing.


One would think silicon binning improves as the process node ages, but I guess not. Maybe the better stuff ended up in Epycs, but at this point who knows.
There's enough **** I've seen on AMD's side that I'm willing to take my chances with Intel for my next upgrade, provided Intel remains competitive of course.

no need to search for whea 18 bone stock, I had to rma mine as it had whea bone stock and after an hour or so might even BSOD. My replacement was fine though, but couldn't push ram too much else the USB dropouts start happening.

AMD still has time to fine tune their CPUs before launch, so I think they'll do decently. Really looking out for their second iteration though, should be properly good since they would have learned from mistakes from their first gen of AM5.
 

haylui

High Supremacy Member
Joined
Jul 18, 2006
Messages
29,763
Reaction score
91
Still having USB driver install/uninstall issues or drop out on a few of my AMD system, including my gaming/workstation PC......Some system just works, some just........

On the production front, I really want to use their consumer CPUs....but then again I gave up.​
Guess it is due to windows update? Yesterday suddenly all my Logitech wireless controller reporting driver not found... Logitech themselves doesn't provide driver some more...
 

haylui

High Supremacy Member
Joined
Jul 18, 2006
Messages
29,763
Reaction score
91
not bery impressive leh. ST only 15% faster than a stock 5950X in cinebench23.....

31% faster in blender than Adl but with 35% more cores than Adl and on newer 5nm....

Rpl will add more E-cores....

quite worrying times for amd
It is 46% faster 31% is lesser time taken
 

darkmatt

Supremacy Member
Joined
Oct 9, 2006
Messages
8,983
Reaction score
370
but 5.5GHz during gaming is quite a lot vs 5950x. Then don't know will have 3D cache version or not, also the 24 cores model.
i wonder if amd can sustain 5.5GHz all core.

if can maybe thats where the 30% came from..

so far no specs on how many cores that 5.5GHz is
 

Phen8210

High Supremacy Member
Joined
Jul 29, 2011
Messages
28,884
Reaction score
8,214
AMD needs to work on the little things that matter. Intel's new architecture isn't problem-free either since it's relatively new. We have to see how they both iron out the issues moving forward.

Stability, Compatibility, and Performance are top priorities. Having extra is a bonus, don't have then too bad. Most of the issues are blown out of proportion and do not matter for general usage such as gaming. In real-world use, it has zero difference.

I wouldn't spend too much time tweaking and nitpicking on things that don't make any impact as it does not make me any $$$.

This is coming from a person where the computer is the rice bowl.
 
Important Forum Advisory Note
This forum is moderated by volunteer moderators who will react only to members' feedback on posts. Moderators are not employees or representatives of HWZ. Forum members and moderators are responsible for their own posts.

Please refer to our Community Guidelines and Standards, Terms of Service and Member T&Cs for more information.
Top