Processors

[Migz-DH]

Did you know that Ryzen 5000 CPUs can't be overclocked to their maximum potential with the most recent AGESA versions? It's true*, **.

Apparently, when the AGESA was made for the 5800X3D, an EDC amp limit was implemented for it, but that limit affects other CPUs as well! This means that PBO is nearly useless for these other processors. In my case, I lost about 5 percent of my CPU's potential when I updated my BIOS this summer. I feel like AMD screwed me out of performance on my Ryzen 9 5900X.

Does AMD have a plan to correct this issue in the near future?

Are there other ways to solve this besides installing an older BIOS-version chip?

Thanks, and cheers!

* https://steamcommunity.com/app/223850/discussions/0/3648503910214432530/#c5002961597081495498

** https://hardforum.com/threads/did-i-degrade-or-destroy-my-5950x-now-with-a-response-from-asus-on-pag...

[thanatoast]

This *really* needs to be addressed soon. It's been far too long at this point.

For lower end SKU's like the 5600x, the issue is much more profound as those CPU's have a much lower default EDC (90). It's still possible to tweak the other PBO values though. Just gotta leave EDC alone.

Where is the next AGESA version?

[Boggy22]

I have recently decided to update my X570 Taichi Razer's BIOS only to find out that my previous 2-3 weeks of manually tuning the curve optimizer was for nothing. The 5950x that boosted to 5.2Ghz now barely goes to 5Ghz. Scores have plummeted, frequencies dropped AND temps have increased (!!) .
From 31k in R23 to 29.5k now, while getting 85C vs 80C. OCCT 87C at 4.15Ghz vs 82C at 4.4Ghz with 1.2.0.3. The difference is huge!

What are you doing AMD? Is ruining a great series of CPUs your tactic of making the 7000 series look better?  And you're wondering why people still prefer Intel. It's because of small -things- like this.

[thanatoast]

Some people seem to think its intentional and i'm not sure I buy that. There might be merit to the theory that this was meant for the 5800X3D too but I'm not sure. Either way, this has been a thing for FAR too long at this point.

Sucks about your PBO CO OC, but you can still salvage it by just putting EDC back to whatever your CPU's stock value is, you'll get your missing voltage back.

I've been wanting to do a proper PBO OC for so long now, but obviously cant due to this bug. Which led me to exploring an All core OC...which my 5600x seems to absolutely hate. So stuck with some super basic PBO settings and -5 curve on all cores...boring.

[goodplay]

Or could it be related to the 'power deviation reporting' from some board makers, and with the arrival of the 5800X3D they were pulled back into line with recommended specs.

[thanatoast]

No, theres no way this is intended behavior for ALL Ryzen 5000 CPU's.

Whenever EDC is raised above your processors default value - EVEN BY 1 - max voltage gets cut down from 1.45v to somewhere below 1.4v (in my case it was around 1.375v I think) which massively impacts maximum boost clocks, especially when using boost offset.

There's a way to actually check the voltage range, and you can observe this change, but I don't remember what the program was called. There's NO WAY this can be be intended behaviour - otherwise it means AMD willingly gimped the potential of their entire Ryzen 5000 line...which I find hard to believe to be honest.

Here's a post I made on MSI forum when I originally thought it was a motherboard specific BIOS issue.
https://forum-en.msi.com/index.php?threads/lower-pbo-voltages-on-agesa-1-2-0-7.376040/

And here's the post that finally shed some light on this behaviour for me, theres a lot of good info here.
https://www.reddit.com/r/Amd/comments/vecyn6/the_agesa_1207_pbo_bug_combined_with_a_65w_tdp/?ref=sha...

And yet nothing but radio silence on the matter. No new BIOS update in ages.

[Boggy22]

This is a limit that was supposed to influence only the 5800X3D, but instead it's affecting all CPU's. It's like they forgot a simple code line like "if (5800X3D)" before the limiter, so it's applied to all CPUs.
Intentional or not, I find it very very hard to believe that nobody at AMD is aware of this, considering the time it has passed from AGESA 1.2.0.3 C which was the last "good" BIOS. I'm guessing some people in this corporation might still have 5xxx CPUs and are overclockers. And if not, someone that might "pass the message" is reading this right now(or at least should be reading it). What really bugs me is that it seems like it's a very simple fix, isolating the 5800X3D for this mod. 

I did go back to EDC140 (stock), but it's affecting my multi-core performance a lot. Also, the temps seem crazy high. OCCT AVX2 with previous BIOS (1.2.0.2)  at 225W yelded 83C. Now it's shooting at 86 in the first seconds before saturating my water cooling system. So at least 5-6C difference at the same power.

Why am I not switching back to 1.2.0.2? Because I cannot run my memory at 3600Mhz / IF 1800Mhz with it (128GB). And I'm also awaiting the 4090 which might not work unless you update the BIOS first. The new BIOS on my Taichi Razer page is especially addressed to "Improve GPU compatibility for GeForce RTX 40 series."

[thanatoast]

I cant roll back to an earlier bios either. The last available BIOS for my board has AGESA 1.0.0.6...quite a gap there. Last time I tried, it couldn't POST afterwards.
Thank god for BIOS Flashback.

Is there actually ANY evidence that this was a change only meant for the 5800X3D?
Because I've heard this thrown around quite a lot with zero actual citation. Regardless though, even if it is or is not, somethings gotta change, or i'm gonna be unhappy for the rest of my 5600x ownership.

[Migz-DH]

Does AMD need some sort of regimented analysis to prove to them that this is a thing, or will only a massive public outcry solve this?

[wim-bart]

Anyone knowing the status of this. Before i found this article i did not know what was wrong. I had the feeling my settings where not applied, i lost over 120 points in CB, temps ate higher, but sadly cores don't go over 5.125GHz anymore. I was thinking i did write my BIOS settings down wrong.

Board: Asus prime x570 pro + R9 5950X

[Migz-DH]

It's still being ignored.

[Migz-DH]

Has anyone tried the new ComboV2PI 1208 AGESA for this issue yet?

ASUS just pushed it out to MBs this past week.

[sarge198]

Not yet, according to this thread the issue is still present: https://old.reddit.com/r/Amd/comments/10z1n8z/new_beta_bios_agesa_combov2pi_1208_offers/j82xuuf/

[Migz-DH]

Four months since I posted this, and nothing from AMD.

I'm definitely thinking of going back to Intel for my next CPU/chipset combo.

[tanazzz]

What is even more infuriating is that if you're running WINDOWS 11 you can't really roll back to 1.2.0.3 C because the intermittent stutters will come back. So the only way is to either use Windows 10 or buy a dedicated TPM module. Just ridiculous stuff. You either leave 5-10% single core performance on the table or up to 15% multi-core. This is genuinely horrendous customer support if AMD's gonna just keep ignoring this.

[thanatoast]

Previous available BIOS on my B450 Carbon is AGESA 1.0.0.6, which is SUPER old and cannot be rolled back to -  last time I tried it bricked the board and I would've been **bleep**ed if not for BIOS Flashback.

In the meantime while waiting for this to get patched (NOT LOOKING TOO GOOD), I explored all core OC instead, and now have dialed in settings that I probably would take over PBO OC any day...

It would be great if there was, like, y'know, ANY sort of response from *someone* at AMD about this.
Literally anybody. Getting sick of this constant information black hole we seem to find ourselves in.
Lord knows how many similarly annoying minor issues I've ran into over the years -  none of them get any attention.

[thanatoast · Accepted solution]

Bit late, but the newest AGESA 1.2.0.A fixed this issue, finally, after so long.
It also fixed the FIT Scalar which was also apparently broken...neato.

[tanazzz]

Still don't have it for my board (ASUS TUF X570 Plus). Hopefully it comes out soon.

[Migz-DH]

Thanks for that information!
That AGESA has been out since May, but the PRIME X570-PRO still has not been updated to it, so I will have to pester AMD Support to have the BIOS updated.

[Migz-DH]

I meant ASUS Support, not AMD. And, I contacted ASUS Support this morning to let them know I would like a new BIOS with that AGESA deployed when they get a chance.

[tanazzz]

They're extremely scared because of their recent BIOS mess-up so they might not even bother out of fear of breaking a platform that's already essentially over.

[Migz-DH]

In BIOS 4802, ASUS added AGESA 1.2.0.A to my board, PRIME X570-PRO, on 14 July, so I get to test out the AGESA on my system soon!

[tanazzz]

Still no 4802 for TUF Gaming X570 Plus.

[tanazzz]

Update: New bios dropped and I can confirm that CPU voltage boosts properly and is not limited to 1.40v like it used to be. I guess AMD's  CPU division actually care about products even if they're EOL unlike their GPU division.

[Migz-DH]

This should be my final post on the topic.

After installing BIOS 4802 with AGESA 1.2.0.A, my Ryzen 9 5900X's max-thread performance has improved by about 4.0%, and single-core performance is the same as before the update or slightly better.

Thank you, AMD, for finally solving this issue. I am pleased that it was finally addressed.