AMD Zen Discussion Thread

X470 gaming plus max is also good

Amazon have had stock for a while on some of the 5000 series too

.G. wrote: »

s without frequent WHEA BSODs.

still nothing for the b550 tomahawk apart from the Beta bios that caused random WHEA shutdowns

Downgraded to the previous AGESA 1.1.0.0 and all back to normal

batmansDad wrote: »

Anyone have any experience using Q Flash on X570-I-AORUS-PRO-WIFI

There is no leds or indicators to confirm it's working or worked. While I can see the USB Stick light blinking for a few seconds the psu never starts so I have my doubts.

Have this motherboard and did this. My PSU did start.
My USB stick doesn't even have a LED, I am 99.9% sure there was a flashing light on the motherboard itself?

It definitely took 3+ minutes.

Maybe you didn't rename the file correctly or similar?

Found this random video that seems to show the whole process with another gigabyte board, hopefully it will help you.

https://www.youtube.com/watch?v=y9uGfIFufZI

As a separate question, anyone that has a X570 board. How do you find the fan noise from the PCH fan?

I'm waiting on components so haven't build the machine yet but when I was flashing the BIOS yesterday the PCH fan was loud!

I'm not 100% sure, but I think the QFLED only lights up once the flash has started. I used a USB with an indicator light and it was on for a few seconds then the QFLED came on. Presume that was the board reading the file then starting the flash. It sounds to me like in your case its not recognizing the bios file so not starting the flash.

Sounds like you have eliminated a lot of the variables. Were the USB sticks 2.0 or 3.0? I saw mention of people having problems with USB 3.0 when trying to flash. A small capacity USB 2.0 seems to be what most people have luck with

I followed that same vid... only thing I did slightly different is only using the 24 pin connector. I didn't plug in the the other 8 pin (manual says one CPU power connector should be plugged in as well but users on other forums had luck with just the 24 pin). Could try that

From this thread it sounds like the boards can be really picky about USB sticks...

https://www.overclockers.co.uk/forums/threads/help-q-flash-wont-work.18907064/page-2

Here's where they mention only using the 24 pin (Dawned on me there that I just pulled the 24 pin from my old machine and plugged it into the new board. Left all the rest of the cabled plugged into the old machine)

https://linustechtips.com/topic/1272021-the-q-flash-plus-on-my-x570-aorus-elite-wifi-isnt-working/

Seems to be total trial and error!

If you have no joy... you could try requesting a boot kit from AMD

https://www.amd.com/en/support/kb/faq/pa-105

CTR: A Review and a Warning

https://www.igorslab.de/en/clock-tuner-for-ryzen-2-0-tutorial-new-version-with-support-for-ryzen-5000-hybrid-oc-phoenix-mode-and-much-more/3/

Different article on CTR 2.0. They have a full guide on how to use it and have given it a glowing recommendation. Make no mistake its no 1 click tune like they said in the first article. I know this is what 1USMUS ultimately wants. Its still RC software and I know he has a roadmap to 2.2 or 2.3 and will make full use of curve optimiser.

EDIT
Seems to have gotten a 14.2% gain at full pelt. power efficiency not being considered etc.
for reference my 5950x gets 9900 CB20... his after the tune was 12292.

Just finished going through the whole process with CTR 2.0

My earlier reference to 9900 in CBR20 was with pbo enabled and curve optimiser too. So essentially you are required to put everything to stock then run the tests. Unfortunately I have a bronze sample chip. I started out at ~9500. It gave me ~10800 under profile 1.... which is basically as fast as it can go at 1.1V (CCX1- 4.325Ghz, CCX2- 4.15Ghz) and then profile 2 topped out at 11485 at 1.25v (CCX1- 4.575Ghz, CCX2- 4.425Ghz)

So I have left this on and with hybrid mode I should essentially have 3 profiles active... PBO, P1 and P2. one you hit the parameters for P1 you go to that profile and likewise for P2. So under heavy load on smaller amounts of cores PBO remains active, then medium high core count loads P2 is active and full load on all cores will engage P1. I'm delighted with the results.

Be aware do not use this if you use Heavy AVX work loads. I dont so im happy to run it like this. Like anything to do with OCing its a risk. But i like seeing what my chip is capable of... even if i was disappointed to see its a bronze sample.

H3llR4iser wrote: »

Yeah I've tried CTR myself, after seeing all the raving reviews from the various tech channels, and had the exact same results as the author of that article - multiple times. Please note - I DID read the manual, followed 1usmus' tutorial to the letter and let it run its course on a 3900x.

According to CTR itself, everything had gone smoothly - and looking at the report there was no reason to think otherwise. IIRC it applied a rather nice undervolting (around 1.25v, IIRC) and managed to "push" the cores in CCX1 to something like 4.35 Ghz, with the others between 4.2 and 4.15. Consider my CPU, stock, settles to 4.1 on all cores during load, this looked like a great result - CB20 reflected the improvement.

The system LOOKED like it was working fine, CB20 would run indefinitely without issues, normal use, a game etc; YET, I decided I wanted to make sure the whole thing was "rock solid" - and it was NOT by any definition - OCCT started reprting errors within 1 minute of testing and Prime95 small FFT test would instantly push the CPU to the thermal limit (Kraken X62, under stock conditions the CPU rarely reaches 75c under load, 80 with small FFT).

I've retried the CTR process multiple times, checking BIOS settings, double checking the instructions, every time with the same exact result.
Now, it could probably be made to work if one really throws a lot of time tweaking the process and resulting settings, but frankly I can't be bothered to do that and went back to the standard PBO setting.

The argument about the tool being talked about as a "1 click everyman OC solution!" is very relevant - and I haven't seen any of the tech channels perform a proper test; 1usmus replies under that article are frankly idiotic in tone, sounding like a preschooler - namecalling on the very first line, and then the icing on the cake - "To my great regret your processor didn’t burn out. ".

From Igorslab review

If I could find fault with anything, it would only be the persistent lack of an “AVX Heavy” test mode. Even though only a very small part of the user base might have a use for this, such edge cases can often become critical. If you start a heavy AVX workload like Prime95 small-ffts with AVX or LinpackXtreme, the system always crashes with an “AVX Light” tested profile. The only workaround is to increase the voltage for the P1 and P2 profiles by 75-100 mV and then manually test for stability for these cases. For the vast majority of games and applications you wouldn’t need it, of course, but if you did need it, you’d wish you’d known about it beforehand.

But Yuri already has this and many more items on his roadmap for upcoming releases mentioned at the beginning. CTR 2.1 with many features around the Ryzen 5000 boost innovations has even been unveiled and is available for testing for its Patreon backers already. So Ryzen owners can be happy about the new 2.0 release today and at the same time look forward to upcoming features that are already in development or testing. For such a commitment to the community one gladly leaves a Patreon subscription!

I completely agree with you about 1USMUS and his comments on that review. Should have said nothing IMO and let the reviews speak for themselves. Plenty of good ones out there. At the end of the day it is only a release candidate of the tool. Its not the finished article so to speak. For me this wasnt far off a 1 click tune. I had to reset my pbo and curve optimiser to stock... which is reasonable. Then just click diagnostic... then tune then save the profile. done. If you want the hybrid mode then tune again and save profile 2. Mine works exactly as intended. I would say Igorslab walk through is far more user friendly and gives you the warnings about the heavy AVX workloads. As I said in my first comment about my experience it suits my needs perfectly. It definitely isn't for everyone and you need to be careful about how you load the processor but then again it isnt the final product.

I think it could be the 1 stop solution. But not as it stands right now. And I feel its closer than you probably think it is but im no coder.. I think just extra profiles that identify the specific workloads could solve the issues.

I think its perfectly fine in most cases, gaming, general use, or score chasing... just not heavy AVX workloads. And thats enough for it not to be for general consumption by those its aimed at (people looking for easy overclocking), as they wont know why they are having issues and could be causing damage to their cpu in the process.

Ultimately I think the software should not be a Release Candidate... I feel its still very much BETA software. This would at least make more sense. If you want to use it you really need to educate yourself fully of all the caveats about it.

1USMUS really needs to be able to control himself a better too. I sympathise with him in one sense.... He's made this potentially really great software (for free, unless you count patreon) which many media outlets have been able to use with little to no issue and give it glowing reviews but then this one media outlet slate it for the issues they experienced when (in 1USMUS's opinion but who knows) they didnt use it properly. I'd probably be pissed about it too but im aware they are entitled to their opinion and this is indicative of what others may experience. But no way would I ever publicly talk to anyone like that. By all means answer them or whatever... but at least be professional about it and try help the situation... His attitude just makes it worse.

I agree and disagree with you... With regards to the PR campaign. These media outlets are there to give their consumers an unbiased and fair opinion. I'm not naive enough to believe this is always the case (recent Nvidia issues show this) but surely 1USMUS is far from someone they are concerned about burning bridges with(again the media pushing back against Nvidia would show this should be the case). And the damage it could do with their userbase due to a bias review would be far more detrimental to them. So as much as yes 1USMUS may be asking them to sell it that way... its their responsibility to give the truth of their experience. And i'm sure 1USMUS doesnt review or censor the media outlets in that regard. I actually went back and looked at the LTT video on it. They didn't review it, but gave impressions of CTR 1. They never said it was safe and mentioned having a few BSOD's during the process but their impressions of the software was about consistent with what I experienced with CTR 2.0 tbh.

I'm just not ready to slate his tool just yet. I obviously went and used this software and was fully aware of all the issues I could encounter and as such I encountered 0 issues. I also knew not to use it and run heavy AVX workloads (CTR 1 had AVX heavy testing you could do so its likely on the way). I researched all of this before I used it. The information is there And in all honesty anyone who decides to overclock should research what they are doing. 1 stop solution or not if you are taking the risk you need to know the risk. Same way before I enabled PBO and curve optimiser I researched the reasonable settings to start at and could push it from there as I wished. At your own risk is always the bottom line when you overclock. Overclocking and safe dont exist together which is why even when you use AMD's own tools to overclock you void your warranty. Ryzen Master literally displays this every time you open it, my mobo gives me the notice every time I go into the AMD overclocking utility and so does CTR. You cant hold this software to a different set of rules.

100% agree that if 1USMUS is critical of others he needs to accept it when he gets it in return. And if he feels compelled to respond he should at least show a certain level of professionalism about it and just address the issues. Im not completely familiar with all the things he has criticised in the past ie PBO and ASRock (im a fan of PBO and curve optimiser). I have seen the little digs he's made with regard to the lack of support from AMD to help his development but did state that they provided him a few CPU's for testing. His response was awful to that review and made him look like a defensive child. My only point about him was that this appears to be more than just a job to him and is clearly a labour of love. Dont get me wrong there is no excuse for the way he dealt with it but he is clearly very emotionally invested in it and cant remove his emotions from it when he is getting criticised, as a professional should.

Slutmonkey57b wrote: »

I think a key difference is that CTR 1 was built largely on open source code, and some contributors were not pleased with 1usmus' lack of clarity about the code, and his refusal to honour some of the licensing agreements, supposedly because it was "beta" and therefore he felt he didn't have to. (shaky ground imo) CTR 2 is supposedly rewritten from scratch not to include any open source code, and he's clearly looking for commercial success with it.

But his conduct has gotten shadier the longer the project has gone on, and this stunt to me is a final nail in the coffin : he's skirted up to the edge on a number of issues but his reaction to this very serious flaw is proof that without the open source code he's stuck, and the quality of his own work isn't reliable, certainly not reliable enough to trust your investment with.

Very fair.

I dont think anyone can recommend it for general use yet. If at all, if the coding is as shady as you suggest. Again this isnt something I'm as aware of. PBO and curve optimiser is the way to go for overclocking.

But I suppose if you do decide to use it dont take CTR 2.0 as a beginner friendly utility at all. In fact, you should be more aware using this than any other utility as this has reported issues, serious issues at that! other utilities have far more fail safes involved. Anyone interested in using it should be very aware before even thinking about using it. If you do decide to use it I do suggest Igorslab guide rather than 1USMUS's. As I found it to be the easiest to follow and gave the best/ most clear warnings about its issues.

cromelex wrote: »

Is there really even a point in using this for Zen3?

From my own experience, with a X570 board and a 5600X, I just had to turn on PBO2 and auto overclock and test a few different settings on the Curve optimiser.


4840mhz boost single core, 4764mhz boost all cores, and it's super stable. Like, no crashes on benchmarks and even Cyberpunk stable (can play it without crashing!). Fair enough, it take a few more clicks, but it is still very simple. Only voltage I have ser manually is the RAM.

Post where I shared previous results months ago.
https://www.boards.ie/vbulletin/showpost.php?p=115410945&postcount=3281

Right now, no I would not recommend this to anyone. There is so many things you need to be aware of while using it. And if you make a mistake you really can damage your CPU.

Stick to PBO and curve optimiser. You should get a little bump and save a little power too if you set it up right. More importantly you can just not think about it after that.

Aodhan5000 wrote: »

I thought PBO2 was better than PBO because the undervolting drops temps allowing higher clocks and it doesn't void your warranty but PBO does. I reckon Zen 3 just PBO2 is the way to go.

If you overclock your CPU in anyway AMD do not have to stand over the warranty. This includes PBO2. I get a big warranting about this before I can enable this in my BIOS. Now that being said I have no idea of anyone who tried to return a chip after causing damage this way, so in practice I have no idea. But officially you void your warranty if you use PBO2. Unless this has changed someone else may know more than I do.

Aodhan5000 wrote: »

Ya just double checked it there on and you're dead right, PBO2 voids warranty as well. Really curious about how they check if you enabled it or if it's just a case like intel saying XMP voids warranty but having no way to check it so they'd just be relying on your word

Yea that exactly why i mentioned that I have no idea how it works in practice. But I think it will come down to if you cause damage... the damage caused might very clearly be from overclocking and not knowing what you are doing. But if the issue isnt clearly overclocking related they might honour the warranty but Im just speculating here... I have no idea how it really works.

I've tried CTR and it defo improved my scores over stock while also dropping voltage. My 5900x under stock BIOS settings with PBO on auto regularly asks for and receives 1.5v from the board which really is not healthy if everything I've read is accurate. With CTR I can get better all core scores, stable performance and lower voltages at higher all core clocks but it still allows PBO to operate in low usage scenarios so the chip is still taking 1.5v when that happens. It boost to 4950mhz on two cores under PBO.

All that said I was able to get much better scores myself playing around with the scalar settings, curve and the boost overhead in the advanced PBO menu but I can never get any of them stable so It's just left on auto. Don't know if its sh!tty bios I have or a poor chip, last bios update certainly improved things, couldn't use +200mhz or the curve at all on the previous version, but its still not great.

I finally updated my BIOS and enabled the base PBO2 features. Got a decent enough bump, all-core settles at 4.525GHz and have 7 of them regularly boosting to between 5-5.1GHz under normal use (previously on PBO1 a single core would make it that far, with the rest between 4.7-4.8).

_CreeD_ wrote: »

I finally updated my BIOS and enabled the base PBO2 features. Got a decent enough bump, all-core settles at 4.525GHz and have 7 of them regularly boosting to between 5-5.1GHz under normal use (previously on PBO1 a single core would make it that far, with the rest between 4.7-4.8).

im on a 5950x and i get nothing like this... what monitoring software are you using and settings? i have seen spikes of 5.1 (just below this) and all core will be 3.9+ but never 4

Managed to buy a 5600X from AMD direct.

Going to run some benchmarks, etc for the next week or 2 to have data for comparison.