Ryzen 5900X overclocking help

Slightly better results. Now testing -10 on first core and -25 on all other cores.
CBR20-5-Feb.jpg

Also passed 10K in CPU Z MC. https://valid.x86.fr/h792y4
And slightly higher Geekbench score. Don't have AIDA since trial expired. https://browser.geekbench.com/v5/cpu/6318357
 

Attachments

  • CBR20-5-Feb.jpg
    CBR20-5-Feb.jpg
    146.1 KB · Views: 0
so got me system put together last , got all the new installed bugs out , biggest one was i never realized that the Crossfire Dark Hero viii doesn't have a pcie 1 slot above the gpu slot for sounds cards so i had to make a my sound car fit with a gpu riser cable ugh. but anyways it is hell tweaking this thing compared to my 9900ks @ 5.1 . so far got to 4.7 @1.34volts and tweaking kept memory at stock so far 3600mhz cl 16 32gb. like that pbo was boosting often hated that it is slower reacting than all core overclock. going to test another week so if i can make some gaming gains on my old 9900KS

Things I like
1.Water cool loop sits at 31c now benching gaming etc , 9900ks would hit up to 33c gaming 35c benching
2.multicore performance is beast mode
3.Memory is easier than Ryzen 2
4.Higher 3dmark benches 6 percent to be exact
5.Pci 4.0 makes my ssd scream ,
6.Love the 24 cores for when you just all over the place

Things I hate
1. Hard to hit 166mhz at 2560x1440 9900KS does it consistently and easily
2. Heaven Benchmark 3 percent slower
3. Valley Benchmark 10 percent lower
4. Tuning is a bitch pbo , pbo 2 , all core overclock or stock there no correct answer to get performance out of these things , and to prove it go look for guides there none complete not even rog itself has done one.
5. No new x570 boards have the pcie x1 slot on top of board , most people using riser cards these days that make pci slots unusable. being board makers they should keep up with trends, and onboard supreme 1220a sound still sucks if your using high end speakers " yes i have a dac

 
PBO will give you better overall results in the real world. For synthetics all core might yield higher MC results.
 
What does the curve optimizer actually do?
It changes voltage by between 3 and 5 millivolts per step, depending on load. For example, if you set a core to be negative 10 in curve optimizer, it will reduce the voltage by 50mV at idle and 30mV under load.

This reduces heat generated if power consumption is capped in PBO and leads to a bit higher clocks. If power usage isn't capped the cores clock much higher until the PB (not PBO) algorithm tells it to stop based on temperatures, silicon quality, and power usage.

So generally, lower heat, less power used, and higher clocks. All good stuff. It's just a bit of a pain in the butt as if you go too low you'll see WHEA errors and reboots, and you want to set your best/second-best and remaining cores to different values.
 
It changes voltage by between 3 and 5 millivolts per step, depending on load. For example, if you set a core to be negative 10 in curve optimizer, it will reduce the voltage by 50mV at idle and 30mV under load.

This reduces heat generated if power consumption is capped in PBO and leads to a bit higher clocks. If power usage isn't capped the cores clock much higher until the PB (not PBO) algorithm tells it to stop based on temperatures, silicon quality, and power usage.

So generally, lower heat, less power used, and higher clocks. All good stuff. It's just a bit of a pain in the butt as if you go too low you'll see WHEA errors and reboots, and you want to set your best/second-best and remaining cores to different values.
I started on - 5 and the computer crashes first minute what am i doing wrong ?
 
It changes voltage by between 3 and 5 millivolts per step, depending on load. For example, if you set a core to be negative 10 in curve optimizer, it will reduce the voltage by 50mV at idle and 30mV under load.

This reduces heat generated if power consumption is capped in PBO and leads to a bit higher clocks. If power usage isn't capped the cores clock much higher until the PB (not PBO) algorithm tells it to stop based on temperatures, silicon quality, and power usage.

So generally, lower heat, less power used, and higher clocks. All good stuff. It's just a bit of a pain in the butt as if you go too low you'll see WHEA errors and reboots, and you want to set your best/second-best and remaining cores to different values.

Thank you, so set voltage to Auto, then use curve optimizer ?
 
Yes, set voltage to auto then use CO to individually undervolt each core.

If your CPU crashes at -5 then you might just have poor silicon. It's a gamble. But make sure you're using AGESA 1.2.0.0, it's much more stable.

Generally you want to use the PBO menu under "AMD Overclocking".
 
Yes, set voltage to auto then use CO to individually undervolt each core.

If your CPU crashes at -5 then you might just have poor silicon. It's a gamble. But make sure you're using AGESA 1.2.0.0, it's much more stable.

Generally you want to use the PBO menu under "AMD Overclocking".
At least on MSI you want to NOT use the PBO menu under AMD Overclocking.
 
Did they release the tomahawk 1.2 bios? Been waiting for it.
1.2 is out. I swear I was checking for it everyday. The site shows release date as of January 26, 2021. LOL.... Give it a go (if you haven't already), I might upgrade my kids PCs on the weekend.
 
Yes, set voltage to auto then use CO to individually undervolt each core.

If your CPU crashes at -5 then you might just have poor silicon. It's a gamble. But make sure you're using AGESA 1.2.0.0, it's much more stable.

Generally you want to use the PBO menu under "AMD Overclocking".
thanks man im try when I get home , I updated to AMD AM4 AGESA V2 PI 1.2.0.0 this morning but never tried the curve optimizer this morning , i was on a older bios when i tried. is -5 the lowest setting ? should i try individual cores , I did all core
 
1.2 is out. I swear I was checking for it everyday. The site shows release date as of January 26, 2021. LOL.... Give it a go (if you haven't already), I might upgrade my kids PCs on the weekend.
After spending so much time optimizing the living crap out of my PC I am extremely weary of trying out the new bios lol.
 
thanks man im try when I get home , I updated to AMD AM4 AGESA V2 PI 1.2.0.0 this morning but never tried the curve optimizer this morning , i was on a older bios when i tried. is -5 the lowest setting ? should i try individual cores , I did all core
Individual cores are better since you can avoid WHEA errors through it. My second best core doesn’t run w/ anything below -10 whereas all other cores are right now functional at -25. Fingers crossed since I still need to wait 3-4 days for that idle WHEA error. Does someone know best way to test for idle WHEA errors other than waiting days lol.
 
After spending so much time optimizing the living crap out of my PC I am extremely weary of trying out the new bios lol.
ahhaha.... I don't blame you :)

BTW.... I'm still on AGESA 1.1.0.0 BIOS. So I'm 2 releases behind. My kids will kill me if I break their setups... LOL
 
Last edited:
thanks man im try when I get home , I updated to AMD AM4 AGESA V2 PI 1.2.0.0 this morning but never tried the curve optimizer this morning , i was on a older bios when i tried. is -5 the lowest setting ? should i try individual cores , I did all core
All core is good to start out with to find where to start. If you say do -5 and it's stable, but -10 isnt then start with all cores at -5 and then tweak individually. It's a laborious process and hopefully CTR 2.1 with auto core optimizer will help with that.
 
After spending so much time optimizing the living crap out of my PC I am extremely weary of trying out the new bios lol.
I updated mine to 1.2.0 and my previously stable ram configuration now hangs windows shortly after boot. Need to start over from scratch it seems...:’ and I haven’t gotten to it yet.
 
I have only heard problems with 1.2.0.0 bios on most motherboards. Glad I didn’t indulge thus far. Will wait for the next one. No need to break what’s working.
 
Realized I had undervolted the wrong core -_-. Now I'm -5 on best, -10 on second best, -20 on others, +200 boost, auto scalar. Can't seem to push any of the "worse" cores past 20 and I'm too lazy to figure out which core can't handle it. Pretty happy so far, seeing boosts up to 5025 on 8 cores and the other 8 are boosting to 4.8-4.9.

CBR20 Results:
11039620

CPUZ Results:
12941 and 672
 
Best and second-best should be set exactly the same; the Windows scheduler will switch single-thread tasks between the best and second-best cores on CCD0 to even out heat distribution. So if you're stable with -10 on second-best, it should work on best too.
 
I have only heard problems with 1.2.0.0 bios on most motherboards. Glad I didn’t indulge thus far. Will wait for the next one. No need to break what’s working.

I tried the beta bios weeks ago, CPU would not boost over 3.3ghz (all core), regardless of my settings, went back to the last non-beta bios which is 1.1.8 I believe, and it's back to normal, Asus just released a non Beta 1.2.0.0, so I may try that one.
 
OK, question, in HWinfo, it shows Effective clock and (perf), which one shows actual clock speed?

Effective clock for me I think has never broken 5ghz, usually 4.8-4.9 is the highest It will go, where as PERF has hit 5.15 on my "best cores"

When running a 1-2 core workload, the effective clocks match what Ryzen master shows, and perf shows a higher number. I am confused :)

(red = perf)
(yellow = effective)

1612677190089.png
 
I don’t think effective clocks are correct representation of actually what the processor clocks are at a point in time. I just use the top one to monitor my processor clocks.
 
OK, question, in HWinfo, it shows Effective clock and (perf), which one shows actual clock speed?

Effective clock for me I think has never broken 5ghz, usually 4.8-4.9 is the highest It will go, where as PERF has hit 5.15 on my "best cores"

When running a 1-2 core workload, the effective clocks match what Ryzen master shows, and perf shows a higher number. I am confused :)

(red = perf)
(yellow = effective)

View attachment 326923
Ive heard that you have to look at effective , perf is only when you proc hits for a sec , but i think it doesn't work correctly
 
What do you guys think about this program ClockTuner 2.0 for Ryzen (CTR) Guide , The writer of the program seems really knowledgeable , seems like its a program to make gains without taking you out of Amd specifications , " you can change perimeters to overclock if want to

https://www.guru3d.com/articles-pages/clocktuner-2-for-ryzen-(ctr)-guide,1.html
I think the internet has mixed feelings about it. Some people swear by it but others think it's easier to just use PBO. Maybe when 2.1 comes out and it can do auto core tuning.

https://www.reddit.com/r/Amd/comments/ld2qec/chips_and_cheese_ctr_a_review_and_a_warning/

I tried it but I wasn't too confident in it and I felt like I had tweaked my PBO2 settings pretty heavily that I wanted to stay with it.
 
When I leave my computer mining and have hwinfo64 up, it crashes after some time(hours), with it closed it will mine for days :/ heads up
 
I updated mine to 1.2.0 and my previously stable ram configuration now hangs windows shortly after boot. Need to start over from scratch it seems...:’ and I haven’t gotten to it yet.
The updated bios (asus hero WiFi 3204) seems to have added another menu for memory settings and voltage. Some of the settings in the main extreme tweaker section don’t seem to work anymore.

after playing with it a bit, it seems Agesa 1.2.0 is more stable that 1.1.X - I can run at IF 2000 for about 9 minutes of cinebench before it crashes (it used to crash just launching). Still trying to find out where its actually stable though
 
For anyone with the dark hero where is pbo2 vs pbo in the bios? I've been using the pbo tab in amd overclocking section and not the one in extreme tweaker. Is that correct? Pbo2 has the curve optimizer right?

Also, I'm curious why people are putting a higher negative value on their bad cores vs the good cores. Wouldn't you want more voltage to the bad cores to keep them stable (ie a smaller negative value)? So instead of negative 10 on good core and negative 20 on bad, vice versa.

I'm currently running an all core -10 with CO and would like to dial it in more if possible.
 
My experience has been the opposite. Pushing a higher negative gets cores to higher frequency. My only second best core was the first one (core 0) that used to throw a WHEA error after 3-4 days.

Since I set it to -10 and all remaining cores to -25 I see my all core boost routinely to 4750 in games. Sometimes even 4800 (but usually only a few seconds). System has been running stable since.

I personally think the good and bad core is all bullshit. You need to see the core that is throwing a WHEA error and give it more voltage to stabilize (which intuitively also makes sense).

I am waiting for a more stable version of bios so I can try a higher IF than 1867.
 
I personally think the good and bad core is all bullshit. You need to see the core that is throwing a WHEA error and give it more voltage to stabilize (which intuitively also makes sense).

That is what I was thinking, so the core that throws a WHEA error you would give it less negative offset right hence the -10 versus -25 on the rest. How did you identify which core was causing the WHEA error?
 
That is what I was thinking, so the core that throws a WHEA error you would give it less negative offset right hence the -10 versus -25 on the rest. How did you identify which core was causing the WHEA error?
It’s in this thread. And yes if core is not stable you give it more voltage (a lower negative offset). Someone posted about the core ASIC ID in the event viewer. For me it was easy as core ASIC ID was 0.

Core 0 will have ASIC ID 0 and 1, core 2 would have 2 and 3 and so on. However, some ASIC ID are missed so you may see ASIC 25 or 26 even though it should’ve gone to only 22 and 23 for last core. That might be trial and error to identify.

The other extremely cumbersome method is to use OCCT single thread test and set core affinity to the one you think has the problem and look for errors. It didn’t work for me since my core had no problem running load. It had problem running idle and with low voltage it would boost higher than it should and crash.
 
My suggestion tune memory. Ignore the rest.
After my experience on 3700x this is absolutely how I’ll be doing with my 5900x.
Get the highest 1:1 fclk to dram setting that is stable, tighten timings, and move on. These babies are fully optimized out of the box anymore.
 
Back
Top