AMD Curve Optimizer any guides / experience

[Octopuss]

But if raising the limit by 200MHz requires lower CO offets, doesn't that mean the cores would only hit lower frequencies than before?

 

[AVATARAT]

But if raising the limit by 200MHz requires lower CO offets, doesn't that mean the cores would only hit lower frequencies than before?

Your CPU will manage lighter load at high frequency, your high load will be the same.

 

[Octopuss]

It's just confusing from defaults point of view.
If lowering the core voltages by negative offsets lets the cores clock higher, increasing the voltages doesn't seem logical to be beneficial for frequencies even if the "cap" is removed. I just can't wrap my head around this.

Btw I couldn't google the max boost frequency anywhere. AMD's website only says 4,7GHz without any details.

So I ran two rounds of iterations of 6m Huge FFT size and these are the maximum frequencies:

Do you think I should increase the limits or the CPU is not good enough?
The current offsets are 30, 15, 20, 15, 30, 25, 15, 25.

 

[AVATARAT]

PBO, +200MHz, you probably need to lower current offsets.

Or something like this, because your mobo is like main (with bios 1602):

 

[Det0x]

Maybe this is the wrong thread for this, but could not fine one dedicated to the 5800x3d (?) Let me know and i will delete the post if there is a problem

Have finally completed my first game comparison between my maxed 5950x, my 5800x3d and golden samples Alderlake cpus @ ~5.6 7200MT/s

https://imgur.com/a/TZdNAzm

vs

https://imgur.com/a/ahnY3nl

Shadow of the Tomb Raider: 1080p lowest:

• 5950x @ ~5100/5000mhz = 353fps average cpu game
• 12900k @ 5750mhz 4300MT/s CL14 = 373fps average cpu game
• 5800x3d @ 4450mhz = 402fps average cpu game

Horizon Zero Dawn: 1080p performance preset, lowest res scale:

• 5950x @ ~5100mhz = 301fps average cpu game
• 12900k @ 5580mhz 7160MT/s CL30 = 321fps average cpu game
• 5800x3d @ 4450mhz = 313fps average cpu game

F1 2020 1080p low dx11: Australia benchmark location and dry weather:

• 5950x @ ~5100mhz = 490 average fps
• 5800x3d @ 4450mhz = 555 average fps

Farcry6 1080p ultra, HD-texture enabled, FSR QTY:

• 5950x @ ~5100mhz = 163 average fps
• 12900k @ 5580mhz 7160MT/s CL30 = 203 average fps
• 5800x3d @ 4450mhz = 176 average fps

Cyberpunk 2077: 1080p low

• 5950x @ ~5100mhz = 252 average fps
• 12900k @ 5580mhz 7160MT/s CL30 = 304 average fps
• 5800x3d @ 4450mhz = 268 average fps

Final Fantasy XV 1080p low: (game engine limited, results with grain of salt)

• 5950x @ ~5100mhz = 23426 score
• 12900k @ ~5700mhz(?) = 23585 score
• 5800x3d @ 4450mhz = 23489 score

Final Fantasy XIV: Endwalker 1440p maximum: (game engine limited, results with grain of salt)

• 5950x @ ~5100mhz = 30553 score
• 12900k @ ~5700mhz(?) = 33891 score
• 5800x3d @ 4450mhz = 33764 score

HardwareLux Counter-Strike: Global Offensive benchmark settings: (only scale with clockspeed it seems)

• 12900k @ 5500mhz 4133MT/s CL16 = 954 fps
• 5800x3d @ 4450mhz = 716 fps

Max tuned Alderlake beats out Zen3 in games, but i have to say i'm pretty impressed by the5800x3
...Considering zero binning/golden samples are required for 5800x3d and they dont scale with memory, so you can use a cheapo x470 together with 3200/3600MT/s memory for nice "low-end gaming machine".

edit
Added some of the Alder lake screens:


*edit2*
My old Spectre install was bloated at this point, so i made a new win10 install for the 5800x3d runs..
(earlier when i did the 5950x runs the spectre install were pretty slim)

And i have no control over the Alder Lake windows installs, but those runs are cherry picked super golden samples with the highest scores i could find.. (binned cores+binned IMC+super cooling+best of the best memory) Anyone feel free to improve the AL numbers if you can

For this to be a even more fair comparison of "maxed out gaming systems" i would need a new motherboard with an external clockgen so i could run the 5800x3d at ~4900mhz instead of 4450mhz, but i dont think that would change the outcome much in the end either.. Alder Lake should still come out on top.

 

[Mussels]

I have just realized I never touched the PBO limit.
Does increasing it by the usual +200MHz have any negative implications?
What exactly does it do anyway?
And what is the default frequency cap of 5800X?

Adds 200Mhz to the low core count boost, from 4850Mhz to 5050Mhz on the 5800x - If there is thermal and power room to spare

If you're undervolting with CO, you may need to reduce it for stability

 

[Ibizadr]

Adds 200Mhz to the low core count boost, from 4850Mhz to 5050Mhz on the 5800x - If there is thermal and power room to spare

If you're undervolting with CO, you may need to reduce it for stability

I'm at +200 with CO. And 4 of my 5800x cores goes to -30

 

[tabascosauz]

But if raising the limit by 200MHz requires lower CO offets, doesn't that mean the cores would only hit lower frequencies than before?

It's just confusing from defaults point of view.
If lowering the core voltages by negative offsets lets the cores clock higher, increasing the voltages doesn't seem logical to be beneficial for frequencies even if the "cap" is removed. I just can't wrap my head around this.

Btw I couldn't google the max boost frequency anywhere. AMD's website only says 4,7GHz without any details.

So I ran two rounds of iterations of 6m Huge FFT size and these are the maximum frequencies:
View attachment 246070
Do you think I should increase the limits or the CPU is not good enough?
The current offsets are 30, 15, 20, 15, 30, 25, 15, 25.

Like I said, you'll just keep fruitlessly going in circles unless you either enable Snapshot Polling or look at Effective metrics. These default "Core Clock" numbers with 25MHz granularity are pointless data. I can see 5050 on "core clock" when I know my 5900X barely manages 4930MHz effective on a good day.

Once you do that, test with +0 and see after corecycler whether you are capping out at exactly 4850MHz. If so, then you have additional headroom on those cores by expanding boost override.

Stock 5800X boost ceiling is 4850, 5900X is 4950 and 5950X is 5050. It's not advertised explicitly, but all Vermeer and Cezanne CPUs have their Global limit set 50-150MHz above their official rated boost.

Setting +200 doesn't hurt anything. If your cores are 4800MHz duds, then nothing happens. If your cores are excellent, then worst case you may need to dial back your undervolt a little bit, but you won't know until you test

Maybe this is the wrong thread for this, but could not fine one dedicated to the 5800x3d (?) Let me know and i will delete the post if there is a problem

Interesting results particularly for 12900K. Might gain more traction to start a new thread for this in General Hardware - I'm sure many people are very interested in seeing more of these comparisons between these 3, it's a hot topic and certainly deserves to have its own thread

 

[Makaveli]

I have a golden sample 5800X and using 200+ with these CO settings

 

[Det0x]

Interesting results particularly for 12900K. Might gain more traction to start a new thread for this in General Hardware - I'm sure many people are very interested in seeing more of these comparisons between these 3, it's a hot topic and certainly deserves to have its own thread

Some more numbers, i found the framtimes very interesting..

Metro Exodus Benchmark: 1080p low:

• 5950x @ ~5100mhz = 322 average fps
• 12900k @ 5580mhz 7160MT/s CL30 = 408 average fps
• 5800x3d @ 4450mhz = 338 average fps
• 5800x3d @ 4450mhz, dx11 = 524 average fps

Metro Exodus Enhanced Benchmark: 1080p low:

• 12900k @ 5580mhz 7160MT/s CL30 = 250 average fps
• 5800x3d @ 4450mhz = 237 average fps

The frametimes for the 5800x3d look much better, no ?

 

[Octopuss]

Snapshot Polling or look at Effective metrics

Where do I find that?
I use HWiNFO for information.

 

[tjtremor999]

Where do I find that?
I use HWiNFO for information.

You can use boosttester to see max clocks. It's low load array shuffle.

I wouldn't tweak CO with high offset like +200. Not worth much since it only hits 1 or 2 cores at most and will get very unstable.

On my 5900x, +150 will get 5.1 boost in games on two best cores and it's not stable past CO-6.
With +100 offset, I get 5050 for best two cores and can go CO -19

After the best core, the others will always boost -25mhz,-50mhz,-75mhz etc even with +200 offset; so there's nothing there to gain on dual ccds.

Not sure about 5800x but boosttester will show you the max possible freq for all cores.

 

[tabascosauz]

Where do I find that?

f

 

[tjtremor999]

all core Y-cruncher (options 7) seem to catch errors very fast vs Prime95 or corecycler that can pass for days with unstable CO.

I'm currently at -19 -19 -19 -6 -19 -18 -14 -14 -15 [] []-19 for y-cruncher. Two more cores to test & I can run Prime95 blend & corecycler.
210w and +100 offset

I think it's mostly a waste of time on my 5900x b2. It already scores high with low temps.

Here's 4 cores -6 & 8 cores -12.

 

[HD64G]

I used the Ryzen Master that tested my new 5600 and concluded that the optimal values are -27-28 for all cores apart from the 3rd one that managed -22. I put manually -25 for the strong ones and -20 for the weak one and added +150MHz for the PBO increased limit at the default 76W value and all good for now.

 

[Mussels]

I used the Ryzen Master that tested my new 5600 and concluded that the optimal values are -27-28 for all cores apart from the 3rd one that managed -22. I put manually -25 for the strong ones and -20 for the weak one and added +150MHz for the PBO increased limit at the default 76W value and all good for now.

Oh yeah i forgot they added that, now i'm on the new AGESA i'll let it test that overnight

 

[Octopuss]

So what am I doing wrong if no cores are boosting over 4850MHz despite adding +200 limit?

edit: WTF, I have just gone in the BIOS and found the power limits changed to manual and the frequency override disabled. I didn't change anything.
Looks like a week of testing wasted, because the thing decided to change itself.

 

[Mussels]

So what am I doing wrong if no cores are boosting over 4850MHz despite adding +200 limit?

edit: WTF, I have just gone in the BIOS and found the power limits changed to manual and the frequency override disabled. I didn't change anything.
Looks like a week of testing wasted, because the thing decided to change itself.

ryzen master can change those settings if you used its test settings

RM has mine set to -30 but the voltages didnt actually change, CO is broken on my current beta BIOS

 

[AVATARAT]

So what am I doing wrong if no cores are boosting over 4850MHz despite adding +200 limit?

edit: WTF, I have just gone in the BIOS and found the power limits changed to manual and the frequency override disabled. I didn't change anything.
Looks like a week of testing wasted, because the thing decided to change itself.

Adrenaline video driver can change it. Be careful, it's known problem.

 

[Octopuss]

I don't use Ryzen Master (tried, looked like bloatware), and I don't have that option in Adrenaline.
No idea what happened there.

And the stupid ass CPU just doesn't boost past 4800MHz anymore no matter what I do. Doesn't matter if I add 200MHz to the limit or not.
I even have slightly lower (actually, higher since negative) offsets on a few cores, and it's actually worse.

I updated the BIOS to the latest version before I started messing with all this again too, hoping I'd get better overclock, but no.

 

[HD64G]

ryzen master can change those settings if you used its test settings

RM has mine set to -30 but the voltages didnt actually change, CO is broken on my current beta BIOS

I think if one want to test with Ryzen Master the same settings altered there need to be at auto in UEFI. My PC stopped booting because I already had applied the CO in Ryzen Master and thought that without deactivation there I could do the same in UEFI. It seemed that the settings added to the previous ones and V was so low for the CPU that didn't booted at all.

 

[Mussels]

I don't use Ryzen Master (tried, looked like bloatware), and I don't have that option in Adrenaline.
No idea what happened there.

And the stupid ass CPU just doesn't boost past 4800MHz anymore no matter what I do. Doesn't matter if I add 200MHz to the limit or not.
I even have slightly lower (actually, higher since negative) offsets on a few cores, and it's actually worse.

I updated the BIOS to the latest version before I started messing with all this again too, hoping I'd get better overclock, but no.

Theres only been two AGESA's since the 5800x3d came out, and it seems like the regular X gets treated like the 3D with some limits

That said... i like the lower temps, and i see no performance loss.

 

[Octopuss]

I think I was on a BIOS from late 2021.

 

[oobymach]

I don't use Ryzen Master (tried, looked like bloatware), and I don't have that option in Adrenaline.
No idea what happened there.

And the stupid ass CPU just doesn't boost past 4800MHz anymore no matter what I do. Doesn't matter if I add 200MHz to the limit or not.
I even have slightly lower (actually, higher since negative) offsets on a few cores, and it's actually worse.

I updated the BIOS to the latest version before I started messing with all this again too, hoping I'd get better overclock, but no.

Some bios versions are better than others, see if you can find 1.2.0.3c for your board, that's the best one for 5000 series (except the 5800X3D)

Also the boost is affected by PPT, TDC, and EDC so you may need to adjust PPT, TDC, and EDC to achieve higher clocks, and don't bee too aggressive with values, higher isn't always better especially TDC and EDC.

 

[Octopuss]

I don't want to mess with that, I just set limits to motherboard and that's it.
Also not interested in using old BIOSes with bugs. I'm just curious why is significantly newer BIOS with numerous improvements over the course of months affecting performance in negative way.