5800x (and other Zen 3 chips) PBO settings/Temperature fix

[The King]

My 5700x with Chrome and some app in the background (dont think it makes too much difference anyway)

View attachment 273245
View attachment 273246

Yeah weird stuff same PBO settings almost same score but showing 87W.

 

[izy]

Yeah weird stuff same PBO settings almost same score but showing 87W.
View attachment 273249

How is that possible? I see your 5800x is using 0.02v more your effective clock is lower and you probably have better cooling than me but the scores are the same.

Edit: Maybe the AutoOC / curve does something , i dont know.

Nice RAM OC , love to see it

 

[The King]

How is that possible? I see your 5800x is using 0.02v more your effective clock is lower and you probably have better cooling than me but the scores are the same.

Running a dual tower aircooler. Its performing ok at the moment because of the cold temps here. What cooling are you using?

Nice RAM OC , love to see it

Not one of my best bins have to run RCDRD 16 on this. but not much difference in performance if it was at 15.

 

[izy]

Running a dual tower aircooler. Its performing ok at the moment because of the cold temps here. What cooling are you using?


Not one of my best bins have to run RCDRD 16 on this. but not much difference in performance if it was at 15.
View attachment 273250

I am using ID-Cooling SE-224-XT, it does the job.

 

[RJARRRPCGP]

Rearranged the front case fans and got this on 11 22H2: (82C max, according to Ryzen Master, at the defaults) I noticed that when the temp gets higher, it seems to be boosting more. Windows build 22621, is indeed 11.

 

[izy]

Rearranged the front case fans and got this on 11 22H2: (82C max, according to Ryzen Master, at the defaults) I noticed that when the temp gets higher, it seems to be boosting more.

View attachment 273265

I am on win10 , what are your PBO settings / curve ? I will add some top exhaust on mine too not that i need it but why not

 

[RJARRRPCGP]

I am on win10 , what are your PBO settings / curve ? I will add some top exhaust on mine too not that i need it but why not

I haven't changed anything yet, except when I tried some different power limits.

 

[The King]

So I dropped EDC to 95A now I am getting 15.9K @ 85W

Dropped EDC to 90A and got even better score.

So if you wondering what happens at 85A you guess it score goes up again!

Anywhoo midnight here need to hit the bed. Hope this was all not just some weird dream.

 

[Ibizadr]

Later I post my results I'm on agesa 1.2.0.7 + 3800cl14

 

[DR4G00N]

So I dropped EDC to 95A now I am getting 15.9K @ 85W


Dropped EDC to 90A and got even better score.


So if you wondering what happens at 85A you guess it score goes up again!


Anywhoo midnight here need to hit the bed. Hope this was all not just some weird dream.

Run with Realtime priority and you'll get a more reliable and higher result, this goes for all CB versions.

On normal priority with my 5600 @ 4.65GHz I get around 11700-11900 but on Realtime I get around 12050-12100.

Dropping the EDC from 160A to 90A (stock) yielded a higher score on normal priority but with Realtime it didn't change.

 

[The King]

Run with Realtime priority and you'll get a more reliable and higher result, this goes for all CB versions.

On normal priority with my 5600 @ 4.65GHz I get around 11700-11900 but on Realtime I get around 12050-12100.

Dropping the EDC from 160A to 90A (stock) yielded a higher score on normal priority but with Realtime it didn't change.

Thanks for the tip but I always do all my runs in real-time only. Every CB run is only done in RT.
In Benchmate is set permanently to real time.

Refer to this post.

5800x (and other Zen 3 chips) PBO settings/Temperature fix

Oh i missed the DOCP/XMP stuff Like everyone else on ryzen, you just need to raise your SoC voltage. XMP raises DRAM voltage and sets the timings, but ryzen CPU's memory controller are only guaranteed to 3200 - above that, you're likely to need to raise the SoC voltage (especially for 4+ ranks)...

www.techpowerup.com

Your 5600 score is about right I get around 12108 with mines so good score. Only noobs do their run in normal priority mode. j/K

 

[Mussels]

You need a nice Thermalright cooler to fix you up. People say X3D is a hot chip.. nah not at all

x3D is freaking amazingly cold for what it is, it only warms up for AVX workloads

Isnt that EDC bug still around where once you pass a certain value, you can get a performance loss?

Effective clock is super important from HWinfo to tell if you're clock stretching, i can drop my voltage/temps a lot more on this CPU but it comes at the cost of effective clocks and therefore performance

 

[The King]

@Mussels I need some help here please.

Leave it to me to buy a 5800X the week before ZEN 4 launches.

So the issue I am having is that this CPU wont do PB0 +50 without bluescreening. Both my 5600 and 5600X can do PB0 +200 and -CO on top of that.
Stock perofmance seems slighly lower than the TPU review. So far I am not that happy with this CPU but it could be because it behaves differently from my 5600 and 5600X.

Here are the core ratings in HYDRA. With PBO disabled. So is it normal for a 5800X to not be able to do PB0 +50? I have tried +25 it can boot into windows but +50/+75 etc mostly unstable crashed before windows loads or after I log in. I am considering requesting a return and exchange on Amazon but before I do that I wanted to check with someone who has more experience with this CPU.

View attachment 262780View attachment 262781

Finally sorted out the PBO +200Mhz boost override issues I had on my B450.

 

[tabascosauz]

Hey guys so I was doing some testing with EDC today and I found some really weird results need
to verify with other 5800X users if this data is correct. Multiple runs I'm getting around +/-84W with 15K R23?

EDC 110A = 84w 15k

EDC 120A = 87w 15k

EDC 125A = 85w 15k

EDC 130A =84W 15K

I think you just found a new bug in AGESA

Safe to say 87W would not get you to 80C, and none of the effective clock is correct for 15k (4.0, 4.2, 4.6). Did you ever figure out what setting broke everything at once

 

[The King]

I think you just found a new bug in AGESA

Safe to say 87W would not get you to 80C, and none of the effective clock is correct for 15k (4.0, 4.2, 4.6). Did you ever figure out what setting broke everything at once

Not sure what is was. But did do a run with ECO mode 65W and 95W. These power figures should be more accurate than those. Temps are also lower here.

 

[Mussels]

Huh go figure, my x3D is only a tiny bit behind your 5800x

as you can see with the eco mode, the effective clock dropped a fair bit - secret is to slowly raise PPT until that stops raising and you've got your max limit, and see if the temps are good there

 

[Deleted member 185158]

Huh go figure, my x3D is only a tiny bit behind your 5800x

as you can see with the eco mode, the effective clock dropped a fair bit - secret is to slowly raise PPT until that stops raising and you've got your max limit, and see if the temps are good there

Yeah, mine too. But it's not an AMD rig unfortunately. I suppose this is an ill comparison. :/

 

[Mussels]

Yeah, mine too. But it's not an AMD rig unfortunately. I suppose this is an ill comparison. :/

View attachment 273908

definitely the weirdest zen CPU i've seen so far

 

[Deleted member 185158]

definitely the weirdest zen CPU i've seen so far

Humps along pretty good though. When looking, it was either that or a 5600X. Which still want to do, but for my sons rig. To replace a 2700x. Just not sure if he needs 16 threads or if 12 will do ok.

 

[MaddoggMiranda]

https://www.patreon.com/1usmus/posts?filters[tag]=HYDRA%20DOWNLOAD

1usmus Hydra is pretty awesome now for finding PBO Custom Curve as well as other things like OCing Radeon GPU's, Memory DRAM Calc etc.
You do have to subscribe to patreon to download it (until final release which will eventually be free but this could be a while yet) but its worth $9.50 as it works flawlessly for PBO in my findings. gained an extra 5%-10% performance over my manual PBO curve and it stress tests too. You can set the level of stress as well depending on your type of PC usage. Give it a try if you havent already its certainly worth a go for sure in my opinion. Hope it helps you all in more ways than one

 

[Diceair]

So I played around with my limits and curve optimizer, Found the following to work for me

ppt =142
TDC = 95
EDC = 110
max boost override 100mhz
curve on -10 best cores and -26 rest

I'm on the following specs
5800x
fractal design s36 360mm
3600mhz ram g.skill overclocked to 3733 cl16-16-16-19
asrock b550 steel legend
Palit RTX 3080 ti gamerock (undervolted)
Windows 11 with latest updates

My phanteks enthoo pro case I modified so I can add 1 more fan in front so layout is as follows
2x cougar vortex pwm 70cfm fans front, 1x arctic p12 pst front
fractal cooler top as exhaust
rear 120mm arctic p12
I took the stock rear and installed it at the bottom as intake
I also removed my pci slot covers that helped with gpu temps a bit.

Now I did a test last night with everything as it is at 100% fan speed I can get about 15850-15910 4620mhz on Cinebench R23 now I have 3 extra arctic p12 fans lef over from the 5 pack so decided to literally put them on top of my case with side panel closed as exhaust and have the cable coming out the pci slot covers. I then started with 1 then 2 then 3 fans and every time I add a fan my scores increased a little bit with 2-3 fans I got consistently over 16000 points with 3 fans or over 16050 +-4700mhz with same settings. I think if I can get them flush with the radiator it will be even better for my clock speeds.

Now I'm thinking how can I improve my temps maybe I should try push/pull on the radiator but don't know if it's a good idea to have different fans on top than on the bottom of the rad so bottom of rad stock fractal fans and top 3x arctic p12. Scared that might not be a good idea. The other thing is have to quickly look if I even did get enough screws to install another 3 fans on top.

TLDR: cooler = fractal s36, stock fans all fans in case and cooler 100% cbr23 = 15850-15910 4650mhz, Added 3x arctic p12 literally top of case full speed cinebench r23 = 16000-16050 +-4700mhz

 

[Mussels]

That's a pretty good setup overall

push pull is perfectly fine with mismatched fans, that's not an issue at all other than setting them to different fan speeds (So they cant all be on the one controller, obviously)

 

[Waldorf]

@Diceair
iinm the fractal are flow, and the arctic are pressure fans, so swap that the arctic are the intake pushing onto the rad,
and the fractal pull behind it.
according to arctic, the P fans are supposed to pull when used alone,
but are better in pushing air, when used in combo (with other fans pulling).

so far seems to be working better, as my temps are same/tick lower, while running less max fan rpm), after switching the P14 to push (280 copper rad, two F14 pulling.

 

[Diceair]

@Diceair
iinm the fractal are flow, and the arctic are pressure fans, so swap that the arctic are the intake pushing onto the rad,
and the fractal pull behind it.
according to arctic, the P fans are supposed to pull when used alone,
but are better in pushing air, when used in combo (with other fans pulling).

so far seems to be working better, as my temps are same/tick lower, while running less max fan rpm), after switching the P14 to push (280 copper rad, two F14 pulling.

yes seems like I don't have enough screws to install 6 fans on radiator

 

[Waldorf]

at least for the arctic, you might be able to use shorter screws between fan and rad,
and use the existing longer set to mount the exhaust fans from outside the case (rad/case/fans).
as long as they have the finer tread and fit the rad mount, even shorter ones should work,
mounting fans straight to the rad (no case metal between), sometimes drive/motherboard screws will fit.