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

[ViperXTR]

I'm running this config for months now. Had to adjust two times because of a WHEA but never again. It's running fine and stable, I ran long cb23 and prime sessions to double check. No problems at all. I do not know what Rev4 offset is, so I guess not

If its stable on idle and on load i guess its fine. Offset mount on our arctic liquid freezer is basically an optional feature for Ryzen 3000/5000 cpus where the core of the CCD is offset the main package chip, using the offset mount makes the cold plate on the center so it will cool optimally.
You can take a look at the manual and/or under your arctic cooler box and see what revision it is. The 4th revision is the best one so far for Zen3, if you dont have it, i think you can just ask arctic about it and they will send you rev4 kit.

 

[St54Kevin]

Oh yes I'm using the offset mount of my mounting system. I need to check the revision though

Okay now Im very confused. I think im not using the offset. There are different manuals online and videos mounting the cooler differently despite using 3000/5000 series. Does it need to go towards the GPU or away from it?

So it was indeed mounted in the standard holes (away from GPU) I checked again and gained about 3C on every test except in prim95 where its the same, at least something, right? Thanks for pointing it out again

Well of course, higher clocks need higher voltages

No one knows, every chip is different - and every core on every chip is different

Of course, but the thing with CO on my system is, with lower voltage I get higher boost clocks. I just tried to just up every CO value by +20 and did my tests again: 2-3C highter temps but in all tests nearly 150Mhz less clock. I may have already found my CPUs sweetspot

 

[Mussels]

Of course, but the thing with CO on my system is, with lower voltage I get higher boost clocks. I just tried to just up every CO value by +20 and did my tests again: 2-3C highter temps but in all tests nearly 150Mhz less clock. I may have already found my CPUs sweetspot

Reducing voltage is giving you higher clocks, because you have less heat to deal with and/or it's keeping you within your PBO voltage limits.

If you cant get higher clocks without the massive undervolts, then it would imply your cooling or PBO settings cant keep up with the added boost

 

[DemonicRyzen666]

Reducing voltage is giving you higher clocks, because you have less heat to deal with and/or it's keeping you within your PBO voltage limits.

If you cant get higher clocks without the massive undervolts, then it would imply your cooling or PBO settings cant keep up with the added boost

Wait what do you mean PBOI can't keep up?
I get the cooling part, but PBO too :-/
What should be my single core score for 4850mhz (5600x)? I can't get higher than 1600 in cinebenchR23.
I'm going to try a negative offset on Stock PBO set up late tonight when I get home from work. See if that changes anything

 

[mama]

I had some stuff really high, but one thing I know this chip has improved speeds just with voltage bump. The voltage bump on the board here is says cpu NB/Soc voltage. The single thread jumped up about 200-600 points. Just from bumping that buy +1 setting.
High like up in the 335 PPT 175 TCD 165 ECD . I doubt it did anything past 125PPT. Also and curve optimizer at like negative 15-17. My all core was sustained 4,700mhz-4,705mhz My single thread was something like 4,850mhz-4875mhz. I did throw in a bump in my bus speed. Because for some reason this boards bus speed is never what it states even on auto it's like 99.8-99.5. If I set 100.065. it comes up 99.9 :/ and that's only the minimum bump for the bus speed.
Went back to what's recommended on here as a start now. The radiator is in the case it's corsair 4000D air flow. How ever I noticed behind the radiator when the fans are at full blast there isn't much air going threw the radiator, with my hand behind it, doesn't feel like a lot was being forced threw. I might just take the fans from push and put them as pull fans instead. only trouble is how to mount the radiator to the front as the fans are what's mounted to the front. Last I could use better thermal paste too, stuff I got isn't' too great.

Push is usually better than pull for fans and radiators. Maybe get an extra set of fans and set up a push/pull setup?

 

[Mussels]

Wait what do you mean PBO can't keep up?
I get the cooling part, but PBO too :-/
What should be my single core score for 4850mhz (5600x)? I can't get higher than 1600 in cinebenchR23.
I'm going to try a negative offset on Stock PBO set up late tonight when I get home from work. See if that changes anything

If you set it to 35W it's not going to boost high

Set it too high, and it'll add extra heat for no performance gain

 

[DemonicRyzen666]

If you set it to 35W it's not going to boost high

Set it too high, and it'll add extra heat for no performance gain

I only see that in Prime95 with AVX.

 

[Mussels]

I only see that in Prime95 with AVX.

Because AVX is more power hungry than any other extension
That's why intel has an AVX offset, underclocking when AVX is used

 

[DemonicRyzen666]

Because AVX is more power hungry than any other extension
That's why intel has an AVX offset, underclocking when AVX is used

I think I've found a bigger problem
widespread core temperature difference

My IOD die must be bowed outward cause it's cool but the chiplet seems to have poor contacnt on the IHS. Almost 20C difference.

 

[phanbuey]

I think I've found a bigger problem
widespread core temperature difference
View attachment 231121

My IOD die must be bowed outward cause it's cool but the chiplet seems to have poor contacnt on the IHS. Almost 20C difference.

Or that thermal diode is off slightly or inaccurate, or the indium solder wasn't uniform or base of the cooler/mounting etc.

Quite a few possibilities tbh. Has remounting done anything?

 

[Mussels]

I think I've found a bigger problem
widespread core temperature difference
View attachment 231121

My IOD die must be bowed outward cause it's cool but the chiplet seems to have poor contacnt on the IHS. Almost 20C difference.

That would have me remounting the HSF and reapplying thermal paste - IF the load levels are the same between the cores - you needed to scroll down lower and see if they were all given the same wattages under core power

 

[DemonicRyzen666]

That would have me remounting the HSF and reapplying thermal paste - IF the load levels are the same between the cores - you needed to scroll down lower and see if they were all given the same wattages under core power

They're like hovering around 0.8 watts to 0.5 watt difference
I did remount it same results. I think the chiplet is concaved or something under the IHS seems core 0 and core 4 are always cooler.

 

[Mussels]

They're like hovering around 0.8 watts to 0.5 watt difference
I did remount it same results. I think the chiplet is concaved or something under the IHS seems core 0 and core 4 are always cooler.

Yes, they are concave - as well as one side being closer to other heat generating parts of the core... but the temp difference should be well under 5C


Turns out my recent temp spikes were due to coolant clogging up my loop... AGAIN


doing another vinegar flush, sigh

 

[Athlonite]

seems core 0 and core 4 are always cooler.

You might find that they're the shit cores that won't clock as high

 

[DemonicRyzen666]

You might find that they're the shit cores that won't clock as high

Nope, this chip is actually a really good clocker it's actually the opposite according to HWinfo lol
My effective all core load clock is 4695-4699mhz. Where those two cores effective clock speed is 4,816mhz
With a little bus speed overclock these two cores have high 4,950mhz. I'm not using an NVMe SSD here, just a Sata SSD.
I probably should get a better power supply, because I don't have the 4pin cpu plug plugged in because this corsair CX 650M PSU only had the 8 pin CPU plug.

 

[Mussels]

Nope, this chip is actually a really good clocker it's actually the opposite according to HWinfo lol
My effective all core load clock is 4695-4699mhz. Where those two cores effective clock speed is 4,816mhz
With a little bus speed overclock these two cores have high 4,950mhz. I'm not using an NVMe SSD here, just a Sata SSD.
I probably should get a better power supply, because I don't have the 4pin cpu plug plugged in because this corsair CX 650M PSU only had the 8 pin CPU plug.

Dude a 4 pin can probably handle your CPU, let alone 8 or 12

 

[DemonicRyzen666]

Dude a 4 pin can probably handle your CPU, let alone 8 or 12

I know when fully loaded the GPU and CPU and it only puts out about 400 watts in total. Coming from a phenom II 1090T that would pull up around 300 watts on over clock at 4.2Ghz on water for the cpu alone; a HD 5870 2GB Eyefinity edition in crossfire (I've had great luck with cpu's lately) to this whole setup being only 100 more than the cpu with a gpu's that's about 12 times faster and a cpu that's around almost 3 times faster in single thread and 4 times faster in multi was huge step up. I'm probably going to Drop in more ram and a that 5800x 3D when out. Then bundle sell the cpu, powersupply, and ram, throw in a 1030 GT for someone on the cheap.

 

[pf100]

I tried lots of different pbo settings as recommended in this thread and settled on manually setting my 5800x's pbo settings to what "eco mode 95w" sets it to which is ppt 128w, tdc 80a, and edc 125a. Then I set curve optimizer to all core -5 because it crashed once at -10 so I just left it at -5. I left clock override at stock. Now I have 10 degree cooler temps with my noctua nh-d15s with an added nf-a12x25 fan on the front. It was running at 82c at sustained full load stock and now it's around 72c. Gaming performance is awesome and is around 70c. With full load multicore I lost a little performance but I don't care because it's not a lot and I don't do full load multicore stuff very often anyway. I like it like this and it works so well I'll probably never touch it again.

 

[Mussels]

I tried lots of different pbo settings as recommended in this thread and settled on manually setting my 5800x's pbo settings to what "eco mode 95w" sets it to which is ppt 128w, tdc 80a, and edc 125a. Then I set curve optimizer to all core -5 because it crashed once at -10 so I just left it at -5. I left clock override at stock. Now I have 10 degree cooler temps with my noctua nh-d15s with an added nf-a12x25 fan on the front. It was running at 82c at sustained full load stock and now it's around 72c. Gaming performance is awesome and is around 70c. With full load multicore I lost a little performance but I don't care because it's not a lot and I don't do full load multicore stuff very often anyway. I like it like this and it works so well I'll probably never touch it again.

Next up is tweaking individual cores with the optimiser and finding which ones can go lower

Some can do -30 while others do -5
I used OCCT with its core swapping setting, corecycler can do it too

Just remember that if you change the setting that lets you add upto 200Mhz boost, it'll possibly become unstable and require tweaking again

 

[pf100]

Next up is tweaking individual cores with the optimiser and finding which ones can go lower

Some can do -30 while others do -5
I used OCCT with its core swapping setting, corecycler can do it too

Just remember that if you change the setting that lets you add upto 200Mhz boost, it'll possibly become unstable and require tweaking again

Thanks for the OCCT tip. I didn't know about that method. I found a reddit post about it here saying to use OCCT's Small Data Set, SSE instructions, 1T and set affinity to the core you're testing. If you have more or better info on methods to set and test curve optimizer with the least amount of agony I would appreciate it. I'd like to avoid having to reboot 100 times and spend a whole day on it. About core boost, I've had set it to +200 before and it boosted to 5050 mhz but I don't think that actually matters in real life over the stock setting does it? I ask because I just left it stock because I figured it would increase stability and reduce crashing. I'm extremely impressed with the performance of the 5800x and is the first cpu I've ever had that I didn't feel a need to overclock it, but I do want the best performance for the least heat.

 

[Mussels]

OCCT has a setting to cycle cores, or lock the affinity - makes it pretty simple.

Set -5, test all of em
If stable, try -10
etc

The moment you start getting errors on any cores, drop back to the stable setting and manually lower the ones that had no issues, and test away... it can be slow, but stressing just the one core i have no issues using web/gaming at the same time

 

[Cutechri]

Kind of pointless to worry about temps on Zen 3, an aggressive architecture designed to redline your CPU temps for the highest boosts possible. If the CPU's voltage regulator is running (which it is if you leave the CPU stock), you don't need to worry about anything unless there's significant cooler noise due to the increased temps, but this has been an issue since Zen 2. In that case just tune the fan curves or get better cooling.

 

[Mussels]

Kind of pointless to worry about temps on Zen 3, an aggressive architecture designed to redline your CPU temps for the highest boosts possible. If the CPU's voltage regulator is running (which it is if you leave the CPU stock), you don't need to worry about anything unless there's significant cooler noise due to the increased temps, but this has been an issue since Zen 2. In that case just tune the fan curves or get better cooling.

the 5800x behaves different to every single other zen 3 chip, thats why theres so much focus on it

Theres a bit of silicon lottery too, with ~15C variance between samples depending on IHS flatness and internal TIM solder quality (i've had someone troubleshooting temp issues swap out CPU's between systems, and some simply run hotter)

 

[Cutechri]

the 5800x behaves different to every single other zen 3 chip, thats why theres so much focus on it

Theres a bit of silicon lottery too, with ~15C variance between samples depending on IHS flatness and internal TIM solder quality (i've had someone troubleshooting temp issues swap out CPU's between systems, and some simply run hotter)

Yeah, it is indeed the hottest Zen 3 part and that's no surprise given the die layout.

 

[kiddagoat]

I decided to go with an all-core OC. The PBO was giving me up to 80C on my 5900x even with some tweaking and curve optimizations.

Now I don't go over 60C ever and I can keep a solid 4.2GHz clock on my CPU. My system startup appears to be snappier and I don't see any slowdowns in gaming or other tasks.

DVDFab seems to like it more as well.