Hot Temps for 5800X3D

[dgianstefani]

What sticks are those? I've got these https://www.overclockers.co.uk/team...00c16-3600mhz-dual-channel-kit-my-002-8p.html

Absolute crap. Can't even get 3600C14. Only XMP actually works. 3600C14 worked for a few weeks, passed memtests for 2 days, corecyclers etc, then it started borking out... Had to run 1.47v (actual 1.5v on my Gigabyte b550i as it always overvolts) for this which is pretty bad, spent over a month trying to tune these only to end up at the beginning :l

1.5 v isn't high if that's B die, stock kits go up to 1.55 V, previously I had a 2x8 TridentZ 3800/14, you can run up to 1.7 daily as long as they're kept cool. I do 1.6 since I'm not pushing anything crazy like 4400/14, other voltages are in screenshot and are also important.

Yeah those 8 pack are B die, so up the voltage, put airflow over and you should be able to do 3800/14 no problem. tRFC is also important as mentioned, but you should be able to copy my subtimings but use appropriate tRFC values off the table I linked, and 3800 instead of 3933.

You'll start to error out above 40 °C.

MemTestHelper/DDR4 OC Guide.md at oc-guide · integralfx/MemTestHelper

C# WPF to automate HCI MemTest. Contribute to integralfx/MemTestHelper development by creating an account on GitHub.

github.com

This can be somewhat helpful as a basis to start with if you don't want to just copy timings.

Remember you aren't forced to run CL 14, try higher CL and MHz, latency still goes down as you increase MHz.

RAM Latency Calculator

 

[Sptz]

View attachment 300734

1.5 v isn't high if that's B die, stock kits go up to 1.55 V, previously I had a 2x8 TridentZ 3800/14, you can run up to 1.7 daily as long as they're kept cool. I do 1.6 since I'm not pushing anything crazy like 4400/14, other voltages are in screenshot and are also important.

Yeah those 8 pack are B die, so up the voltage, put airflow over and you should be able to do 3800/14 no problem. tRFC is also important as mentioned, but you should be able to copy my subtimings but use appropriate tRFC values off the table I linked, and 3800 instead of 3933.

You'll start to error out above 40 °C.

MemTestHelper/DDR4 OC Guide.md at oc-guide · integralfx/MemTestHelper

C# WPF to automate HCI MemTest. Contribute to integralfx/MemTestHelper development by creating an account on GitHub.

github.com

This can be somewhat helpful as a basis to start with if you don't want to just copy timings.

I've followed that guide to a T. Thing is these kits don't have a sensor, so no idea what their temp is at, airflow is good on my build, mainly from the cpu cooler dripping over the sticks but then again, I don't know, 1.5v seemed too excessive for 3600c14. I couldn't even boot 3800 on any timings, loose, mega loose, auto, just didn't boot, had to clear cmos everytime, it did boot 3866 but even at loose timings it was a WHEA fest. I honestly put it down to either the sticks or the CPU being a bad bin.

 

[dgianstefani]

You have to increase related voltages not just VDIMM.

Nothing is excessive unless you're going above 1.7 V VDIMM.

 

[Mussels]

Newer 5800X3Ds aren't as good as launch ones for temps. Regardless, 75 C instead of 70 C I don't see an issue.

This has been noted by YouTubers who've bought more than one, e.g. OptimumTech.

I assume this is because the best dies are going to server chips.

I also have a fairly early production 5800X3D, I bought mine two months after release. Recent production X3D chips are binned, the most efficient bins seem to be reserved for Epyc chips.

It's not something to worry about though ~ 60 C or 80 C is practically meaningless for a desktop CPU, especially one that can't be overclocked, you may use 1-2 W extra from voltage/leakage scaling due to the temperature, and maybe 25 MHz from boost scaling, but it's irrelevant, these aren't KS series chips you're trying to eke out 6.4 GHz instead of 6.3, where every °C matters.

Typical gaming load looks something like this, currently temperatures where I live are 20-25 C. The GPU is not bottlenecked, it's running a 236 FPS frame cap in NVControlpanel, and the CPU frequency is averaged, since gaming doesn't use all cores. My 240mm radiator manages to cool 400 W with no issues.

View attachment 300671

View attachment 300661View attachment 300662

@Mussels if that RAM is B-Die you can get your tRFC down by quite a lot, besides absolute latency, tRFC latency is one of the biggest Zen performance variables for gaming, lower = hotter though, so ensure you have good airflow over the DIMMs. Mine have been repadded with 15 W/mK pads I had leftover from the GPU repad. You could probably get sub 300ns without even changing voltages. It's one of the easiest methods to tune for performance with RAM, because you can simply use this chart, and adjust your values by one step then test, until you become unstable.
View attachment 300665

For every 10trfc drop you get .1ns lower latency on AMD.

500trfc vs BDie 250trfc is 2.5ns lower latency just from tRFC value.

If you don't have BDie you'll struggle to get lower than 400 ns though.

I've got 190 ns which is decent, but there's guys out there running 150 ns lol.

I'm very curious to see how my temps change when I apply Conductonaut Extreme on both CPU/GPU, and upgrade GPU pads to 20 W/mK ones. Currently only have Conductonaut (original) on the CPU.

It's not B-die, it's listed in my system specs

Why don't I see the CPU temperatures when gaming on that page?

I only see the "Load" temperatures on that page...

Because it was added in newer reviews, after lots of comments/complaints/requests about how gaming and synthetic tests don't line up at all any more with high core counts

 

[Sptz]

You have to increase related voltages not just VDIMM.

Nothing is excessive unless you're going above 1.7 V VDIMM.

Yes, ofc. Vsoc + vddgs. Those on auto are already quite high on my Gigabyte, default vsoc is 1.1v which is crazy. Nothing made a difference

 

[freeagent]

Why is 1.1v SOC crazy?

 

[Mussels]

Why is 1.1v SOC crazy?

as a default it's high - to my knowledge AM4 doesnt change SoC voltages based on XMP, so it should be 0.95v
He's got the same mobo i have in my ITX system, i'll check what its defaulting to here

I was running 1.125v for 3200MT/s on my 3700x

"Auto" with XMP on, is showing 1.0875v - so yes, 1.1v and minor droop makes sense there.

Spoiler: ranting at gigabytes shitty BIOS

Edit: i really hate gigabytes motherboards right now, they've got some sort of issue with posting on 16:9 displays regardless of GPU - both my 1440p 144Hz and 4K60Hz have issues with not showing the BIOS, and i need to connect them to a 1080p display on HDMI to get an image
It tries forcing 1680x1050 as far as i can tell, and it doesnt work
It's triggered by disabling CSM, also known as enabling ReBar

Edit 2: I was close, it's forcing 1920x1200 or 1680x1050 depending on the display and the BIOS lags *HORRIBLY* if you can get into it like this
Ctrl-F6 lets you adjust the settings, and 1366x768 is actually usable

[/spopiler]

 

[freeagent]

as a default it's high - to my knowledge AM4 doesnt change SoC voltages based on XMP, so it should be 0.95v

On both of my boards default is .975v, I think at 1600 it changes to 1v, and for 1800 it changes to 1.1, where it wont go any higher when using AUTO.

 

[Mussels]

edited previous post, adding screenshots here
My other boards do not change SoC voltage - they have a default and stick to it. This is the first time i've seen SoC voltage above 0.95v, but i have seen lower on x370 boards
All voltages on defaults (and auto/normal mean the same thing, apparently)

"Auto" with XMP on

"Normal" is the same at 1.0875v
Normal, auto, RAM speed, XMP, FCLK - tested them all, none change the SoC voltage. It's always that one value, let's say 1.09v for accuracy sake


So yeah - this board apparently runs a perfectly safe and reasonable SoC voltage, but also higher than any other AM4 board i've ever seen default to
It's 100% going to boost RAM compatibility, without causing any safety or heat issues - I guess AM4 not being able to adjust SoC based on RAM/IF/XMP status means they just picked a default and ran with it

 

[freeagent]

Interesting.. I will play with mine tomorrow, I haven’t actually used XMP in any meaningful way in recent memory..

Nitey nite

 

[Mussels]

Interesting.. I will play with mine tomorrow, I haven’t actually used XMP in any meaningful way in recent memory..

Nitey nite

On AM4, the boards can only load some settings from XMP - being an intel standard, they load the most basic settings (Primary timings, voltage) and the rest are often the stock board defaults, or like we saw on AM5 the boards can implement custom 'overclocking' settings for secondary/tertiary timings such as their presets for SoC voltages

I can set TRFC in my custom XMP timings but none of my AM4 boards will load it - they have their own defaults generated based on the MT/s value and ignore XMP entirely
(TRFC goes up with each increase in DRAM speed, and that could go up too loose or too tight for various ram kits - many other timings would behave the same)

Personally I like XMP as a basic time saver but it's always worth knowing the same kit of ram will have different settings in different motherboards especially between brands - and it's why some people have such wildly different experiences with the same kits of RAM. (SoC 0.950 to 1.09v is going to be a HUGE changer for dual rank memory compatibility)

(At least this reminded me to get 3600MT/s working on the ITX system, what I thought was boot failures originally was just the BIOS refusing the give me an image over HDMI)

 

[dgianstefani]

Yes, ofc. Vsoc + vddgs. Those on auto are already quite high on my Gigabyte, default vsoc is 1.1v which is crazy. Nothing made a difference

Lower voltages can sometimes mean better stability with VSOC if it's set too high, VDDG and VDDP have a relationship, VDDP always should be slightly lower.

Post your Zen timings.

 

[Sptz]

as a default it's high - to my knowledge AM4 doesnt change SoC voltages based on XMP, so it should be 0.95v
He's got the same mobo i have in my ITX system, i'll check what its defaulting to here

I was running 1.125v for 3200MT/s on my 3700x

"Auto" with XMP on, is showing 1.0875v - so yes, 1.1v and minor droop makes sense there.

Spoiler: ranting at gigabytes shitty BIOS

Edit: i really hate gigabytes motherboards right now, they've got some sort of issue with posting on 16:9 displays regardless of GPU - both my 1440p 144Hz and 4K60Hz have issues with not showing the BIOS, and i need to connect them to a 1080p display on HDMI to get an image
It tries forcing 1680x1050 as far as i can tell, and it doesnt work
It's triggered by disabling CSM, also known as enabling ReBar

Edit 2: I was close, it's forcing 1920x1200 or 1680x1050 depending on the display and the BIOS lags *HORRIBLY* if you can get into it like this
Ctrl-F6 lets you adjust the settings, and 1366x768 is actually usable
View attachment 300768
[/spopiler]

Agree with everything. Even on the latest version it lags horrendously, mainly trying to smash down arrow on a list of say voltages, if you smash 20x and release it will keep going. Absolute crap. Plus the way it overvolts EVERYTHING on auto

Lower voltages can sometimes mean better stability with VSOC if it's set too high, VDDG and VDDP have a relationship, VDDP always should be slightly lower.

Post your Zen timings.

So this is what I was running 24/7 with the teamgroup b-die sticks. I can't find a more recent pic but TRCDWR was at 8 and VDIMM at 1.47v (which hwinfo actually reported 1.5v as well as the bios, since this mobo overvolts), -30 CO. This was fully stable, aida64 benched at 59ns. Passed TM5 1usmus and another recommended profile for 24 hours, passed corecyler running dual as well as single, then all of a sudden it stopped maintaining CB23 for 30 mins run, it would crash at like 15 mins or so, not a hard crash but a software crash (popup saying it was logged but the log file never could be found), backed off CO to -25, same thing, then got a CRITICAL_PROCESS_DIED BSOD that I think actually killed my Sabrent Rocket 4 Plus, it never showed up in BIOS after that, after power cycles, clear CMOS etc. on both M2 slots.

And this is simply running XMP:

Changed nothing apart from VSOC to 1.070v which vdroops to 1.0563v. Everything else is fully default. VDIMM is actually reported as 1.4v.

 

[dgianstefani]

Agree with everything. Even on the latest version it lags horrendously, mainly trying to smash down arrow on a list of say voltages, if you smash 20x and release it will keep going. Absolute crap. Plus the way it overvolts EVERYTHING on auto



So this is what I was running 24/7 with the teamgroup b-die sticks. I can't find a more recent pic but TRCDWR was at 8 and VDIMM at 1.47v (which hwinfo actually reported 1.5v as well as the bios, since this mobo overvolts), -30 CO. This was fully stable, aida64 benched at 59ns. Passed TM5 1usmus and another recommended profile for 24 hours, passed corecyler running dual as well as single, then all of a sudden it stopped maintaining CB23 for 30 mins run, it would crash at like 15 mins or so, not a hard crash but a software crash (popup saying it was logged but the log file never could be found), backed off CO to -25, same thing, then got a CRITICAL_PROCESS_DIED BSOD that I think actually killed my Sabrent Rocket 4 Plus, it never showed up in BIOS after that, after power cycles, clear CMOS etc. on both M2 slots.

View attachment 300788

And this is simply running XMP:

View attachment 300789

Changed nothing apart from VSOC to 1.070v which vdroops to 1.0563v. Everything else is fully default. VDIMM is actually reported as 1.4v.

At first glance -

Your tRFC is too ambitious for low bin B die, especially since you don't have a temperature sensor to flag when errors are due to heat, tRFC lowering increases heat, try 342 or 360 for 190/200 ns at 3600 MT, but you really should be trying for at least 3800 MTs and thus 361/381, you'll see more improvement - 3800 MT is the sweet spot for Zen 3, higher is better, but not by much.
tRCDWR doesn't gain much by lowering, but can lead to instability fast.
VDDP and VDDG should be closer together, 0.05/0.03 V difference.
VSOC can go up by another 0.05-0.1 V safely, assuming there is benefit, put it up as high as it needs to be within that range to get stability.
You can disable GDM, it changes subtimings.
Many of your subtimings are too ambitious, most of those barely make a difference performance wise, but going too low quickly introduces instability.

Again, I suspect your maxed out -30 all core CO is causing instability on at least one core.

1.4 VDIMM is low for B die if you're interested in running them at their potential.

 

[Sptz]

At first glance -

Your tRFC is too ambitious for low bin B die, especially since you don't have a temperature sensor to flag when errors are due to heat, tRFC lowering increases heat, try 342 or 360 for 190/200 ns at 3600 MT, but you really should be trying for at least 3800 MTs and thus 361/381, you'll see more improvement - 3800 MT is the sweet spot for Zen 3, higher is better, but not by much.
tRCDWR doesn't gain much by lowering, but can lead to instability fast.
VDDP and VDDG should be closer together, 0.05/0.03 V difference.
VSOC can go up by another 0.05-0.1 V safely, assuming there is benefit, put it up as high as it needs to be within that range to get stability.
You can disable GDM, it changes subtimings.
Many of your subtimings are too ambitious, most of those barely make a difference performance wise, but going too low quickly introduces instability.

Again, I suspect your maxed out -30 all core CO is causing instability on at least one core.

1.4 VDIMM is low for B die if you're interested in running them at their potential.

Btw, thanks for the detailed post. Very much appreciated!

I hear you on tRFC, that makes sense!
I can't boot at 3800, no matter what. Tried everything I could think of, got sick of clearing CMOS every single time.
I also tried disabling GDM with these timings (XMP), also no dice, didn't boot. Even with higher voltage.
Regarding CO, how can one actually test it properly, OCCT, corecycler gave me 0 errors from -10 to - 30 lol

 

[dgianstefani]

You're likely erroring out when heat build up occurs during gaming. Memory testing software only taxes the memory, sometimes not even the CPU. In actual gaming your GPU and CPU will both be pumping out heat into the case, even with good airflow that will increase the memory temperature.

Obviously there's no way for me to be certain, but I suspect you can't run 3800 MT because you haven't increased VDIMM voltage. The general instability is due to the reasons I already outlined.

B die scales well with voltage, and negatively scales with temperature, keep those two in mind and RAM OC is less tedious.

Hilariously, removing the stock heatspreaders often improves DIMM temperatures.

In my case I repadded them for an improvement instead, but still.

 

[Mussels]

CO is hard to test for, and something you should disable til you've got the RAM nailed down
Keep in mind that you get instability from both the CPU or the RAM getting hotter, let alone both - other heat sources (GPU, NVME's etc) all pile that up quite differently when gaming vs ram testing.

Since the SoC has some of your USB ports and most of your PCI-E lanes, it's voltage is going to droop very differently under different types of load - IF/SoC/WHEA errors usually show up in the transition states as devices suddenly go from low to high power usage

DG has a lot more experience with B-die than I do, so definitely follow his timing advice - I've always ended up with hynix due to B-die being worth more than gold here.

Personally, i've found the following rules are the basics:

• XMP (and EXPO on DDR5) do not contain SoC voltages, or anything other than DRAM voltage - that's always upto the board, and they all choose different defaults.
• You aren't getting above 3900 MT/s stable, Zen3 just doesn't wanna do it.
• More ranks means more SoC voltage is needed, always. Doubling ranks has needed about +0.025v here
• Higher power draw from DRAM/PCI-E may need slightly higher vSoC to compensate the droop. (gen4 NVME's or bright RGB LED's on the RAM, if the boards got weaker power delivery)
• IF stability and DRAM stability are totally independent, you can test this by overclocking one and not the other and ignoring the performance hit temporarily
• Higher CPU temps makes the memory controller unstable, especially past 75c
• Higher DRAM temps makes B-die lose stability faster than other RAM, especially with lower TRFC
• SoC voltage directly controls how high your IF can go
• DRAM voltage helps B-die tighten times or raise MT/s - not always both.
• DRAM voltage helps Hynix raise MT/s, but NOT tighten timings (3200 to 3866 going from 1.35v to 1.40v, but C16 isn't fully stable at 3200 no matter what voltage is used)
• Every board deals with SoC voltages differently, dont forget you can alter the LLC settings as well
  

 

[Sptz]

Really grateful for both of your responses. When you guys OC ram etc, do you have a specific second win installation to test everything as to not corrupt the main installation? I've reinstalled windows more times than I can count last month.

And if you do, there's no chance when crashing, bsod'ing etc that the other windows installation is affected right?

You're likely erroring out when heat build up occurs during gaming. Memory testing software only taxes the memory, sometimes not even the CPU. In actual gaming your GPU and CPU will both be pumping out heat into the case, even with good airflow that will increase the memory temperature.

Obviously there's no way for me to be certain, but I suspect you can't run 3800 MT because you haven't increased VDIMM voltage. The general instability is due to the reasons I already outlined.

B die scales well with voltage, and negatively scales with temperature, keep those two in mind and RAM OC is less tedious.

Hilariously, removing the stock heatspreaders often improves DIMM temperatures.

In my case I repadded them for an improvement instead, but still.

I did try it with 1.5v if I recall correctly, no dice. all on auto. I think I also tried with loading XMP, then changing to 3800 (plus FLCK ofc) no dice. I think I also tried with loose primaries like cl19 or something, also no dice! I don't have active cooling on the ram, that area is a bit cramped in but it does get quite a lot of air from the cpu cooler.

this is the build:

CPU: AMD Ryzen 7 5800X3D
CPU Cooler: Thermalright Peerless Assassin 120
Motherboard: Gigabyte B550I AORUS PRO AX Mini ITX
Memory: TEAMGROUP Dark Pro 8PACK Edition 32 GB DDR4-3600 CL16
Storage: Sabrent Rocket 4 Plus 2 TB
GPU: Gainward Phantom GeForce RTX 4090
Case: Asus Prime AP201
PSU: MSI A1000G PCIE5 1000W

 

[GenericNinja]

Motherboard: Gigabyte B550I AORUS PRO AX Mini ITX

I have the same board and it is unstable at 1800 FCLK ever since I updated above BIOs F13H to get 5800X3D compatibility, have tried all 4 5800X3D supporting versions. On the current bios versions my 5950X / B-die kit which ran at 1900 FCLK completely stable won't even post at 1833 FCLK. At 1767 FCLK system is stable with both CPUs and two different memory kits.

 

[freeagent]

AGESA 1206 and up are for X3D imo. Anything after 1203 hurts older CPUs imo.

 

[GenericNinja]

AGESA 1206 and up are for X3D imo. Anything after 1203 hurts older CPUs imo.

Unfortunately Gigabyte managed to put two different major security vulnerabilities in their older BIOs, plus Sptz and I are on 5800X3Ds, I just put my old 5950X back in for testing after my 5800X3D was unstable.

 

[19Romulus84]

Yesterday I noticed that while playing Battlefield V, of course multiplayer, the temperature stayed at 90-91 degrees throughout the game. It started to worry me a lot, the clocks are normal even at 90-91 degrees all the time 4300MHz on all cores, but I'm worried about such high temperatures. I don't know how to lower them, even because GIGABYTE BIOSes don't have Curve Optimizer or PBO2. They just don't have it because I have the latest BIOS, f64a and these functions are not there, I was looking for an hour, I searched all the tabs, there is no. I installed the latest BIOS while sitting on the R5 3600, is there a chance that if it would overwrite the same F64a BIOS again, these functions would appear? Although I doubt that it will give anything because all BIOS manufacturers always write next to what is in a given version, e.g. AGESA has been improved etc etc. There is nothing like that here. The latest BIOS from gigabyte for the AORUS b450 pro board does not support either PBO or Curve Optimizer.

B450 AORUS PRO (rev. 1.0) Support | Motherboard - GIGABYTE Global

Lasting Quality from GIGABYTE.GIGABYTE Ultra Durable™ motherboards bring together a unique blend of features and technologies that offer users the absolute ...

www.gigabyte.com

 

[freeagent]

I use the latest for my X3D and 1203C for 5900X and 5600X

 

[19Romulus84]

I'm not talking about overclocking the memory, you have BIOS f64, you have BIOS from GIGABYTE and you record from AORUS? Do you have Curve Optimizer and PBO in this BIOS because I don't?

 

[Mussels]

I have the same board and it is unstable at 1800 FCLK ever since I updated above BIOs F13H to get 5800X3D compatibility, have tried all 4 5800X3D supporting versions. On the current bios versions my 5950X / B-die kit which ran at 1900 FCLK completely stable won't even post at 1833 FCLK. At 1767 FCLK system is stable with both CPUs and two different memory kits.

Post your zentimings - i've got mine at 3866 12+ hours stable so far, on the current FBE bios (rev 1.2)
Make sure to load optimised defaults and dont use curve optimiser while testing this stuff - you can easily have an undervolt instability and mistake it foran IF/RAM one.
Oh and make sure your CPU isnt overheating if you're over 80c in gaming, fix the cooling, enable ECO mode, whatever you need to do.

Edit: 3800 is easy to get stable, 3866 is refusing to play ball on the latest BIOS.
I'll have to compare some before/after zentimings results and see what voltages or timings changed to cause that.
Edit 2: 3800 is running stable at 1.10v SoC, drooping to 1.08v (stock, more or less)
1.05v is unstable, with WHEA errors within 5 minutes of running memtest5

Last time I tested needed 1.13v for 3800 with four ranks with the same CPU

Edit 3: Aorus B550I is 100% overvolting the RAM, 1.34v comes out as 1.42v
be careful with setting high voltages as it may go higher than you expect

Really grateful for both of your responses. When you guys OC ram etc, do you have a specific second win installation to test everything as to not corrupt the main installation? I've reinstalled windows more times than I can count last month.

And if you do, there's no chance when crashing, bsod'ing etc that the other windows installation is affected right?



I did try it with 1.5v if I recall correctly, no dice. all on auto. I think I also tried with loading XMP, then changing to 3800 (plus FLCK ofc) no dice. I think I also tried with loose primaries like cl19 or something, also no dice! I don't have active cooling on the ram, that area is a bit cramped in but it does get quite a lot of air from the cpu cooler.

this is the build:

CPU: AMD Ryzen 7 5800X3D
CPU Cooler: Thermalright Peerless Assassin 120
Motherboard: Gigabyte B550I AORUS PRO AX Mini ITX
Memory: TEAMGROUP Dark Pro 8PACK Edition 32 GB DDR4-3600 CL16
Storage: Sabrent Rocket 4 Plus 2 TB
GPU: Gainward Phantom GeForce RTX 4090
Case: Asus Prime AP201
PSU: MSI A1000G PCIE5 1000W

Don't forget the UCLK div on these, it's that stupid setting that reads like UCLK==FCLK (It only matters above 3600, or auto halves it)
What did you run at 1.5V? The DRAM?
That's extremely high if the ram isnt directly air cooled, and not safe for anything that isnt B-die
I use my main OS and never have corruption issues, but i'm also not jumping into extremes like that - I do small changes over time

You can use rufus and make a portable USB for windows, it intentionally locks you out of local storage access for safety/security reasons so it cant corrupt anything but itself
Specific link on "Windows to go" section of the FAQ

AGESA 1206 and up are for X3D imo. Anything after 1203 hurts older CPUs imo.

That's fixed in 1.2.0.A, the EDC bug is gone

Yesterday I noticed that while playing Battlefield V, of course multiplayer, the temperature stayed at 90-91 degrees throughout the game. It started to worry me a lot, the clocks are normal even at 90-91 degrees all the time 4300MHz on all cores, but I'm worried about such high temperatures. I don't know how to lower them, even because GIGABYTE BIOSes don't have Curve Optimizer or PBO2. They just don't have it because I have the latest BIOS, f64a and these functions are not there, I was looking for an hour, I searched all the tabs, there is no. I installed the latest BIOS while sitting on the R5 3600, is there a chance that if it would overwrite the same F64a BIOS again, these functions would appear? Although I doubt that it will give anything because all BIOS manufacturers always write next to what is in a given version, e.g. AGESA has been improved etc etc. There is nothing like that here. The latest BIOS from gigabyte for the AORUS b450 pro board does not support either PBO or Curve Optimizer.

B450 AORUS PRO (rev. 1.0) Support | Motherboard - GIGABYTE Global

Lasting Quality from GIGABYTE.GIGABYTE Ultra Durable™ motherboards bring together a unique blend of features and technologies that offer users the absolute ...

www.gigabyte.com

That says your coolings no good, simple as that. Repaste and re-attach the cooler, gaming should never run that hot

The settings you're looking for were added in 1.2.0.7 for the 3D CPU, they're going to be in the AMD overclocking menu rather than gigabyte branded settings area if they were added by AGESA and not the boards manufacturer.