3600x overclocking weirdness

[Rage Mage]

Hi all,

I am experiencing some strangeness on my 3600x, and I am trying to figure out what is at fault.

I run it on a Gigabyte X370 AORUS Gaming K7 with the latest BIOS, with a Deepcool Captain 240 AIO watercooler and 16GB of Samsung B-die RAM. My PSU is a modular Corsair 650W one, don't remember the model but it was either gold or platinum 80+ rated.

Initially I had only enabled PBO (advanced, 10x, +200Mhz, motherboard limits) on the BIOS but due to my need for increased performance on Cyberpunk (futile, I know) I started monitoring the speeds and temps more closely.

What I found out was that while playing the game I saw max core speeds of around 4.1Ghz, which lead to trying cinebench - that was running up to a max of 4.0Ghz. All this was with temperatures around 70C. Even on single core Cinebench tests the highest I saw was between 4.2 and 4.3Ghz.

This seemed strange to me, so I installed Ryzen Master that said max speed was 4600 on the CCX, applied AutoOC and left PBO on at the BIOS. All this changed was that Cinebench started at 4.05-4.1Ghz for a couple of seconds, and then dropped back to 4.0.

At this point I started asking myself if it is the motherboard due to it being X370 and not very good at the PBO stuff, the AIO, the CPU, the PSU, etc. I did a lot of things after this including multiple CMOS resets, RAM overclock tests (tried with default 2133 speed too), Fabric speed tests, played with voltages and LLC and found another "bug" where if I set dynamic voltage on the BIOS, not matter what I add or subtract it always shows 1.1V in windows. The AIO is on top of the case, now configured as intake (used to be exhaust) with open case and no intake filter/restrictions.

So I opened Ryzen Master and started manually overclocking - I reached 4.250 on all cores with a voltage of 1.18 to 1.20 but got very strange temp readings while running Cinebench - immediate jumps to 85 - 92C specifically, after which I stopped pushing. I reseated the AIO and reapplied thermal paste just in case with no change. When remounting the AIO plate the paste pattern looked normal before I cleaned it off, but I had mounting issues with it in the past on my R7 1700.

Does anyone here have a clue as to what might be going on here and possibly any suggestions so I can ideally avoid ending up in a mental hospital?

 

Thanks

[GorujoCY]

It might not be reporting well the temperature and maybe it does.

 

If you have a thermal camera, can you try measuring it there and see what your thermal camera reports, I think even a laser temperature meter will work well. That would be my first step but if you have an alternative way of checking temperatures physically, you do you...

 

Edit: I think I'm wrong, but it's a good measure to check, but it must be an architecture limitation or something along those lines

 

 

 

[Rage Mage]

12 minutes ago, GorujoCY said:

It might not be reporting well the temperature and maybe it does.

 

If you have a thermal camera, can you try measuring it there and see what your thermal camera reports, I think even a laser temperature meter will work well. That would be my first step but if you have an alternative way of checking temperatures physically, you do you...

 

 

 

 

 

Hi neighbour (Greek here) and thanks for the reply,

 

unfortunately I do not have a thermal camera of any kind, if I did have one would it be a good idea to try it on the radiator fins? They should be getting hot even to the touch when the CPU is reporting 90+ degrees right?

[GorujoCY]

13 minutes ago, Rage Mage said:

Hi neighbour (Greek here) and thanks for the reply,

 

unfortunately I do not have a thermal camera of any kind, if I did have one would it be a good idea to try it on the radiator fins? They should be getting hot even to the touch when the CPU is reporting 90+ degrees right?

hmm yeah maybe.

 

From what i can tell, it can be a temperature behavior where it throttles below the clock to keep it's temperature

 

but it can also be a cpu/motherboard behavior which puts me into the mystery hole

[Tristerin]

When you increase voltage you increase heat, and Cinebench will eat up a 240mm AIO.

 

My R7 1700 was on a 240mm AIO, overclocking headroom is not much with that.  Custom loop puts me at 4.2ghz all core no problem (which paces with 3600 series CPUs)

[Rage Mage]

25 minutes ago, GorujoCY said:

hmm yeah maybe.

 

From what i can tell, it can be a temperature behavior where it throttles below the clock to keep it's temperature

 

but it can also be a cpu/motherboard behavior which puts me into the mystery hole

As I don't have much experience with other Matisse CPU's, do you reckon that my temps on 1.2V are normal? Do they usually go that high during full synthetic loads?

 

Hopefully someone with a 3000 series will show up and share their temps to verify if this is normal or my AIO is dying

[Rage Mage]

6 minutes ago, Tristerin said:

When you increase voltage you increase heat, and Cinebench will eat up a 240mm AIO.

 

My R7 1700 was on a 240mm AIO, overclocking headroom is not much with that.  Custom loop puts me at 4.2ghz all core no problem (which paces with 3600 series CPUs)

Yeah same, I bought this AIO with my 1700 and I could get it up to 3.85 with temps around the 85 range but I remember that being normal for that generation and core count, especially due to the voltage being around 1.4 at that point.

[Rage Mage]

So after wasting a few more hours on this turns out I had to re-flash the BIOS, now it's boosting 4.025-4100 on all cores on synthetic benchmarks at 70C max and Cyberpunk is boosting to 4.275-4.3 at 66C max.

 

Even the all core OC now makes sense, as I have to go to 1.375 to get a 4.2 Ghz.

 

Thanks for the help.

[GorujoCY]

6 minutes ago, Rage Mage said:

So after wasting a few more hours on this turns out I had to re-flash the BIOS, now it's boosting 4.025-4100 on all cores on synthetic benchmarks at 70C max and Cyberpunk is boosting to 4.275-4.3 at 66C max.

 

Even the all core OC now makes sense, as I have to go to 1.375 to get a 4.2 Ghz.

 

Thanks for the help.

well I'm glad you found a solution...!

[rkdee]

Glad you fixed it.

 

Honestly, GorujoCY is right with the thermal camera thing. I always have one at hand, I used to use it on mining rigs to do like a general inspection. I used the Hti 301 which is a little camera that you plug into your smartphone. It's a bit on the expensive side, but the quality is just perfect. Made sense for me because the rigs were worth much more than that. But you can get a standard Seek Compact if you don't want to spend much.  

 

The temp readings from the software are not to be relied upon, it's always good to have a 3rd party that's not part of the system.