Ryzen 9 3900X OC.

[trickson69]

Okay well I did it with 1.39 vcore but it does jump up to 1.4v when benchmarking no issues at all yet. temps are hitting 82 and 70*C the bigger being the CPU die and package and the lower the CPU.
I am happy!
for now have to do a OCCT run this is where I get nervous.

 

[freeagent]

No idea. I don't OCCT.
It's not real world load. Everything will get hot. Doesn't matter which cooler.

I am curious how PA120 would hold up with your 3900X under it compared to TS140P

I can squeeze 240w PPT out of my 5900X with PBO and surprisingly PA120 can take it, but not as well as FC140 can.

 

[EarthDog]

well if that is the case I can assure you that your system is not stable. If you can at least pass occt benchmark, then you can say you are stable.
Also yeah it might not be real world, but it is stable world and also tells you if your system is going to overheat.
It does take your system to the limit making sure it is stable.

What's your goal bud? To pass OCCT so you think you're stable? You say it spit out temp warnings, but never mentioned what temps.... now suddenly you aren't close to 95C/the limit? You've got 10C before id stop going. Are you still on auto votage??

Also, id imagine there to be a negliglbe difference between that CM AIO and the nhd-15.

I haven't touched OCCT in a decade. For stress testing, I use AIDA64. Disable AVX512, check off cpu/fpu/cache/memory and off I go. If that passes for a few hours, its stable for MY USES (the important part in caps - id also bet my uses are more streasful).

but it is stable world and also tells you if your system is going to overheat.

It's going to tell if you if you will overheat in unrealistic load conditions. Unless you're a user who utilizes all cores and threads of a cpu in your workflow you'll never come close to what mkst stress tests do. Gaming, for example is 10-30C less for me. Any photo or video editing is still 10-20C less for me.

It's still worth it to stress test, don't misunderstand me, so take it closer to the limit so you're not putting a glass ceiling on the clocks.

I am curious how PA120 would hold up with your 3900X under it compared to TS140P

I can squeeze 240w PPT out of my 5900X with PBO and surprisingly PA120 can take it, but not as well as FC140 can.

Don't forget, amd had a 500w gpu out cooled by a 120mm cooler.

 

[Mr.Scott]

Stability is in the eye of the beholder. If you can do everything you do with no error or crash, it's stable.
I'm not trying to launch a missile or anything.

 

[trickson69]

Anytime you receive a overheat warning from the bios then things are bad.
I would imagine if it is hitting the temperature max that would be 95*c or 100*C I am not sure but for it to give that error and not any other tells me it reached its thermal maximum.
Stability is not a restart and error message from the BIOS stating overheat protection.
This CPU cannot take the voltage at all. At 1.35volts it is temp stable at 4.350MHz any voltage above this give the overheat warning and restart.
First time I have ever seen this from a CPU OC.

 

[EarthDog]

Anytime you receive a overheat warning from the bios then things are bad.

Huh? How are you getting an overheating warning from the bios? Tha doesn't make any sense considering there's no load...

I would imagine if it is hitting the temperature max that would be 95*c or 100*C I am not sure

Be sure! How can you move forward not being sure??!!! Lol!

Coretemp reads temps and wattage. Hwinfo64 also does a great job and provides a lot more information.

Stability is not a restart and error message from the BIOS stating overheat protection.

What?!!! See above. I dont understand what you're trying to say here either. I get the restart, but how you're seeing a "bios" message in windows when stress testing is beyond me. Is OCCT telling you this? If so, what are the temps?????

I'll ask again.... are you still on auto or using manual voltage?

 

[trickson69]

Okay first off at 100% load at 1.4volts the system gives a overheat warning from the BIOS this after a restart.
So after running even cinabench R23 in loop the same resulting high Temp error comes up. System restarts BIOS gives a warning of CPU over temp.
Granted I have never seen anything like this before myself. But this is what is happening.

So, to be clear I am not getting any crashes just over temps.
I do know the difference.

Strange as this sounds to me as well. The system will restart then once it boots into the BIOS the system tells me that an over temp error occurred restarting the system.

 

[EarthDog]

That's a bit more clear, but post a screenshot of this bios error, plz. Have usb stick on the system, and hit f12...should save to usb stick (simple thing to figure out).

So, to be clear I am not getting any crashes just over temps.
I do know the difference.

So you can run CB and OCCT. They don't crash now. When you restart the system on your own, you then get a temp error????? Again, screenshot this error please...

Edit: I think I see what you're saying..... a "restart" because of temperatures is a "crash"!

Just have coretemp up at the same time of your stress test and watch.. if it starts to hit 95C and crashes, you can confirm temps. But you have to 'know' these things when overclocking/troubleshooting, dude. Side note, I never liked OCCT...

 

[trickson69]

Yes what I am saying is that the system crashes from an overheating CPU due to CPU overheat protection causing a restart of the computer. This in turn gives me the CPU overheat warning on the BIOS screen.
But I was able to get a screen shot of the current high temp I get while running cinabench.
and the score after was as good as the one I got with 4.4GHz.

 

[freeagent]

Okay first off at 100% load at 1.4volts

That right there is your problem. Because when you see the board give 1.5v when it is on auto mode, it's not the same as what you are doing. With Zen 2 1.325v would be my max all core voltage, maybe 1.3375v if I could keep it cool. That is all load voltage, such as Linpack or something heavy like that.

 

[trickson69]

That right there is your problem. Because when you see the board give 1.5v when it is on auto mode, it's not the same as what you are doing. With Zen 2 1.325v would be my max all core voltage, maybe 1.3375v if I could keep it cool. That is all load voltage, such as Linpack or something heavy like that.

Yeah, this chip is a hot one and at 1.35volts it hits the limit any higher say 1.37 or 1.4 volts it overheats I need some kind of freezer on top of this CPU to keep it cool.
But hey I can't ***** 4.350Mhz is smokin fast!

 

[Mr.Scott]

Okay first off at 100% load at 1.4volts the system gives a overheat warning from the BIOS this after a restart.
So after running even cinabench R23 in loop the same resulting high Temp error comes up. System restarts BIOS gives a warning of CPU over temp.
Granted I have never seen anything like this before myself. But this is what is happening.

So, to be clear I am not getting any crashes just over temps.
I do know the difference.

Strange as this sounds to me as well. The system will restart then once it boots into the BIOS the system tells me that an over temp error occurred restarting the system.

I have never gotten anything like that. Not even close.
Think you just got a bum chip.

 

[Zerileous]

I think we need to see actual temps. I believe if you hit a certain temp you will shut down (though I'm not inclined to try and find out).

Are you saying that it reboots while running the load, and when the system reboots it gives a high temp error? This is why you gotta monitor your temps when you start overclocking.

I definitely had??? issues running over 1.4v on my custom loop with my 3700x, hence all my questions. Not saying my loop is perfect, I've had a few issues with CPU temps (clogged jets, etc) so it is totally possible plausable for off the shelf hardware to beat my loop, given that I am more lazy than I should be about cleaning.

Edit: So I lied. Or maybe I never tried it after clearing the clogged jet. I do enjoy a little OC now and then but I tend to be busy with other things lately. Quick and dirty 4.4GHZ all cores (just 8 in my rig), 1.45v (droop was down to 1.4v tho), 91.5c max, 110w to the package. This was one run R15 on a cool (not from shut down, but also not from load) loop.

Also to clarify, don't expect the 3900x (12 core part) to require similar cooling as my 8c part.

 

[trickson69]

yeah my Ryzen7 3700X does 4.4 on air. it is a way cooler chip.

 

[EarthDog]

I think we need to see actual temps. I believe if you hit a certain temp you will shut down (though I'm not inclined to try and find out).

Are you saying that it reboots while running the load, and when the system reboots it gives a high temp error? This is why you gotta monitor your temps when you start overclocking.

+1

Been asking for exact temps (suggesting apps to read it from) and we still don't know (maybe I missed it, lol) if he's manually changing the voltage or not...if not, as was said previously, you can turn the voltage down and try testing....

 

[trickson69]

I think it is just a "Bad" Chip to OC.
This one is fine at stock speeds and all but once you start to tamper with the voltage and speed it starts to give CPU overheat errors.
I am looking for "Better" cooling but not sure what would really work if this is that "Bad".
Maybe just buy a new one I got this one-off Ebay for 200 bucks.

What is the difference if I post a screen shot (If I could, I would have) of the temps when it crashed? It is not easy to take a screen shot when the system is crashing, and I do not know of any program that does.
It is defiantly hitting 95C for it to shut down restart and tell me that the CPU overheated right?
I really think this CPU is faulty in some way.

Even after a reseating and reapplication of thermal grease I still see extremely high CPU temps at just 4.3Ghz this has to be the CPU.

 

[EarthDog]

Trickson69, are you changing the voltage manually or just adjusting the multiplier? If you are changing it manually, what mode (offset, override, adaptive)?

At stock, what voltage does it run at when it isn't overheating? (what does CPUz say? when you're running OCCT?)

Have you tried setting the voltage to that (making sure your load voltage is exactly the same in windows) and seeing if you're able to reach 4.4 GHz?

Have you tried using LESS than 1.35V for 4.4 GHz?

I am looking for "Better" cooling but not sure what would really work if this is that "Bad".
Maybe just buy a new one I got this one-off Ebay for 200 bucks.

I don't think better cooling is going to do much. At least, not worth the price anyway. You're going to drop $150 on 3x120mm AIO for 75MHz (maybe)? I mean, your call.... but just saying that's what it will amount to.

It is defiantly hitting 95C for it to shut down restart and tell me that the COU overheated right?

Short answer is yes. But, I wanted to cover all bases. I really want YOU to see ANOTHER temperature application reach a peak and see it shutdown.... just to make sure it jives.

I really think this CPU is faulty in some way.

Even after a reseating and reapplication of thermal grease I still see extremely high CPU temps at just 4.3Ghz this has to be the CPU.

No. I highly doubt that. Welcome to modern processors, my man. The newer they are, the less headroom that's left in them, the more power they use, and the more difficult they are to cool (higher transistor density).

 

[trickson69]

Trickson69, are you changing the voltage manually or just adjusting the multiplier? If you are changing it manually, what mode (offset, override, adaptive)?

At stock, what voltage does it run at when it isn't overheating? (what does CPUz say? when you're running OCCT?)

Have you tried setting the voltage to that (making sure your load voltage is exactly the same in windows) and seeing if you're able to reach 4.4 GHz?

Have you tried using LESS than 1.35V for 4.4 GHz?


I don't think better cooling is going to do much. At least, not worth the price anyway. You're going to drop $150 on 3x120mm AIO for 75MHz (maybe)? I mean, your call.... but just saying that's what it will amount to.



Short answer is yes. But, I wanted to cover all bases. I really want YOU to see ANOTHER temperature application reach a peak and see it shutdown.... just to make sure it jives.



No. I highly doubt that. Welcome to modern processors, my man. The newer they are, the less headroom that's left in them, the more power they use, and the more difficult they are to cool (higher transistor density).

Short answer I did it all.
I changed voltages manually; I've let the bios do it I set the offset + and every time 4.4Ghz overheats not so much at 4.3Ghz as long as I keep the Vcore at 1.32v.
Yeah it is a monster of a chip for sure. How are people cooling them? It seems a bit insurmountable IMHO. I wouldn't care to drop the cash as long as I know it would do the job.
Thanks for your Help I was looking at that very cooler and may just get it.

 

[Zerileous]

You need to first set up the monitoring, then incrementally make changes and evaluate the impacts each incremental change has to the temps. I was initially impressed by your max temp of 76c at 1.35v in your original OCCT post. That's pretty decent for off the shelf cooling. Maybe I should have shared each I used last night, and I can when I get home, but I started with a max of like 71c stock and ended up with a max of 91.5c at 1.45v set (but actual load voltage was only 1.4v). Don't expect the system to exceed 1.4v without custom cooling, you're still doing really well. The goal is not to try and capture a screenshot as it shuts down or crashes. The goal is to set a more conservative personal limit that is less than the absolute limit of the hardware, I believe 95c was suggested, and then gradually increase voltage and clocks to find the highest clock at lowest stable voltage without exceeding that temperature.

Edit: It seems like 1.35v and 4.3GHz is a good starting place. Then you will increase your clock by 25MHz and run a load (what load you use is a personal decision, as Mr. Scott touched on earlier, the goal is to make sure that it can serve your needs without issues, not fly it to Mars). I think for roughing it in one or two cinebench runs is good, then you can test for longer once you get dialed in. This is where you monitor your temperature, not after you have problems. Say you do a couple increments of increasing and now the system crashes at 4375MHz, and you know it didn't overheat because you were monitoring your temperature. Now you can increase your voltage by 25mV and try 1.375v and see if it will run the cinebench at that level. Keep doing that until you reach the personal / soft thermal limit (i.e. 95c). Then you can try something harder or run cinebench on a loop if that gives you more peace of mind.

In your screenshots above you have temperatures but it doesn't show the clock and voltage, and we still don't understand exactly how you are setting those figures either.

EDIT: when we use the term "manual" (well maybe I should just speak for myself), but I interpret that to mean setting a fixed v-core, i.e. 1.35v or 1.45v. I interpret the term "offset" to mean you're setting a "core voltage offset" to +0.05v or -0.05v or +0.10v so that whatever the CPU asks for, the motherboard gives it more or less. I suppose the ideal terminology would be "fixed voltage" vs "offset voltage" since both are entered manually, vs. leaving the settings on automatic. I am not familiar with a good usage of + offset voltage. Some chips will run better with a small negative offset, as it allows the boosting algorithm go higher without hitting thermal limitations, but not every chip will be stable. My advice is to either use "fixed" voltage with "fixed" *****, or offset voltage with automatic or boosting clocks. I think trying to use a fixed clock with an offset voltage creates too many variables.

 

[EarthDog]

not so much at 4.3Ghz as long as I keep the Vcore at 1.32v.

So 4.4 GHz crashes at this voltage?

Thanks for your Help I was looking at that very cooler and may just get it.

I didn't mention a specific cooler........ just a 3x120mm AIO of which there are dozens out there.

How are people cooling them? It seems a bit insurmountable IMHO. I wouldn't care to drop the cash as long as I know it would do the job.

.... with custom loops and 3x120mm AIOs.... or they are sitting where you are. But again, it's 100MHz at most you MAY get out of it. Look at some reviews of the best 3x120mm AIOs before you go off willy nilly buying things.


I really wish you would answer the questions asked of you...this is getting tiresome.

At stock, what voltage does it run at when it isn't overheating? (what does CPUz say? when you're running OCCT?)

Have you tried setting the voltage to that (making sure your load voltage is exactly the same in windows) and seeing if you're able to reach 4.4 GHz?

Have you tried using LESS than 1.35V for 4.4 GHz?