Normal overclocking, abnormal temps

karanbraich

Honorable
Sep 10, 2013
24
0
10,510
Hey guys,

I recently build a high end rig and tried overclocking in accordance to the guide posted on Asus ROG forum - http://rog.asus.com/216552013/crosshair-motherboards/guide-overclocking-fx-8350-to-4-8ghz-on-crosshair-v-formula-z/2/

However, my CPU temps on the Asus Ai Suite show temperatures hitting 80C!!
I stopped the Prime95 test. Is this normal guys?

Mobo: Asus Crosshair V Formula Z
CPU: AMD FX-8350
Cooling: Noctua NH-D14
Test: Prime95
Temp sensor: Asus Ai Suite

I have a very high end air cooler however it is not able to run at 100% load for 10mins. Please let me know what I can do here.
 
Solution
ok. i'll buy your cpu hit 80C and then failed. In reality it's probably closer to the 75C your socket temp is reporting. so lets go with the lower end of my working hypothesis of your cpu temp really being reported as +5C too high. 75C is at the high end for a temp failure. Atleast we now know where your cpu fails, and at what temps.

Frankly your NH-D14 is performing fantastic with a 1.5000V core... so lets work on this overclock a bit alright?

get a note pad, and write down the vcore and multiplier settings that successfully work as you go. you need to learn your cpu.

go into your bios and reset everything to stock settings. save and restart. This overclock will be a quick and dirty just to establish where you chip wants...
ok... the reported amd cpu temps are not actual temps taken be a thermometer. AMD hasn't used an on chip thermometer since the phenom 1 days... what they use is some sort of mathematical algorithm or something. anyway, the end result is the core temps reported by AMD cpus are more of a "best guess" then they are fact (not that i'd put any more stock into intel core temps... cpu temp thermometers on an intel chip tend to be poorly calibrated and are little better then the amd "best guess")

Because we don't know if you're getting an accurate reported temp or not (piledriver temps tend to be reported inaccurately), i need you to look at your motherboard temps (not the cpu temps). the socket temp on the motherboard (the hottest reported temp on the motherboard) tends to be pretty close to the temps as your amd cpu is supposed to report them. So if your socket temp is off by more then 5C from your reported core temp i think we can safely say your cpu isn't reporting the temps right.

Now... lets say your cpu IS reporting temps right. The next most obvious issue would be how well your cpu cooler is seated and the application of your thermal paste. take the cpu cooler off, clean the surface of it and the cpu using rubbing alchohol (make sure the pc is unplugged). make those surfaces shine like a mirror, then put a ball bearing sized drop of thermal paste in the middle of the cpu heat spreader, set the heatsink down on it, and give it a little twist (not more then 30%) left and right before attaching it. make sure you don't rock it, or air will get into the space between the two parts. which will kill the heat transfer ability of the paste.

IF that doesn't fix it then we're going to have to look at your overclock. it's likely you're shoving too much vcore into your cpu, and i'll walk you through how to overclock it properly.
 

karanbraich

Honorable
Sep 10, 2013
24
0
10,510
Asus Ai Suite2.01.01 tells me,
At idle- CPU:37C and MB:28C
With Prime 95, temps quickly rise on the CPU to 70+ whereas temps on the MB remain at 28C.



Dear sir,
I have already paid due care to the installation process. I put on a pea size thermal paste and twisted the cooler to ensure it spread nicely.
 


1.49V is quite high, I personally wouldn't want to go much past 1.5V. Have you tried a different temp monitoring program like coretemp?
If so does it report the same temps are HWMonitor?
 

karanbraich

Honorable
Sep 10, 2013
24
0
10,510


Yes it does. Peak at around 80C.

What should I lower the Vcore to?

Ill be happy with 4.6Ghz. I dont want to push it to 4.8 anymore.
 


If that is the genuine temperature then it's really very high, I wouldn't want to go past 65-70C for stress testing with your CPU. It's differcult to say whether its a faulty reading from the software programs or an actual cooling problem...

Anyway, try 1.45V for 4.6Ghz, if that's stable then decrease it alittle and re-test for stability. Rinse and repeat until it becomes unstable then up the voltage by one notch to regain stability.
 

karanbraich

Honorable
Sep 10, 2013
24
0
10,510
A bit of a success. For the first time (1.45V) I am able to run Prime95 for 5 mins. It looks like the peak temp is 70 on HWmonitor, 68 on Asus Ai Suite and 72 on Coretemp.

HOWEVER, Prime95 is reporting Hardware failure on Worker#2 and 4. What can I do now sir?
 


It's not stable unfortunately, if any workers (CPU cores) fail or error then they are not stable, you probably need more Vcore.
However your temps are quite high so you may consider reducing the frequency (say to 4.4Ghz) to regain stability.
 

karanbraich

Honorable
Sep 10, 2013
24
0
10,510
Does this mean I have an ordinary chip which is not able to overclock to 4.6-4.8 under high end air coolers?

I have changed the multiplier to 22 so its running at 4.4Ghz and still getting the same temperature readings. Vcore is 1.464.
 


Yeah the idea of lowering the frequency is to regain stability, if you kept the voltage the same then the temps will be roughly the same.
 


No, start from where you were (1.464V) and then lower the voltage slightly, say to 1.450V and re-test for stability, if that's stable then lower it slightly again ETC...

When it becomes unstable, up the voltage alittle to re-gain stability.
 

karanbraich

Honorable
Sep 10, 2013
24
0
10,510


Yes it is a ROG board. I have updated the bios but still the same temps. I will be replacing thermal paste tomorrow morning. Will update after this.

Looking forward to your suggestions and guides.
 
ok. just want to make sure you can trust the reported temps or not. i wouldn't trust them because most fx cpus fail at that reported temp. which just means we're going to have to figure out how hot your cpu really is. there are a few ways to do this, but i think the most likely solution would be to simply "assume" your reported temps are probably 5C-8C too high... at idle your cpu will likely only be +5C over ambient (i'm assuming your ambient temps are pretty close to that 28C reported on the motherboard)... just a guess really... but we'll work with that as our "working hypothesis" about your cpu.

I can believe your cpu functions at -5C or -8C from your reported temps. lets see what happens with the thermal paste re-apply. if it still is kicking out similar temps we'll have to assume your cpu is off a little bit on it's reported temps... and that your cpu can function on the higher end of the FX temp scale.

Then i'll start walking you through overclocking this cpu right. I just don't want to have you overclock this cpu, depending on temp readings we're not certain about. and while crashing due to temps won't kill a cpu, it will shorten it's life.
 

karanbraich

Honorable
Sep 10, 2013
24
0
10,510

Alright sir, I have reapplied the thermal paste. Rotated the cooler 30 degrees left and right. I can notice the temperature increase slowly under prime95 test. My peak was 66C before my 8th core failed. Ambient temperature is 25C.

This is the latest screenshot of HWMonitor, Asus Ai Suite and Core temp with 4.6Ghz.
http://i942.photobucket.com/albums/ad270/karanbraich/Temps_zpsf3dca6e1.png
 
ok. i'll buy your cpu hit 80C and then failed. In reality it's probably closer to the 75C your socket temp is reporting. so lets go with the lower end of my working hypothesis of your cpu temp really being reported as +5C too high. 75C is at the high end for a temp failure. Atleast we now know where your cpu fails, and at what temps.

Frankly your NH-D14 is performing fantastic with a 1.5000V core... so lets work on this overclock a bit alright?

get a note pad, and write down the vcore and multiplier settings that successfully work as you go. you need to learn your cpu.

go into your bios and reset everything to stock settings. save and restart. This overclock will be a quick and dirty just to establish where you chip wants to go. there is a lot of other things you can do if you want to go nuts with this overclock. at the moment i want to find what volts your cpu actually needs for what speeds.

Turn on LLC, change your cpu multiplier to manual, turn the vcore to manual, set the vcore at the stock vcore for a 8350 at 1.3500V, now bump your cpu multiplier up +0.5, save and restart, see if the computer will load into windows. if it does, restart the machine, and bump your multiplier by another +0.5, keep doing the above until the computer fails to POST or fails to load into windows (it probably will fail to load into windows before it fails to post... but who knows?). if you fail to post, don't worry, after 3 failed posts, your motherboard will reset your bios back to stock settings, just remember to set everything back to your last successful posting settings, and restart, make sure you can still post to windows.

We now know you need to start to add vcore, so what we're going to do is add vcore to get your cpu prime stable. Bump your vcore by .0125V, restart, see if you can load into windows. If you can't bump it by another .0125... if you can load up prime and see if you crash (you may or may not... we'll see). if you crash back to windows, a core fails, the computer freezes and locks up or you black screen reset your cpu is still undervolted. Keep an eye on your cpu's temps as well, you'll notice as temps get higher the cpu will get more unstable and need more vcore to stabilize, which of course will make your cpu unstable. basically once you start to add vcore you are nearing the end of the overclock (or atleast its on the horizon, depends on the chip).

So once you have found the vcore needed to keep the cpu running through a few passes of prime, and there are no serious heat issues and your vcore is still under 1.45V (1.45V-1.5V is generally the safe max on air cooling for a fx cpu, to put more vcore into the chip you either need a seriously magical chip that has no heat issues for magic reasons, or a custom water loop) you will go back into the bios and bump the multiplier up another +0.5. load into windows and run prime95 for a few passes. if the cpu makes it, jump back into the bios and bump your multiplier up another +0.5, continue the cycle of bumping multiplier, plus running a few passes of prime and maybe when it crashes bumping vcore until it can run prime.

When your vcore reaches 1.45V (or more) and your temps are breaking 70C when running prime, you've reached the end of your overclocking with that chip. There are more things you can do to squeeze more performance out of your system (such as playing with your FSB/Cpu frequency) but i'm just doing a quick and dirty here to find the "near" limits of your cpu. if you're unhappy with the result, i'll walk you through FSB overclocking and perhaps your FX will respond better to that instead.
 
Solution

TRENDING THREADS