AMD FX-8320 (OC'ed to 4.0Ghz on A Corsair H100) Unstable Idle Temps?

[Cloud92]

 

Oh damn, guess I got lucky:

 

 

 

a 5min run in prime95 please. I still don't believe it,lol. You either got the best chip in the world which I doubt or something's wrong. and a screenshot while it's running the benchmark with CPU-z,I still want to see the voltage.

 

oh and 1 more thing,I have the same cooler,212,I'm pretty sure it's not supposed to handle a 8 core at over 4.5 - 4.6ghz without mauling the chip from the heat,are you 100% sure it's not throttling down?That's why I'm asking for screenshots while it's at 100% load

[DaCoder]

a 5min run in prime95 please. I still don't believe it,lol. You either got the best chip in the world which I doubt or something's wrong. and a screenshot while it's running the benchmark with CPU-z,I still want to see the voltage.

 

oh and 1 more thing,I have the same cooler,212,I'm pretty sure it's not supposed to handle a 8 core at over 4.5 - 4.6ghz without mauling the chip from the heat,are you 100% sure it's not throttling down?That's why I'm asking for screenshots while it's at 100% load

 

Well, just changed all the settings as you instructed, to a tee.

Core speed is 1.4 Ghz when not under load, and then jumps straight to 4 Ghz every few seconds (probably from CPU usage)

CPU usage btw: is under 3 percent always (even with chrome open)

 

Voltage: drops down to .888 V idle (chrome open) and with prime95 stress: 1.248 V

Note: when idle and it kicks up to 4 Ghz, voltage is 1.270 V

 

Temperatures are 66 degrees C stress, and 29 to 44 C idle.

 

EDIT: lol this scared me silly, i played some CSGO, cus i noticed that one of my GPU fans were not spinning, thinking it was a feature to save noise, power, etc. I started playing, and suddenly i hear: *tick tick tick tick* really fast. I am sitting here thinking its my H100 pump, but then I see the fan is hitting a 6 pin to 8 pin molex connector, which must have been pushed in there from the trip back from QuakeCon *WHEW*

[KeltonDSMer]

a 5min run in prime95 please. I still don't believe it,lol. You either got the best chip in the world which I doubt or something's wrong. and a screenshot while it's running the benchmark with CPU-z,I still want to see the voltage.

 

oh and 1 more thing,I have the same cooler,212,I'm pretty sure it's not supposed to handle a 8 core at over 4.5 - 4.6ghz without mauling the chip from the heat,are you 100% sure it's not throttling down?That's why I'm asking for screenshots while it's at 100% load

 

Oh those CB runs are not at a voltage that would be stable in P95. Those were panel open, box fan blasting, "let's see what this thing can do without melting" type of bench. You can tell from the scores that there is most certainly no throttling...the numbers have near linear scaling compared to any 8320 at 4.0ghz. There is a thread on Cinebench scores with a GoogleDoc that has lots of 8core piledrivers right there with my numbers.

 

The CPU hit 70C for a split at the end of the R15 run, I don't want the chip to get that hot again.

 

The highest stable OC with the panels on and at least two hours P95 stable is 4.5ghz at 1.32v under load. There is no doubt that this cooler is pretty limited on this CPU as you said!

 

I just tried to run CB again at 4.88 on the same profile and it failed...those screenshots were right on the edge of stability. It failed when I took the screenshot during the run lol

 

All I could get is another shot of the desktop at 1.32v for 4.88....hardly much proof for you sorry. I will soon test again trying to bump past 4.9 on some cool morning. I'll make a note to send you some screenshots or use the @ thing to notify you of the thread I post them in.

 

Sorry OP, didn't mean to jumble your thread.

 

@vern021 , I would use a higher setting LLC as you go up in frequency before adding more base voltage. There is such a big Vdroop under load and LLC can take care of that for you without using an unnecessary voltage for idle and light loads. Also, I would disable the setting that steps the frequency up and down, "Cool and Quiet" I think it is called (not positive). That can really goof an otherwise stable OC profile.

[DaCoder]

Oh those CB runs are not at a voltage that would be stable in P95. Those were panel open, box fan blasting, "let's see what this thing can do without melting" type of bench. You can tell from the scores that there is most certainly no throttling...the numbers have near linear scaling compared to any 8320 at 4.0ghz. There is a thread on Cinebench scores with a GoogleDoc that has lots of 8core piledrivers right there with my numbers.

 

The CPU hit 70C for a split at the end of the R15 run, I don't want the chip to get that hot again.

 

The highest stable OC with the panels on and at least two hours P95 stable is 4.5ghz at 1.32v under load. There is no doubt that this cooler is pretty limited on this CPU as you said!

 

I just tried to run CB again at 4.88 on the same profile and it failed...those screenshots were right on the edge of stability. It failed when I took the screenshot during the run lol

 

All I could get is another shot of the desktop at 1.32v for 4.88....hardly much proof for you sorry. I will soon test again trying to bump past 4.9 on some cool morning. I'll make a note to send you some screenshots or use the @ thing to notify you of the thread I post them in.

 

Sorry OP, didn't mean to jumble your thread.

 

@vern021 , I would use a higher setting LLC as you go up in frequency before adding more base voltage. There is such a big Vdroop under load and LLC can take care of that for you without using an unnecessary voltage for idle and light loads. Also, I would disable the setting that steps the frequency up and down, "Cool and Quiet" I think it is called (not positive). That can really goof an otherwise stable OC profile.

 

So do you know why my CPU has such bad voltage / temps? I am left to guess a bad seating of thermal paste, temps are causing it to throttle temps.

 

Will re-seat tomorrow, aswell as dust stuff.

Also: I might have bent a pin on install, had to straighten it again, i will check to see if it went in correctly.

 

EDIT: found some extra thermal compound around, used it. Found that the spreading wasn't spread properly, so i restarted with new (hand spread with finger) paste (I hear it makes air-pockets?) and now the thing is working great! Or at least it was... once again I see that the voltage is 1.24 ish, so I cranked the Ghz back to 4.2 (where it was stable for 3 months before) and did a prime95 test. Well after about 15 milliseconds half the threads failed, half kept going. Voltage was an all time low: 1.247 V and my temps skyrocketted to 70 C. Where i turned off the test pronto.

I then lower the Ghz back down to 4.0, now the temps rocket up to 70C and past it, at least none of the threads fail, also the idle fluctuation is gone after the reseatting, but I still get over 30 C change from opening Chrome.

 

So it went from being perfect, under 50 C under full Prime95 stress, to rocketting past 70C!? All while having low voltage...

 

EDIT2: Changing the Corsair fan speed has no effect on idle temps nor stress temps. Leading me to believe contact surface (i just re-did that), or pump.

Corsair pump RPM: 2333 RPM avg.

 

Doing some research, and apparently the H100 cooler is not very efficient at cooling, at all. Should I go ahead and get a newer Corsair unit?

[KeltonDSMer]

So do you know why my CPU has such bad voltage / temps? I am left to guess a bad seating of thermal paste, temps are causing it to throttle temps.

 

Will re-seat tomorrow, aswell as dust stuff.

Also: I might have bent a pin on install, had to straighten it again, i will check to see if it went in correctly.

 

EDIT: found some extra thermal compound around, used it. Found that the spreading wasn't spread properly, so i restarted with new (hand spread with finger) paste (I hear it makes air-pockets?) and now the thing is working great! Or at least it was... once again I see that the voltage is 1.24 ish, so I cranked the Ghz back to 4.2 (where it was stable for 3 months before) and did a prime95 test. Well after about 15 milliseconds half the threads failed, half kept going. Voltage was an all time low: 1.247 V and my temps skyrocketted to 70 C. Where i turned off the test pronto.

I then lower the Ghz back down to 4.0, now the temps rocket up to 70C and past it, at least none of the threads fail, also the idle fluctuation is gone after the reseatting, but I still get over 30 C change from opening Chrome.

 

So it went from being perfect, under 50 C under full Prime95 stress, to rocketting past 70C!? All while having low voltage...

 

EDIT2: Changing the Corsair fan speed has no effect on idle temps nor stress temps. Leading me to believe contact surface (i just re-did that), or pump.

Corsair pump RPM: 2333 RPM avg.

 

Doing some research, and apparently the H100 cooler is not very efficient at cooling, at all. Should I go ahead and get a newer Corsair unit?

 

Are you sure you are reading the core temp and not the socket temp? 70C on the core at only 1.275v should not happen with that cooler, but I could see that being the case for the socket temp.

 

What are you using to read temperature?

[Cloud92]

Oh those CB runs are not at a voltage that would be stable in P95. Those were panel open, box fan blasting, "let's see what this thing can do without melting" type of bench. You can tell from the scores that there is most certainly no throttling...the numbers have near linear scaling compared to any 8320 at 4.0ghz. There is a thread on Cinebench scores with a GoogleDoc that has lots of 8core piledrivers right there with my numbers.

 

The CPU hit 70C for a split at the end of the R15 run, I don't want the chip to get that hot again.

 

The highest stable OC with the panels on and at least two hours P95 stable is 4.5ghz at 1.32v under load. There is no doubt that this cooler is pretty limited on this CPU as you said!

 

I just tried to run CB again at 4.88 on the same profile and it failed...those screenshots were right on the edge of stability. It failed when I took the screenshot during the run lol

 

All I could get is another shot of the desktop at 1.32v for 4.88....hardly much proof for you sorry. I will soon test again trying to bump past 4.9 on some cool morning. I'll make a note to send you some screenshots or use the @ thing to notify you of the thread I post them in.

 

Sorry OP, didn't mean to jumble your thread.

 

@vern021 , I would use a higher setting LLC as you go up in frequency before adding more base voltage. There is such a big Vdroop under load and LLC can take care of that for you without using an unnecessary voltage for idle and light loads. Also, I would disable the setting that steps the frequency up and down, "Cool and Quiet" I think it is called (not positive). That can really goof an otherwise stable OC profile.

NOW THAT is more like it,as I was saying,at 1.320v yes it can handle stability tests at 4.5ghz,mine does it as well,but I was saying that there is no way in hell it's stable 1.320v and 4.9 ghz,I bet if you turn down the llc it won't even load windows without bsod. I knew it can't possibly handle the clock at that voltage during a stability test and the temperature since I have the same exact setup and voltages. Your chip seems to be even better than mine,mine does 4.4 100% stable stock 1.32,not sure about 4.5,never stress tested it to see if it bsods. But anyway. not bad,you lucky bastard,you have one hell of a chip,and I thought mine was good with 4.4 1.320v

[DaCoder]

Are you sure you are reading the core temp and not the socket temp? 70C on the core at only 1.275v should not happen with that cooler, but I could see that being the case for the socket temp.

What are you using to read temperature?

HWmonitor

[DaCoder]

This is just impossible,it's out of this world,something is terribly wrong,it's either the chip or the motherboard,it's most likely a HORRIBLE chip that cannot overclock at all,or something is wrong with the motherboard.Try running prime95 on stock settings with turbo on and see if it fails,maybe you're lucky and you can RMA the chip. That is just not right,70 degrees on 1.24v , you shouldn't even pass 40 degrees at that voltage at 100% load with your h100. But if you're saying you ran 4.2 stable for 3 months and now you can't it looks like your chip has degraded alot,that's not normal,better RMA it

oh shit double post

you dont think its the H100 failing, or the pump not keeping stable speed? Gonna hook it up to molex as one person stated.

 

I have found several other postings of people who solved this exact issue by replacing the cooler, only thing is, why is the voltage so low?

 

Redid the connectors just to check, reseatted CPU again, and:

 

So its working great right now, I would like a higher clock speed from the CPU, but I am afraid that opening UEFI will cause that strange problem again.

[KeltonDSMer]

you dont think its the H100 failing, or the pump not keeping stable speed? Gonna hook it up to molex as one person stated.

 

I have found several other postings of people who solved this exact issue by replacing the cooler, only thing is, why is the voltage so low?

 

Redid the connectors just to check, reseatted CPU again, and:

 

So its working great right now, I would like a higher clock speed from the CPU, but I am afraid that opening UEFI will cause that strange problem again.

 

You ask why the voltage is so low, what is it set to in the bios? What is CPU LLC set to?

 

That last screenshot is close to what we need, can you run P95 for like 5 minutes with the HWmonitor window showing Vcore, socket temp, and cpu temp? The window in your picture just shows core temp and it looks like P95 was just started. The tab under your motherboard in HWmonitor should have a temp section that shows socket temp, that's what I am really curious about.

[DaCoder]

You ask why the voltage is so low, what is it set to in the bios? What is CPU LLC set to?

 

That last screenshot is close to what we need, can you run P95 for like 5 minutes with the HWmonitor window showing Vcore, socket temp, and cpu temp? The window in your picture just shows core temp and it looks like P95 was just started. The tab under your motherboard in HWmonitor should have a temp section that shows socket temp, that's what I am really curious about.

 

Will post, and voltage is auto.

[Bagmup]

You positive the pump is running on the H100? Doesn't sound like it is.

[DaCoder]

You ask why the voltage is so low, what is it set to in the bios? What is CPU LLC set to?

 

That last screenshot is close to what we need, can you run P95 for like 5 minutes with the HWmonitor window showing Vcore, socket temp, and cpu temp? The window in your picture just shows core temp and it looks like P95 was just started. The tab under your motherboard in HWmonitor should have a temp section that shows socket temp, that's what I am really curious about.

 

Well, one day later, the temps are right back up to where they should not be.

 

Idle temps are back up to 40 degrees average. This is ridiculous, I will go get another water cooler and update you guys on if that fixed the issue.

Also, the temps on the motherboard did hit 70, but only after a long time. (5 min)

[KeltonDSMer]

Well, one day later, the temps are right back up to where they should not be.

Idle temps are back up to 40 degrees average. This is ridiculous, I will go get another water cooler and update you guys on if that fixed the issue.

Also, the temps on the motherboard did hit 70, but only after a long time. (5 min)

 

Thanks for posting another picture, there are definitely a few things to figure out from what I can see. The socket and chip are getting a bit too hot for sure.

 

The PSU voltages are also very concerning to me. The 12V and 5v rails are reading like 30% low...I think maximum delta is 5% in either direction. Either HWmonitor is reading incorrectly, or there is a serious power supply issue that needs to be resolved. Do the PSU voltages read the same in other monitoring software?

 

I don't really see how the hot socket/chip and the low PSU voltages could be related unless the mobo VRMs are needing to draw so much extra current to supply the system because of the low PSU voltage. I didn't even know a system would run with 8v on the 12v system, but the VRMs would have to work very hard to power the system if they are only getting 2/3 the voltage they are meant to run. The cpu temp, however, I would think should be unaffected if the Vcore is correct; 1.2xx volts is super safe and should not be producing that much heat on the chip with any AIO.

 

This has me stumped, just doesn't make much sense. I sure hope someone with more experience can chime in and help you get this resolved. I am going to think about it some more but that low 12v and 5v reading really has me concerned. I would run stock speeds at as low of voltage as possible until you figure this out.

 

I am not familiar with the GB mobos, but socket temps above 70-72c could surely make the power delivery throttle which could explain the instability you have been experiencing.

 

Best of luck man, I would definitely figure out what is going on with those PSU voltages while you are waiting for the cooler. Even if the AIO is bad and that is why the chip is heating up so much, the socket temps are still concerning and need to be addressed somehow. The stock cpu fan blows down and provides airflow to the power delivery components, so using an AIO or tower cooler like the CM HE212 can negatively impact board cooling in that regard.

[DaCoder]

Thanks for posting another picture, there are definitely a few things to figure out from what I can see. The socket and chip are getting a bit too hot for sure.

 

The PSU voltages are also very concerning to me. The 12V and 5v rails are reading like 30% low...I think maximum delta is 5% in either direction. Either HWmonitor is reading incorrectly, or there is a serious power supply issue that needs to be resolved. Do the PSU voltages read the same in other monitoring software?

 

I don't really see how the hot socket/chip and the low PSU voltages could be related unless the mobo VRMs are needing to draw so much extra current to supply the system because of the low PSU voltage. I didn't even know a system would run with 8v on the 12v system, but the VRMs would have to work very hard to power the system if they are only getting 2/3 the voltage they are meant to run. The cpu temp, however, I would think should be unaffected if the Vcore is correct; 1.2xx volts is super safe and should not be producing that much heat on the chip with any AIO.

 

This has me stumped, just doesn't make much sense. I sure hope someone with more experience can chime in and help you get this resolved. I am going to think about it some more but that low 12v and 5v reading really has me concerned. I would run stock speeds at as low of voltage as possible until you figure this out.

 

I am not familiar with the GB mobos, but socket temps above 70-72c could surely make the power delivery throttle which could explain the instability you have been experiencing.

 

Best of luck man, I would definitely figure out what is going on with those PSU voltages while you are waiting for the cooler. Even if the AIO is bad and that is why the chip is heating up so much, the socket temps are still concerning and need to be addressed somehow. The stock cpu fan blows down and provides airflow to the power delivery components, so using an AIO or tower cooler like the CM HE212 can negatively impact board cooling in that regard.

 

I feel relatively certain it either three things: bad PSU, since it was half off! And/or a bad cooler since it was half off! (I buy things when they are on sale, alot)

Or it could be the motherboard, I returned one that was so broken, you couldnt even use BIOS/UEFI! This is the third one I got from Microcenter.

[DaCoder]

Update: my PSU was defective, it was sending the incorrect voltages to all of my hardware. 12v was getting 6v etc.

Note to self: never buy TT PSU ever again