FX-8120 temps

[Joni]

Are these temperatures normal for FX-8120. Idle 36 C and in Prime95 blend (8 cores) it gets to 61 C in about 10 minutes, thats when I stopped it. This is on stock speeds, case open, Noctua D-14 with the two fans on ulna adapters and Noctua high grade thermal paste. Temperature readings are from CPU (NOT socket readings, those are about 16 C idle).

I've read alot of conflicting reports about 8120 @ stock speeds, some say they have 23 C idle on air, makes me wonder if mine should be that low or are they reading core idle temps.

 

[RGone]

Idle 36 C = that should certainly be 'high' for that cooler if CORE TEMPS are measured with HWMonitor. Which makes me believe your CPU cooler is not seated very well or it is not getting cool air and a good exit for the air after it passes thru the cpu cooler.

 

[Joni]

It's an open case atm with all the fans still on so it must be something else, just to clarify CPU is idle 36C while cores are at 16 C with HWMonitor.

Heatsink doesn't feel warm at all, Radeon 4870 right below it feels too hot to touch, it's showing as 82 C.

 

[bassnut]

36*C at Idle and 61*C on prime thats your socket temp. what you want to keep an eye on is your cpu temps. I found on my 8120 that once the socket temp gets up over 70*C ish it can cause a bit of instability. Some people have added extra cooling to the back of thir motherboards to help with this and shave off a few *C. Your temps look normal for an FX8120.

 

[Joni]

36*C at Idle and 61*C on prime thats your socket temp. what you want to keep an eye on is your cpu temps. I found on my 8120 that once the socket temp gets up over 70*C ish it can cause a bit of instability. Some people have added extra cooling to the back of thir motherboards to help with this and shave off a few *C. Your temps look normal for an FX8120.

I thought socket and cpu temps were the same thing? Sorry, last time I overclocked was years ago, Compaq with 300mhz k6-2, multiplier had to be changed from the motherboard with jumper .

Core temp tells the CPU temp? It has Tj. max 90 C and CPU #0 36 C? Also AMD said safe operating temperature is below 61 C, are they talking about socket or CPU temperature? What about bios CPU temperature, is that from the socket or CPU?

 

[trents]

If you are using CoreTemp to monitor temps then "CPU" temp is the same as core temp. With HWMonitor, core temp is core temp and CPU temp is socket temp. Sometimes HWMonitor will display CPU (socket) temp as one of the TMPINx lines when it cannot specfically identify the socket temp sensor as such and will give it a generic label.

 

[Joni]

I'm now at 4.0 GHz, stock voltage. Anything above that freezes as soon as I start Prime95. On HWMonitor socket idle is same 39 °C on Prime95 it gets to 62 °C in about ten minutes, thats when I stop.

On wPrime it goes to 61 °C but newer above it, I wonder if this would be safe to keep at 4.0 GHz, I mean would I ever get near the stress levels of these apps on normal usage?

I would like to get to 4.2 GHz but dunno what's the problem. I have disabled all the usual stuff on bios, windows power management is set to performance. I tried to add voltage + 0,15, but still freezes. I wonder if this is the limit for my Enermax 550W PSU.

I wasn't expecting to get this mutch though.

 

[trents]

What does HWMonitor say your core temps are when it says your CPU temps are 61 when stressing with Prime95? With good aftermarket cooling like you have core temps, the really important ones (when the sensor is calibrated accurately, at least) are typically less than socket (CPU) temps, often about 10 C less. We often tell people that AMD CPUs are stable up to a core temp of about 55 C. Sounds like you could have a little temp room yet to add some core voltage and stabilize that overclock at 4.0 anyway.

It would be helpful if you would run Prime95 blend with HWMonitor open on the desktop and attach a pic of HWMonitor using the Go Advanced button at the bottom of any new post window. Then click on the little paper clip icon at the top of the Advanced post window. That will cause a file browser/upload window to appear and the rest will be obvious. But first frame, crop and save the pics with Snipping Tool in Windows accessories.

Also, have you supplemented the CPUNB voltage and the ram voltage any? That can often help progress an overclock.

 

[Joni]

Below is a pic of HWMonitor when it reaches 61 °C (4.0 GHz). I haven't touched any voltages or overclocked NB/RAM. My memory is Kingston 1866 MHz, I had to add the timings manually to run at 1866, could 1866 MHz memory cause some problems when overclocking?

I'll try if adjusting NB and RAM voltages helps.

 

[ssjwizard]

Your core temps are only at 46C which is the important measurement. That 61C CPU temp is the air space below the CPU. Its stagnant air and should be ignored. Your Core #0, #1, Ect should not go over 55C for maximum stability. For benchmarking I have pushed my 8120 as high as 64C core temp without an issue. Go ahead and run Prime blend for 1-2 hours to verify stability. Possibly some 3dMark if your going to be using your system for playing games also.

I am running my 8120 @ 4.0Ghz 1.25 vCore that bumps up to 1.28 under load with LLC set to minimum.

 

[mjw21a]

Your core temps are only at 46C which is the important measurement. That 61C CPU temp is the air space below the CPU. Its stagnant air and should be ignored. Your Core #0, #1, Ect should not go over 55C for maximum stability. For benchmarking I have pushed my 8120 as high as 64C core temp without an issue. Go ahead and run Prime blend for 1-2 hours to verify stability. Possibly some 3dMark if your going to be using your system for playing games also.

I am running my 8120 @ 4.0Ghz 1.25 vCore that bumps up to 1.28 under load with LLC set to minimum.

Unless your case has an air intake behind the socket and space for a fan that is. The majority of cases on the market don't so (as usual) ssjwizard is likely right.

 

[ssjwizard]

Unless your case has an air intake behind the socket and space for a fan that is. The majority of cases on the market don't so (as usual) ssjwizard is likely right.

Both my rigs have a fan port behind the motherboard and I still dont use them. Seems like a wasted source of noise to me. Why worry about the resistors on the back of the board when everything thats heatsinked already isnt overheating.

 

[mjw21a]

True. If it needed cooling, they'd have placed the bits on the other side of the board around the socket. Besides, I'm sure heat is factored into the mainboard design.... Well, with decent brands like Asus and Gigabyte at least.

 

[trents]

I wouldn't be so quick to dismiss the CPU socket temp as unimportant. One thing we know by now is that core temp some motherboards have thermal down-throttling mechanisms that operate independently of the core temp and so in those cases you have to consider the socket temp as well. Having said that, it would appear Joni that you do have some room to increase your CPU voltage and progress your overclock.

 

[Joni]

FX-8120 @ 4.4 GHz now Temps look good but needs some more testing.

 

[trents]

Let's see the "Memory" and "SPD" tabs as well from CPU-z.

 

[Joni]

Memory and NB is stock. Hmm, looking at those tabs I think I have set tRC wrong.

 

[Woomack]

Nah, tRC is fine as long as it's stable. You can change it but it won't make bigger difference. Funny is that 2800 profile ... good luck with 2800 9-9-9

 

[Joni]

I don't understand, this just doesn't make any sense to me. On default settings this CPU uses these vcore voltages when using prime95:

3.1 GHz 1.236 (8 workers)
3.4 GHz 1.284 (4 workers)
4.0 GHz 1.380 (1 worker)

If I disable amd turbo mode and change multiplier to 20x (4.0 GHz) everything is fine on prime95 with default voltage 1.212. Then I change to 21x (4.2 GHz) and it freezes as soon as i start prime, then I add voltage, eventually up to 1.4 and still the same thing, freezes as soon as I start prime. Now if I turn LLC high (50%) everything works fine with extensive prime testing again on default voltage, about 1.29.

 

[trents]

CPUs are like that in their "scaling" with regard to frequency overclocking and voltage need to be stable. You can get 885-90% of the max overclock on a very modest voltage increase but that last .1 ghz takes a lot more than the first .6 -.8 ghz overclock did. It's the "wall". You have to decide if that last little bit of overclock that needs so much more juice and generates so much more heat is worth the greater wear and tear on the system.