FX-8320, Sabertooth 990FX, GTX 970 - overclock (OC) and temperatures

0v3rl0rd

Reputable
Dec 3, 2014
37
0
4,530
Hi

First things first - I have the following configuration:
- MBO: Asus Sabertooth 990FX R2.0
- CPU: AMD FX 8320 + Cooler Master 212 Evo cooler (with Phanteks PH-F140HP PWM fan on it)
- RAM: 2 x 4 GB Mushkin Ridgeback (997046) @ 1600MHz
- GPU: Asus GTX 970 OC Strix
- PSU: OCT Fatal1ty 750W
- Storage: Samsung 840 Evo 120GB + WD Purple (WD10PURX-64D85Y0)
_ Case: Cooler Master N500 (3 front intake 120mm fans, 1 exhaust 120mm at the back).

I decided to overclock a little bit and downvolt it also. The following settings are in BIOS:

**************************************************
**************************************************

AI Tweaker:
- Ai Overclock Tuner: Manual
- CPU Ratio: 20.0
- AMD Turbo CORE technology: Disabled
- CPU Bus Freq.: 200
- PCIE freq.: Auto
- Memory freq.: DDR3-1600MHz
- CPU/NB freq.: 2200MHz
- HT Link Speed: 2600MHz
- CPU Spread Spectrum: Disabled
- PCI Spread Spectrum: Disabled
- EPU Power Saving Mode: Disabled

- CPU & NB Voltage: Manual Mode
- CPU Manual Voltage: 1.212500
- CPU/NB Manual Voltage: Auto (1.175V)
- CPU VDDA Voltage: Auto (2.510V)
- DRAM Voltage: 1.50000
- NB Voltage: Auto (1.103V)
- NB HT Voltage: Auto (1.107V)
- NB 1.8V Voltage: Auto (1.800V)
- SB Voltage: Auto (1.100V)
- VDD PCIE: Auto (1.100V)
- VDDR: AUto (1.200V)


DIGI+ Power Control:
- CPU Load Line Calibration: High
- CPU/NB Load Line Calibration: High
- CPU Current Capability: 130%
- CPU/NB Current Capability: 130%
- CPU Power Phase Control: Standard
- CPU Voltage Freq.: Auto
- VRM Spread Spectrum: Disabled
- CPU Power Duty Control: T.Probe
- CPU Power Response Control: Auto
- CPU/NB Power Response Control: Auto
- CPU Power Thermal Control: 130
- DRAM Current Capability: 100%
- DRAM Voltage Freq.: 300
- DRAM Power Phase Control: Optimized

CPU configuration:
- Cool'n'Quiet: Always Disabled
- C1E: Disabled
- SVM: Disabled
- Core C6 State: Disabled
- HPC mode: Enabled
- Apm Master Mode: Disabled

**************************************************
**************************************************

So, as You can see, I overclocked it to 4 GHz (just because I wanted it "rounded" instead of 3.5 GHz) using multiplier of 20.
I tried with voltages of 1.1875, 1.19375, 1.2, 1.20625 with no success: worker 8 was throwing errors.

Then, I set the voltage to 1.2125 volts and tested it:
- 10 hours of "Small FFT" with Prime95
- 10 hours of "Blend" with Prime95
- 2 hours with Aida64 stability test (with everything checked except LOCAL DISKS).
It passed everything without any errors or bluescreens.

The readings from AIDA and HWmonitor were the following:

2427wd2.jpg


*Note1: LOAD temperatures are the highest that I recorded in all tests together.
*Note2: Ambient temperature at 20°C.

**************************************************
**************************************************

Now, I have a few questions:

1) At one point (when the test was on, so, at full CPU LOAD - Prime95), HWMonitor registered a max volts on a CPU as high as 2.508. Is that even possible if that is set to 1.2125 volts in BIOS? Or is it just a reading error?
2) Are those LOAD temperatures "normal" or even "safe"? The ones I'm refering to are:
- CPU (socket)
- CPU (core)
- Motherboard VRM's - is this supposed to fly so high (even higher than the CPU itself?)?
- GPU diode & GPU VRM
3) AFAIK, the CPU1, CPU2, ..., CPU8 are those CORE temperatures that are safe until 60°C, and the CPU is the socket that is safe until 70°C. Is that true?
4) This question is about IDLE temperatures: is it normal that this CPU idles at 35°C (socket)? Shouldn't it be lower considering the following:
- I lowered voltage to 1.2125
- Cooler Master 212 Evo
5) This is about "what after overclocking". Since this is my first time I actually overclock (had to gather as many info as possible before I had the guts to do it :D ), I'm have no idea what should I do now that I reached a stable overclock & undervolt.

What I mean is: when all the settings in BIOS were at their default values, the CPU voltage (in Windows) would range from 0.728 to higher (1.272, etc.) - depending on the CPU usage of course. Now, when OC, the voltage is, let's say, fixed to set value.
What should I do to make it "drop" when the CPU usage is very low (idle)?
Also, do some of the settings (Cool'n'Quiet, C1E, SVM, C6 state, HPC - disable?, APM master mode) need to be re-enabled to make the CPU run cooler at IDLE?
 
Solution
1) False reading probably.
2) Temps are fine. Yeah the VRM's can get pretty hot, it's normal.
3) I believe the max CPU socket temp for your cpu is 61C. You shouldn't really look at your core temps.
4) You have CnC disabled (which is good while OC'ing, so leave it on), so all cores will always run @ 4Ghz. That's why it's that "warm".
5) To run your pc as stable as possible, keep your voltages at a fixed value.
What's wrong with 35C idle? It's a nice temp.

KenGrinsven

Reputable
Dec 15, 2014
365
0
4,960
1) False reading probably.
2) Temps are fine. Yeah the VRM's can get pretty hot, it's normal.
3) I believe the max CPU socket temp for your cpu is 61C. You shouldn't really look at your core temps.
4) You have CnC disabled (which is good while OC'ing, so leave it on), so all cores will always run @ 4Ghz. That's why it's that "warm".
5) To run your pc as stable as possible, keep your voltages at a fixed value.
What's wrong with 35C idle? It's a nice temp.
 
Solution

0v3rl0rd

Reputable
Dec 3, 2014
37
0
4,530


1) & 2) & 3) Thanks, You confirmed my doubts
4) Yes, I thought that CnC might be the reason for "dropping" voltage. Isn't there something that I could possibly turn on so that it doesn't run at 4GHz all the time? Because, mostly I don't need it to run at that freq > I leave my PC 24/7 on, and I'm not sitting in front of it all of that time. so, that's the main reason.
5) For example, right now when I'm writing this post:
  • - running: aida64 and chrome with only few tabs opened
    - my CPU usage is less than 3-5% in 95% of time
    - it's 20°C ambient
    - and my CPU temp is 39°C.
If the temp stayed a few degrees under (that 35°C) I would be ok, but right now, it is at 39°C and I'm not even using it :??:
 

KenGrinsven

Reputable
Dec 15, 2014
365
0
4,960


Hmm.. You could always try to enable CnC. But I think it won't work.
If it crashes, you could lower your OC and try again.

The thing is, if you're gaming (for example), and CnC lowers the voltage too much, your pc will probably crash, because your cpu will just need more voltage when it's overclocked.
 

0v3rl0rd

Reputable
Dec 3, 2014
37
0
4,530



I tried enabling CnC but it didn't work. I tried every possible combination of CnC, C1E, C6 state, HPC and APM mode, and nothing allowed it to undervolt.

I reverted back to stock settings (3,5 GHz) - undervolt using a minus offset (-0,1375) and I get voltages from 0,876 to 1,176, and that is stable on clock freq - me likez :)

Now I doing some more stability tests and temperature logging to see the differences before I try to undervolt it more. And after that, OC to 4 GHz using offset (not manual cause I think that that's the one that fixes the voltage to a static non-changing value even if CnC enabled.


The wierd thing is, if I set a negative offset to:

  • - 0,05 ... stable
    - 0,0625 ... I got BSOD
    - 0,1 ... stable
    - 0,1375 ... stable

As You can see, I got a blue screen on a -0,0625 offset.. What I do remember is that I was monitoring the stats with multiple tools (Aida, CPU-z, HWMonitor, and AI Suite II in the background) at the same time, and also, Prime95 small FFT's was running.
[strike][/strike]
What I thought of at first was: "it was stable when undervolting with TurboV Evo to 1,175 volts."
Why did I get the BSOD if the offset of -0,0625 did not bring the voltage down even near 1,175 (as I earlier got to with TurboV Evo)? It dropped the voltage down to just 1,260 (max) and 0,912 (min).

Is it possible that BSOD happened because of multiple readings from different tools?

Also, one "point" to that - I turned on Aida64 logging, and when I got back today (when Prime95 testing on -0,1375 offset), some minimum temps were 0 and some max were 95:
o5tb42.jpg

This, I presume, is a false reading. Could this be partly the reason of the above described behavior?