OC FX8320, several concerns

Lucasrio267

Honorable
Feb 1, 2014
8
0
10,510
Hi,
I bought a few days ago a FX8320 to replace my Phenom II x4 965 BE. Willing to OC it, first I tested it with stock frequencies and I'm rather disappointed. :(

Here are my current specs :
-FX8320 @Stock
-Asrock 970 Extrem 4 (UEFI updated)
-8 Go Corsair Ballistix PC12800
-Sapphire Radeon HD 7950 OC
-Noctua NH NH-U12P SE2
-Antec HCG 620W modular
-Factal Design Define R4 with 6 fans
-SSD Samsung 840 120Go + 1To HDD

There are the charts I get from OCCT :
-With large data set, 16 Threads for 2 hours this night :
Frequencie : http://puu.sh/6FVJt.png
Temperature : http://puu.sh/6FVJL.png
Vcore : http://puu.sh/6FVJU.png
CPU usage : http://puu.sh/6FVK3.png
FSB frequencie : http://puu.sh/6FVK7.png

-As the variations were too big and unreadable, I made a new stress test this morning for 10 minutes, with the same settings :
Frequencie : http://puu.sh/6FVWO.png
Temperature : http://puu.sh/6FVX1.png
Vcore : http://puu.sh/6FVX5.png
CPU usage: http://puu.sh/6FW11.png
FSB frequencie : http://puu.sh/6FVWE.png

And here are the settings I changed from the UEFI regarding CPU and RAM :
-CIE disabled
-C6 disabled
-Cool n Quiet disabled
-Spred spectrum disabled
-Thermal Throttle disabled
-Turbo core disabled
-Vcore set to 1,4V @stock value
-CPU load line calibration :1/2
-Frequencie, multiplicator etc... set to their stock value
-Profil XMP 1,3 1600MHz loaded

What I find abnormal :
-Vcore is unstable, are variations from 1,35 to 1,38 normal ?
Yet my PSU and my MB aren't that bad. Although they aren't high end, they shouldn't be in pain @Stock frequencies et had no problem rising my Phenom II up to 4,2GHz.
-Why is there this periodic fall in Vcore, frequencie and temperature ? Is it a measure bug ?
-Why is the temperature so high whereas my cooler et my case are far from being so bad ?
Room temperature is around 20-21°C. NH U12P is mounted with its two fan with maximum speed, two big 140mm and a 120 expel air from the top and three breathe from the bottom. I use the Zalman ZM-STG1 as thermal grease (maybe not the best but easy two apply).
Also, expeled air from the top is almost cold ! Way cooler than the one I had when my Phenom II was running @4GHz. All this makes my think that the temperatures reported by OCCT, although plausible, are incoherent.
-Beside, since I changed my processor, I don't have the same probes in OCCT anymore : I now don't have the cores' temperatures anymore but I managed to get them back using AMD Overdrive. Still they are quite odd : One is blocked around 85, an other at 125 and the last ones at 29. Also, new probes have appeared : Package, Air Flow, Assembly, and TMPIN0. Is there one that might be important ?

Where could the problem come from ?
Is it possible that the cooler/the thermal paste are poorly mounted ? Still I didn't have the impression to do something different from usual and all the fans are running well.
Can the probes be poorly calibrated ? Can I recalibrate them ?
Is my UEFI well set for OC ?

Thanks for reading, I hope someone will be able to help me ;).
Waiting for an answer, Lucas.

PS : I tried my best to write proper English as it's not my primary langage, I hope it's understandable.
 
Solution
I don't think your "VDroop" is outside of a range that could be expected from a motherboard with 4+1 PWM. At 0.025v +/- seems about average to me, but I've never had the pleasure of an AsRock 970 AM3+ motherboard.

If I see board temps that are out-of-line I do a 'finger test' If I pull back a smoking nub, I know I have an issue. I don't really know the best utility for your motherboard, but HWMonitor (and maybe AOD) are likely as good as any.

Is there a reason you are running 16 threads across an 8-core processor? That may be contributing to some of your wacky numbers.

I don't think your "VDroop" is outside of a range that could be expected from a motherboard with 4+1 PWM. At 0.025v +/- seems about average to me, but I've never had the pleasure of an AsRock 970 AM3+ motherboard.

If I see board temps that are out-of-line I do a 'finger test' If I pull back a smoking nub, I know I have an issue. I don't really know the best utility for your motherboard, but HWMonitor (and maybe AOD) are likely as good as any.

Is there a reason you are running 16 threads across an 8-core processor? That may be contributing to some of your wacky numbers.

 
Solution

Lucasrio267

Honorable
Feb 1, 2014
8
0
10,510
Unfortunately, I don't think the boost brought by the OC would worth the money to buy a new MB.
About the 16 threads, I just ticked the box "auto" on OCCT, and the option was set this way. I'll try with 8 to see if it changes anything.
Thanks for your help.
 

Lucasrio267

Honorable
Feb 1, 2014
8
0
10,510
What do you call the multi- back .5 ?
Indeed, my cooler's fans are running under 1000RPMs and there aren't that much hot air going out of the case.
I tried to enable the Turbo UCC :

"Turbo UCC technology provides you an all-in-one solution to unlock the hidden cpu cores in the CPU, get the amazing system performance boost by auto-overclocking, and running the system in a certainly lower power consumption level as well. When you press "X" when switch on the computer, the system will automatically reboot itself, after that, the system will execute the Turbo UCC function. Enabling Turbo UCC, it will automatically help you to unlock the extra CPU core to enjoy an instant performance boost, and the system performance will boost up by overclocking CPU frequency, memory frequency and all related voltage settings! Different from normal high-energy consumption for overclocking or system upgrading process, ASRock Turbo UCC can allow an energy-saving overclocking or system upgrade computing operation!"
-http://www.asrock.com/mb/AMD/970%20Extreme4/index.asp

And I ran a stress-test :
FSB : http://puu.sh/6GaKH.png
Frequency : http://puu.sh/6GaKV.png
Temperature : http://puu.sh/6GaL8.png
Vcore : http://puu.sh/6GaLw.png
You can see that Turbo UCC increased FSB by 15MHz but decreased Vcore by 0.1V, and the strange falls disappeared ! Temperature has also gone down of 1°C...
Paradoxically, it put down the multiplier to x16.5 which means that the frequency only raised of 50MHz....
I went to the UEFI and checked what changes were made :
-FSB is set to 215MHz
-Multiplier is set to x16.5
-NB multiplier is set x16 -> 2150MHz (2000 previously)
-Vcore is set to 1.275V (1.4 previously !)
-NB voltage is set to 1.125V (I don't know the previous value)
-XMP 1.3 profile is still loaded but the RAM frequency is down to 1145MHz (1600 previously) !
-Every other voltages are set on auto
-AMD Application Power Management is disabled
-Cool n Quiet, C1E and Core C6 are enabled (I had them disabled previously)
-Every other settings including Turbo Core, Spread Spectrum, CPU load Line Calibration and Thermal Throttle are set on auto (They were disabled previously but I think auto mode disable them anyway)

So you think I could undervolt my CPU and decrease the temperature ? I'll try this tonight.
 
I was just noting that you may adjust the CPU multiplier in .5 increments if it helps to reduce temps and volts.

I think you should carefully continue tweaking and gaining experience with you rig -- AMD processors can take some volts BUT more volts = more heat and you have to find the sweet spot for your chip.

The auto-clock function dropped your memory divider to 533. That's why your RAMs speed has dropped (215MHz x 5.33 = 1146MHz).

If you have DDR3 1600 and want to OC using the system clock, you should drop the memory divider to 667 instead of 800. This creates a 'sweet spot' in the system clock at 240MHz (6.67 x 240MHz = 1600MHz) and your RAMs will be running spec speed.

The system clock is not only tied to the CPU frequency, it is also tied to your IMC/NB, HT Link, Turbo & RAMs, so you have to dial them back when OC'ing.

Set the IMC/NB multiplier so your speed is 2200MHz+. The HT Link at 2000MHz or so is fine.

With the clock at 240MHz and your CPU multiplier at 16.5, you should be running 3.96GHz across your cores (with Turbo disabled). You may then work the CPU multiplier up as temps and volts allow. Leaving C&Q on should have no effect on your OC.