Fx 8320 OC

dreadloxx

Distinguished
Feb 1, 2009
154
0
18,680
Hello all. I have a quick question, I'm new to overclocking and I could use some sound advice. I have the fx 8320 in a gigabyte 990fxa-ud3 rev4.0 mobo with a cm 212 Evo, at stock clocks I never go beyond 35 deg. I want to oc to 4.2ghz so I have the multiplier set to x21. My question is about voltage, when it's at stock my voltage is 1.42xx in bios, if I set it at 4.2ghz with amd od my voltage stays at 1.32 with a few jumps and dips using hwmonitor and cpuz. What do you suggest a good starting place is to set the manual voltage and since I'm under what auto/stock is I'm actually better off even if I go up to 1.32~1.37 right? Please don't judge if it's a stupid question. Again this is my first oc experience and I don't want to fry anything or do something stupid
 
Solution
The automatic system voltages can't be used as a good reference, for several reasons.

One, they fluctuate aggressively, which means they go up and down rapidly rather than staying at a predetermined setting. This doesn't allow heat to build as much since the voltage drops before that happens.

Two, the sytem's automatic features generally over exaggerate the voltage requirements to err on the side of stability, since the system knows it is unlikely to stay there long or can be throttled if temperatures exceed specifications.

For a manual overclock you always want to use the minimum voltage that can be stabilized. In some cases, undervolting even, if the chip has especially good characteristics.

I'd disable cool n quiet in the bios...
Of course every chip is different. What one uses, another may not be stable at. What one is stable at, another may be able to be made stable with a lower voltage. My 8320 on a Sabertooth 990FX R2.0 sits at 1.35v and is rock solid at 4.5Ghz. Yours may be different. I'd suggest that may be a good starting place though. I would recommend you NOT use AMD OD, in fact, I'd completely remove it from the system. Some people like it, swear by it even, but I've seen nothing but problems using it on various systems and have seen it override bios settings even after having been fully removed or reset to stock settings.

Here's a couple of good guides to fill in some of the blanks and to outline the procedures you should use when overclocking.

http://www.overclock.net/t/91/ultimate-overclocking-guide

http://www.tomshardware.com/reviews/overclocking-guide-part-1,1379.html
 
And my settings, just for reference. Your chip may have different characteristics than mine, so make adjustments according to stability and thermal behavior, not based on what somebody else's chip was able to stabilize at. For 4.2Ghz though, I'd think that 1.35v or even less would be fine.

jfb52p.jpg



2hpon84.jpg



4j82eq.jpg



j7w4m0.jpg
 

dreadloxx

Distinguished
Feb 1, 2009
154
0
18,680
@darkbreeze
Yes I agree about amd od. The only reason I used it in the first place was to get an idea about how much voltage the chip was using with all the power saving features turned off. I ran prime95 at 1.30v and it froze so I bumped up to 1.32 and had the 8th core fail. It seemed to be stable in a 10 min test ~1.36xx. Temps stayed at 44c. Since I'm way under 1.42v I'm pretty safe from doing any harm right? I watch my Temps like a hawk and they never break 45c
 

dreadloxx

Distinguished
Feb 1, 2009
154
0
18,680
And thank you for the help. I'll delete od when I get home and tinker more with the voltage. I'll try at your voltage and cross my fingers. I have also read it's a good idea to run cinebench and real bench for a good gaming oc test instead of prime95 what do you think
 
What version of Prime are you using? You need to use version 26.6, and only version 26.6. Other versions will run AVX instructions on your CPU which creates unrealistic thermal conditions and is not the recommended method of testing thermal limit or stability. Both stability and thermal limit should be tested using Prime95 v26.6 on Small FFT. Do not use large FFT or blend. Small FFT uses a steady state load.


Prime95 version 26.6: http://windows-downloads-center.blogspot.com/2011/04/prime95-266.html


Run Prime for ten to fifteen minutes to verify thermal compliance. Run it for one hour after each multiplier/voltage change to test preliminary stability. Once your desired OC has been reached, run Prime 95 for 15-24 hours to determine "absolute" stability. Nothing is "absolute" but a 24hr run of Prime is a close as you're going to get.


P95 v26.6 works equally well across all platforms. Steady-state is the key. How can anyone extrapolate accurate Core temperatures from workloads that fluctuate like a bad day on the Stock Market?

I'm aware of 5 utilities with steady-state workloads. In order of load level they are:

(1) P95 v26.6 - Small FFT's
(2) HeavyLoad - Stress CPU
(3) FurMark - CPU Burner
(4) Intel Processor Diagnostic Tool - CPU Load
(5) AIDA64 - Tools - System Stability Test - Stress CPU

AIDA64's Stress CPU fails to load any overcloked / ovevolted CPU to get anywhere TDP, and is therefore useless, except for giving naive users a sense of false security because their temps are so low.

HeavyLoad is the closest alternative. Temps and watts are within 3% of Small FFT's.


-Computronix
 
The automatic system voltages can't be used as a good reference, for several reasons.

One, they fluctuate aggressively, which means they go up and down rapidly rather than staying at a predetermined setting. This doesn't allow heat to build as much since the voltage drops before that happens.

Two, the sytem's automatic features generally over exaggerate the voltage requirements to err on the side of stability, since the system knows it is unlikely to stay there long or can be throttled if temperatures exceed specifications.

For a manual overclock you always want to use the minimum voltage that can be stabilized. In some cases, undervolting even, if the chip has especially good characteristics.

I'd disable cool n quiet in the bios, but leave the control panel advanced power settings minimum processor power state setting at 5%, max at 100%. This will still allow some fluctuation of the multiplier and voltage, but not as aggressively as CnQ will. Alternatively, you can simply disable CnQ and set both the min and max processor power states in the advanced power settings to 100% which will effectively lock the CPU at one speed and voltage.

 
Solution

dreadloxx

Distinguished
Feb 1, 2009
154
0
18,680
I'm not sure what version I downloaded it last night, ive been using the small FFT. 24hrs! I didn't know so much went into overclocking! :p so if lets say 1.35v failed on the 19th hour a whole day has been wasted and the next bump is another day of testing. I can see why it's called "stress" testing
 
Ha ha! Yeah. That's NOT why, but it can be stressful to more than just the machine. I feel you on that.

Ten hours is probably fine to make a fairly positive assurance of stability, at least at first. At some point after verifying stability I'd do a 15-24hr run though, just to be safe. Preferably before any large downloads or file transfers are done, since unstable systems can and do introduce small errors into the data that can accumulate and cause irrevocable damage to your files over time on a system that hasn't passed stability testing of a stringent nature.


Better one day wasted then 6 months of data lost.

 

dreadloxx

Distinguished
Feb 1, 2009
154
0
18,680
So I downloaded your version of prime95 and found a "stable" voltage of 1.368, nothing if failing but my clock speeds will drop from 4219mhz to 27xxmhz on random cores Cpuz shows the voltage dropping to 1.2xx when this is happening any idea?

Edit: 1.3500v...cpuz was showing 1.368v for some reason
 
Even when stable, it may fluctuate somewhat. Try using HWinfo or CoreTemp for monitoring of thermals and specifically HWinfo for sensor data. Don't use HWmonitor or Open hardware monitor, I've found them to be inaccurate sometimes.

HWinfo (Run "Sensors only"): http://www.hwinfo.com/download.php

CoreTemp: http://www.alcpu.com/CoreTemp/


I tend to use CoreTemp for thermals, and HWinfo for everything else.

As I said before, if you go into Control Panel, Power options, Change plan settings, advanced settings, and change the Processor power management setting to 100% on both min and max, the multiplier and voltage will stay locked at your full setting as configured in the bios.

BUT, it's not totally necessary as anytime the processing power is needed, it will stay at it's max speed, and when it's not, it will drop down to save power and reduce thermals. Either way is fine. Also, make sure you disable Cool N Quiet in the bios. Later, if you wish, you can turn it back on.

Pay close attention to thermals using core temp when running stress tests. It even has an option to change the thermal reading from actual to thermal margin in the options. Thermal margin is what AMD OD uses and is supposed to be more accurate, but I don't know about that.

 

dreadloxx

Distinguished
Feb 1, 2009
154
0
18,680
I did what you said and changed the power management to high performance which helped a little. I would still freeze in p95so I started researching my board and found that the llc is touchy so I set it to medium which others had a good experience with and I also disabled the APM and bingo, I am stable at 4.2ghz @1.3500v. I may be able to go lower but as of right now I am 24hrs stable thank you very much!!! Also my Temps never got above 45c on the core and 50c on the socket
 
Did you also disable turbo core? Leaving turbo core enabled on a system that has a manual overclock is a bad idea. In fact, it's a very good way to CAUSE instability.


I'd also read this: http://www.tomshardware.com/forum/id-2345618/reaching-cpu-overclocking-stability.html

And I pretty much follow this:


Should I enable or disable Application Power Management and HPC?

As an overclocker myself, I would say No. Do not disable Application Power Management, and don't enable HPC Mode.

All disabling APM does really is cause your CPU to run outside the 125w TDP range. In essence, your drawing more power and voltage, and creating more heat for very little benefit. The same goes for HPC Mode. Unless your running some cluster server with very specific applications, enabling HPC mode is just going to generate more heat and power draw for very little benefit.

I can say 99% of CPU throttling problems on FX CPU's are due to either buggy BIOS in need of an update, or the more commonVRM throttling. Your motherboard will throttle the CPU if the VRM phases get to hot or outside of a safe zone coded in the BIOS. This is hard-coded in and set to help prevent frying your motherboard.

This is most common on the AMD 970 chipsets and the lower end boards that have 4+1 power phase designs. These are not high end power phase designs, and even an FX CPU at stock will throttle on these boards. Moving up to a 990FX board with a 6+2 or greater power phase design and good beefy heatsinks on the VRM and chipset will result in throttling problems going away without even having to change or mess with those BIOS settings.

In my experience, on an Asus M5A99FX Pro Rev 2.0 board, enabling and disabling APM and HPC Mode had no discernible effect whatsoever in any of the applications or games I run. I was still able to push 60 FPS solid in Crossfire on games such as Skyrim, Devil May Cry, Tomb Raider, and other titles with APM enabled and HPC Mode disabled. the only difference I saw between APM disable and HPC Enabled was higher core temps, high socket temps, and more power draw. The performance of the games and applications was identical.

The throttling many say these CPU's do, on the right motherboard with a 6+2 phase design or better, without buggy BIOS's are merely cosmetic. The FX 8350 dropping to 2.9 GHz for 1 millisecond or less before jumping back up to 4.2 GHz Turbo in game will not be noticeable whatsoever, as these drops are algorithm based to do so when the CPU has room to do so. There is more too these settings then just throttling, sometimes, it may just slightly drop Vcore while maintaining base clock rate to lower heat and keep the TDP profile, othertimes, it will boost voltage and multiplier to give you a boost.

The only time and situations I would recommend Disabling APM(Application Power Management), and enabling HPC Mode is if you have:

1. A very good preferably high end liquid cooling solution for your CPU for planned high overclocks in the 4.9 to 5GHz range that would go over the TDP limit anyways.

2. You have custom heat sinks and active fans on your motherboard's VRM, Northbridge, and other chips on your motherboard.

If your not overclocking that high and don't have the active cooling for your motherboard, I would err on the side of caution against it. If your having throttling issues, they are more then likely related to motherboard VRM throttling or buggy BIOS. In these cases, either invest in a better motherboard, or see if you can get your motherboard manufacturer to update the BIOS to fix the issue.
 

dreadloxx

Distinguished
Feb 1, 2009
154
0
18,680
OK with apm enabled I don't freeze but the frequency of every core just bounces between 4219mhz to 2913mhz and back. Constantly all cores are under 100% load and nothing is failing. My Northbridge is setting at 40c so I assume that means the vrms are cool. Not sure why it's doing this. With apm disabled the frequency is pegged at 4219mhz on all cores
 
Did you set the min and max processor power states to 100%. I don't think just setting it on "Performance" will do that. I think you need to manually set those advanced power options to 100% if you want to lock it in place, turn off CNQ, turn off turbo core, disable HPC mode (BIOS) and set APM to "Auto" rather than enabled. Then again, if it works for you with it disabled, and your temps stay within reason under load, it should be fine.