FX-6300 OC strange temps

[Barthmord]

Hey Folks!

This is my first attempt at OC so bear with me.

System Specs:

http://pcpartpicker.com/p/nbzGHx for easy specs linking.

CPU - FX-6300
Cooler - Noctua NH-D14 using in box compound
Mobo - ASUS M5A99FX Pro R2.0
Case - Corsair 500R White
RAM - G.Skill Ripjaws X Series 8GB (2 x 4GB) DDR3-1866 Memory
GPU - Powercolor PCS+ R9 290
SSD - Samsung 840 Evo 250GB
HDD - WD 500GB
PSU - Corsair TX 650w
Disc - Asus BW-12B1ST/BLK/G/AS Blu-Ray/DVD/CD Writer


So, I started working on this yesterday afternoon and basically followed this guide for BIOS settings with the exception of the voltage numbers, multiplier, and I turned off CnQ mode:

Windows set to performance mode. Did not mess with LLC or FSB.

Order of ops for me(though I realize now i probably didn't need to bump voltage for lower Ghz tests):

All adjustments done through BIOS. Temps using Core Temp(high temp) and AMD Overdrive(margin).

1. Bumped up to 1.35v for cpu and 1.2v NB then set cpu to 4.0Ghz - Prime95 Small FFTs for 1 hour no errors. High temp - 44, Smallest Margin - 25.6
2. Stayed at 1.35v and 1.2v then set cpu to 4.2Ghz - Prime95 Small FFTs for 1 hour no errors. High temp - 46, Sm Margin - 24.88
3. Stayed at 1.35v and 1.2v then set cpu to 4.4Ghz - Prime95 Small FFTs for 1 hour no errors. High temp - 47, Sm Margin - 23.38
4. Stayed at 1.35v and 1.2v then set cpu to 4.6Ghz - Prime95 Small FFTs for 1 hour no errors. High temp - 49, Sm Margin - 21.38 (!?!?!) This is not possible right?
5. Thought i was a rockstar of OCing....played some Far Cry 4 for a few hours with no problems. Temps on second monitor didn't go anywhere near Prime95 numbers.
6. Set up a Prime95 Blend test to run overnight and went to sleep. Next morning found that rig had frozen not 10 mins into the test.
7. Restarted and ran Blend test again - fatal errors in 7 min.
8. Back to BIOS and slipped back to 4.4Ghz and ran Blend test(currently still running at 2+ hours) with no errors at all. High temp - 45, Sm Margin - 25.12
9. HWmonitor now says my highest Vcore is 1.53v and lowest is 0.588?? This program has been flaky for me so i haven't relied on it at all. Using CPU-Z for voltage checking and AMD OD/Core Temp for temps.


Questions:

Are those temps at all possible or is my monitoring software derpy? Ambient temp is around 20 C in the house.
Safe to bump voltage up a hair and get that 4.5Ghz?



I'm just confused based on other posts I've read.


Screen shot is after ~1h15m of Prime95 Blend test at 4.4Ghz from this morning.


Any and all input is welcome!

Thanks!

 

[Mandrake4565]

Barthmord welcome to the forum, first off don't use multiple temperature programs at one time they sometimes will cause mis readings. I am going to assume the package temperature reading is wrong because your cpu "socket" temp is at 52c. Usually when the cpu is too hot both of them will be alarmingly high. We usually use Hwmonitor to monitor temps and I haven't heard or seen any issues with it. It may just be conflicting with the others. All that said, seeing screen shots of the following tabs in Cpu-Z, Cpu, Spd and Memory will let us have a peek under the hood so to speak to see what is what. What I feel you should do is close down all the other software monitoring programs, open Hwmonitor, and run Prime "Blend" 20 minutes. If the cpu temp or package temp go above 70c or 62c respectively stop the test. Then post back with screen shots of Hwmonitor and the Cpu-Z screenies I asked.

 

[Alaric]

HWMonitor's Package Temp readings seem to be not as expected for more than one FX chips. Mine reads funny as well. I think Rgone mentioned it is actually thermal headroom , but I haven't found his post yet. Also , the FX are not designed to be accurate at idle or low loads. What it reads under Prime 95 should be the only (accurate) temp to worry about. I use AMD OD for monitoring temps when I'm playing with OC also. Maybe Rgone will take an interest in this thread or I'll be able to find the info.


My Package Temp is well below ambient.

 

[Barthmord]

Here are the requested screen shots.

I had read that HWmonitor was not always right with the FX chips and figured AMD OD would be able to properly read their own chips. Core temp was right in line with AMD OD so i figured it was also accurate. Though I'm just a beginner as all this.

 

[Alaric]

Found the part about inaccuracy when not under load here http://www.overclockers.com/forums/...curate-temps?highlight=HWMonitor+Package+Temp

And the rest http://www.overclockers.com/forums/...ive-is-crazy?highlight=HWMonitor+Package+Temp

Last , but not least , if you want to OC a Bulldozer or Piledriver , you should read these threads. A lot.
http://www.overclockers.com/forums/...ormance-Scaling-Charts-max-OCs)LN2-Results-co

http://www.overclockers.com/forums/...X-Bulldozer-Piledriver?highlight=package+temp

 

[RGone]

Now that we can see Min/Max of both cpu temps under load it is easy to see that at this time, there does not seem to be a problem with temps.
Current cpu/socket temp of 49c and showing a max of 50c and the corresponding core/package temp of 42c currently and having maxed at 43c is showing a very good 'delta' between socket and package temps.

Suggested Max Cpu Temp is ~72c and the Suggested Max core/package temp is 62c and currently under load of P95 Blend mode...you seem very well within any suggestions for temp. We have found that AMD is not in the HVAC business and has their cpus calibrated to be accurate at the higher temps in order to 'judge' how hot the cpu is in order to shut the cpu down and not die from extreme heat. Thru multiple posts and copies of emails from AMD representatives in forums, it was found that AMD is not the least bit interested in idle temps. Just the higher temps and thus the accuracy is not good on core temps until the cpu reaches about 40c. That number may vary slightly from cpu to cpu as does everything else tend to vary in the computer business. So we always look for temp readings under load of P95Blend to cut thru all the BS and get some temp numbers that will have a bearing on something real and necessary.

One other thought. The delta/difference between the core/package and socket/cpu temps if all is fine on an air-cooled rig is 'normally' about 5-6c difference, with the socket/cpu temp being that approximately 5-6c higher. As the overall temps go up and the HS on cpu begins to lose the heat fight, the two temps may go closer together. This coming together is if the case has very good air recycling going on and the VRMs are not underspec'd/undercooled, so that they may do their job without adding undue heat to the system inside the case.
RGone...ster.

 

[Barthmord]

Thank you.

So I'm good to keep pushing it to the 4.5-4.6 range on this set up? (sounds that way)

IE: Try 4.5 with Prime, probably have to bump voltage and always keep temps under 60-62. Then same thing with 4.6 until i run out of thermal margin.

I'll only use one temp program from now on.

 

[RGone]

IE: Try 4.5 with Prime, probably have to bump voltage and always keep temps under 60-62. Then same thing with 4.6 until i run out of thermal margin.

Sounds like a plan.
RGone...