Sign in with
Sign up | Sign in
Your question

Almost there...I think. Phenom II x4 965BE OC

Last response: in Overclocking
Share
January 19, 2013 4:17:59 AM

Hi! Long time reader, first time poster. I built my first rig 2+ years ago and tried my luck at OC'ing for challenge. Not really a gamer, but more an enthusiast to see how much I can get out of the components. Needless to say my initial attempts did not go well, life got busy, had a baby and the rig has sat ignored since then. Which is why all the components are circa 2010-11. Recently got the bug again and fired it up(after a thorough cleaning!) Upgraded to the H100i and viola! it is so much easier with a good cooler.

So, onto my question:

I have the system 'stable' for 45 min to an hour on Prime95 then crash. It will pass all the benchmarks I have, including 3DMark11, 3DMark Vantage, AIDA64 stress, Cinebench 11.5, Memtest, SuperPi 32M without any errors, but can't get stable for Prime95 past the 45 minute mark.


Here is some initial info:
CPU Mult 20x
FSB: 200
DRAM 1600
Timings: 8/8/8/24/40/1T
CPU/NB 2600
HT 2000

CPU LLC: Full
CPU/NB LLC: Full
CPU volts: 1.4875
CPU/NB volts: 1.3000
CPU VDDA: Auto
DRAM volts: 1.725
HT volts: 1.3000
NB volts: 1.2625

CPU spread spectrum: disable
Cool and Quiet: disable

Temps under load:
CPU/cores - 46-48
NB - 58-60
SB - 50

What other information can I give you guys/gals? And can anyone tell me how to screen cap in BIOS?? Thanks

A little more info:
The BSOD info says 1033, and once Prime95 dropped a worker with a rounding error, then the other 3 continued to work for 3+ hours until I got home and stopped them.

Any advice/assistance/suggestions would be greatly appreciated. My goal is at least 4GHz and then push for more. Heat doesn't seem to be an issue with the new cooler. I didn't realize how terrible my Xigmatek Balder was!

Thanks for helping yet another newbie!

More about : phenom 965be

January 19, 2013 4:24:15 AM

Here is the rig info, not sure if it posted from my profile. Sorry for being so new...

CPU: AMD Phenom II X4 965BE@4.0
Cooler: Corsair H100i
Mobo: ASUS Crosshair IV Formula
RAM: OCZ3BE1600C8LV4GK
Video: XFX Radeon 5850x2 Crossfire OC'd
HDD: OCZ Vertex II SSD 60GB
HDD: WD Caviar Black 1T
Case: CM HAF 922
PS: Corsair HX850
OS: Windows 7 Ultimate 64

a c 103 à CPUs
a c 190 K Overclocking
January 19, 2013 12:20:22 PM

First thoughts,
The NB doesn't necessarily need to be 2600, and I like to keep HT and NB at the same speed personally
Loosen up the Ram timings, 2T on command rate and put the voltage for it back to stock, 1.725v is too high for 1600 Ram
disable LLC if you can
Don't forget to read through the Black edition Oc guide up top and the AMD Overclocking club as well for tips
Moto
Related resources
January 19, 2013 1:27:06 PM

Thanks for the ideas! Will do some more homework, and give the suggestions a go today as well.

Aaron
a b à CPUs
a b K Overclocking
January 19, 2013 2:13:16 PM

Yeah, voltage to memory is too high. CPU too. I can run my 965BE at 4.0ghz with stock voltage or a small bump in voltage. Here is my 2C...run everything stock except the multiplier to take you to 4.0ghz. Benchtest. If it fails, bump voltage. Bench. Fail? Bump. You should find that you won't need many changes to get there. Once you have a solid cpu OC, then you can start tweaking other things.
a c 103 à CPUs
a c 190 K Overclocking
January 19, 2013 6:50:16 PM

** CPU too. I can run my 965BE at 4.0ghz with stock voltage **
Maybe you can run yours with that but thats not to say everyone can,
chips are funny things where voltage is concerned, my old rigs X3 435 will hold 3.5GHz on 1.5v. my other halfs won't go over 3.2 no matter what I do
and my 975 (Same chip as your 965's) can validate at 4.6GHz whereas most Denebs won't even touch 4.4GHz, sometimes its the luck of the drawer over skill
(Thankfully for me :p )
Moto
a b à CPUs
a b K Overclocking
January 19, 2013 8:18:47 PM

I always had trouble with stability with high multiplier try upping the fsb to 240 and bring the multiplier to 17, any benchmark testing I did as well showed better performance for gaming with higher fsb although i know you said its not for gaming.
+1 to the memory voltage too high as well
January 20, 2013 1:08:28 AM

Thanks guys! Sorry the day got busy and haven't had much time to spend on the system. I did reset everything and starting over. 3.8GHz stable with V core 1.4125. DRAM back to 1.65. Hopefully more tonight...

I do have another question.....surprise, surprise.

The measured voltages on CPUZ and AIDA64 are significantly higher then the setting in bios when I am under load. Now, I do understand the Vdroop concept as it applies to generators and house curves and load sharing. I am assuming that this is similar, but is it that the Mobo is trying to do me a favor and increase the voltages for stability? Example: Vcore set manual in bios - 1.40000. When loaded it measured 1.4100 to 1.4180 on the monitoring software. This is true of all the other voltages I had set manually before. Now, that said, the ones that are in 'Auto' I would expect the system to have control and make changes appropriate for stability. In short, I am wondering why a manually set voltage is increasing under load.

Now..... LLC...... It seems to be all about opinion. ON/OFF/50%...etc. and that is fine, seems to be success stories in all categories. I understand that it is there to correct for Vdroop with loading, hence, LLC. On this board, I am seeing increases in voltage under load with or without the LLC enabled.

This trend for voltage increasing with load seems to be a function of the Mobo... right? Because loading a PSU drops voltage. Unless, there is some compensating circuitry/logic measuring load and adjusting source voltage. Is there any way to control this with the motherboard. I am not using anything set to offset. BIOS is all manual. But manual doesn't really seem like manual. If I set it to 1.40000 manually then load it 100% it should either be lower, or the compensation should sense the load and correct for the Vdroop to bring the Vcore back to the manual setting of 1.40000. Closed loop feedback system to achieve and maintain a setpoint, like a voltage regulator.

Is there anything else I should change? or just know that when I set a voltage in BIOS in MANUAL under load it will be 0.01 - 0.02V higher


Thanks again.

Aaron

January 20, 2013 2:18:19 AM

Please disregard the portion of my last post stating that LLC on or off is increasing voltage under load. I had an error in my head and was correlating 'Auto' with 'off' for some unknown reason. LLC in Auto and 100% are raising voltage. LLC set to 0% and 50% are actually not maintaining voltage and crashing almost instantly under load.

So I guess the question should have been "Is it a correct statement that with LLC set to 100%, the increase in voltage I am seeing is due to an over correction in the load calibration?"

Is there any chance that this is a PSU issue...I hope not, really thought the Corsair 850W should be enough? Maybe I will take out a graphics card? But with Prime 95 they are just idling anyways??


Anyways, just stepped up the CPU voltage crash by crash and it is back to 1.4875 running Prime95 now. Is it possible that my chip is going to need 1.5ish to be stable at 4GHz?

Sorry again for my error regarding the LLC settings.
a c 103 à CPUs
a c 190 K Overclocking
January 20, 2013 9:38:20 AM

I'm not sure why it overcompensates tbh, but I can say that its a common enough problem, probably the reason LLC and similar were developed, I know mine and many others have voltage set to x.xx but on screen various measures will report it as x.yx or x.yy which depending on set ups can be .001v over or much more,
I have my 4.6GHz overclock set at 1.55v (975BE, basically same chip as You) and it will often report as almost 1.58v using Cpu-z etc
It may be that your chip wants 1.5v for 4GHz but with the H100i clean and working well you should have enough cooling to handle the heat,
If you can I'd suggest full pushpull on the radiator, give it the best cooling you can
Moto
January 20, 2013 9:02:22 PM

Thanks Moto and others for the suggestions/advice/info. So I started over, re-read the AMD OC guide with multipliers and here is where I am so far....

BIOS: Vcpu 1.40000, DRAM 1.65, all other voltages auto

First tested memory: 9/9/9/24/40/2T@1600
- No errors with SuperPi 32M and Memtest for an hour

3.6GHz
Re-tested memory with SuperPi 32M no errors
Vcore load 1.368
Cinebench 11.5 - 4.05
3DMark11 - P5982

3.7GHz
Re-tested memory with SuperPi 32M no errors
Vcore load 1.368
Cinebench 11.5 - 4.12
3DMark11 - P6022

3.8GHz
Re-tested memory with SuperPi 32M no errors
Vcore load 1.368 -
Crashed during Cinebench

BIOS Change: LLC from Auto to 100% CPU and CPU/NB
Vcore idle - 1.402
Vcore load - 1.422
Cinebench 11.5 - 4.12
3DMark11 - P6032

3.9GHz
Re-tested memory with SuperPi 32M no errors
Vcore load 1.422
Cinebench 11.5 - 4.27
3DMark11 - P6090

4.0GHz
Re-tested memory with SuperPi 32M no errors
Vcore load 1.368
Crashed during Cinebench

BIOS change Vcore increased to 4.125
Cinebench crash
BIOS change Vcore increased to 4.21875

Completed Cinebench - 4.41
Freeze during 3DMark11(Physics)

BIOS change Vcore increase to 4.21875
Cinebench completed 4.55
Crash during 3DMark11(Physics)

BIOS change: Increased Vcore to 1.437500
Cinebench completed 4.53
Crashed with 3DMark11(Physics)

BIOS change: Vcore increased to 1.45000
Crashed 3DMark11(Physics)

**At this point it seemed unreasonable that the gain on 1GHz should require this much Vcore increase. Several of the forums have mentioned the need to increase CPU/NB at higher OC's, especially on AMD. So, I did.

BIOS change: CPU/NB volts changed from auto to 1.25000
Cinebench - 4.56
3DMark11 - P6177

Tried dropping Vcore back down after increasing the CPU/NB.
BIOS change: Decreased Vcore to 1.4375
Crash 3DMark11(Physics)

Increased back to previous/current settings:
Vcore: 1.45000
CPU/NB volts: 1.25000
DRAM: 1.65v
LLC: 100% CPU and CPU/NB
All other BIOS settings Auto/stock

Restested
Cinebench - 4.54
3DMark11 - P6195

Tried Prime95 blend - crash in under 60sec
Vcore load 1.46800
Noted mention of trying small FFT in OC guide
Prime95 Small FFT runs for an hour then crashes.

BIOS change decreased DRAM speed to 1333
Restarted Prime95 with small FFT - has been running 30min.
Vcore load 1.46800
Vcpu/nb load 1.283
Core temps at load 47C
NB temp at load 55C

Question: I am concerned that the stability issue is related to the RAM/NB? From what I have read the OCZ RAM I have is not the best... Several of the forums also mention needing to increase NB freq and possibly voltage to account for the increase in CPU processing speed to prevent data errata errors. Should I try that next?

This seems like a good guide: http://www.overclockers.com/the-importance-of-northbrid...

Sorry for the length of the post. Trying to provide all info I can.
Aaron

January 21, 2013 3:28:06 AM

Update:
So Prime95 Small FTT crashed again at the 100min mark with the RAM at 1333.

BIOS changes(not all at once, just shortening the post)
- NB to 2600
- Vcpu to 1.46250
- Vcpu/nb to 1.2625
- NB voltage 1.20000

Cinebench 4.67
3DMark11 P6290
Stable for AIDA stability test 1 hour

OCCT - crash
Prime95 - Blend crash

The more I look at the info, the more it seems this chip does not like 4.0GHz... And of course my lack of experience and knowledge aren't helping, but the amount of changes seem excessive to get this thing stable going from 3.9 at almost all stock settings to seemingly high voltages and still not stable....

So, to try to clarify things a little I am going to try to run a Prime95 blend stability test at 3.9 and see if it will hold. Initially I will use the current settings and if successful lower the voltages and the minimal values for stability at 3.9.

Mult 19.5
HT 200
NB 2600
HTT 200
Vcpu 1.46250
Vcpu.nb 1.26250
Vnb 1.20000
RAM 1333 9/9/9/24/40/2T
DRAM 1.65
LLC 100% CPU and CPU/NB
All other bios settings Auto

-Aaron



a c 103 à CPUs
a c 190 K Overclocking
January 21, 2013 7:40:35 AM

Glad to see you documenting your progress man, its a good habit to form :-)
Remember the deneb voltage limit is 1.55v so don't be shy about hitting 1.5v to stabilise that 4GHz,
Try dropping your ram to the tens, 10,10,10,30,40 2T I think it is, and try keeping the Nb and HT at the same speed, my chip seems to like 2400, any higher and I hit issues
I hear you on the massive volts for small clock jumps, but some chips need more V's to keep ticking than others, its finding what your own chip 'likes' that will see you through, keep going man, and check the AMD overclocking thread for more users findings on their chips, maybe someone in there has already posted the info you need :-)
Moto
January 21, 2013 3:24:53 PM

Thanks again Moto. I am trying to keep logical with the changes this go around, so thanks for the support. I do understand that for the final increases in Freq it may take some big steps and I have had it at 1.5 in the past, but it was when I was... let's just say, 'less organized' in my methodology for changes.

I guess just need some success at this point, so that was why i dropped back down to 3.9GHz. The run I mentioned in last post stayed Prime95 stable for 4+ hours and I stopped it. I will try to include a screen cap.

Mult 19.5
HT 200
NB 2600
HTT 200
Vcpu 1.46250
Vcpu.nb 1.26250
Vnb 1.20000
RAM 1333 9/9/9/24/40/2T
DRAM 1.65
LLC 100% CPU and CPU/NB
All other bios settings Auto

Going to proceed with dropping voltages back similar to when it passed the benchmarks when I was stepping up, and see the minimal volts it takes for stability at the 3.9GHz. Then some practice tweaking for performance gains on RAM, etc.. Then I will try yet again for the elusive 4.0GHz. Always reading, when time allows.

Aaron

Is there an instruction for attaching screen caps? Seems like have to host it somewhere and link to it? Sorry so new...
a c 103 à CPUs
a c 190 K Overclocking
January 21, 2013 5:41:46 PM

Most of us use Photobucket or Imageshack and paste the Direct link or Imgcode into the posts here,both are ree sites and Pb is quite easy to get used to, I haven't used I.s. though,
Good luck man, I'm sure you are going to be successful in this and its great to see your approach, many a future Googler will find this post interesting and useful
Moto
January 21, 2013 8:23:00 PM

Thanks again Moto for the support. I have learned so much from reading here I hope I can help someone sometime. Here is an attempt at the screen caps from the 3.9GHz run.











Still tweaking to find the lowest stable voltages. Will post when I find something.

Aaron

a b à CPUs
a b K Overclocking
January 22, 2013 1:22:37 AM

Looks like you're getting the hang of this OCing after all. Just remember that patience and cooling are your best friends when overclocking.

Like Moto was saying everyone's chips are different, my 955 required substantial voltage to run at 4.0Ghz and substantially more for 4.2Ghz (still stable) and just beyond recommended voltages for 4.3Ghz stable.

I can boot up and get into windows at 4.5Ghz but that requires well beyond recommended max voltages and I have yet to get more than 30mins into prime at 4.5Ghz.

Once you have a OC that is stable and you're happy with head over to the AMD overclocking club in my sig and post it up there. I look forward to seeing what you end up with. Good luck and happy OCing
January 23, 2013 1:46:26 AM

Thank you mlcaouette, for the tips and passing on the experience and the encouragement. I will definitely head over to your link and post the details when I am done with this one.
Update would be that I am currently fine tweaking the memory and seeing how tight I can get the timings and still stay stable. Just finished 2+ hours Prime with RAM@1600 9/9/9/24/40/2T. Now down to rated 8/8/8/24/40/2T still 1.65000v. SuperPi 32M is half done and no errors. Will update with final screen caps and numbers from this 3.9GHz OC.

-Aaron
a b à CPUs
a b K Overclocking
January 23, 2013 2:49:50 AM

I don't see why you would have any problem passing stability testing with stock settings on the ram. Keep us up to date on your success though.

In my opinion it won't be worth gong beyond stock speed/timings on the ram since the returns are going to be minimal at best and may adversely affect the 24/7 stability of your machine.
January 24, 2013 12:40:06 AM

So, all bench marks are passed with RAM at rated settings 1.65v, 1600 8/8/8/24/40/2T and Prime is running now. I think I am done with adjustments that I can think of..... Will post results when I have them for any further reccs from everyone on things I could tweak/tune to get any more improvement in performance. Thanks again to all for their contributions during this endeavor.

Aaron
January 24, 2013 1:12:07 PM

Here are my final settings and screen caps. Thanks to everyone who assisted me with this. I am happy to have a decent OC done and stable and plan on trying for more after reviewing everyone's input on these numbers/results. Thanks again!


Phenom II x4 965BE @3.9GHz
ASUS Crosshair IV Formula
Mult 19.5
CPU Bus 200
DRAM 1600 8/8/8/24/2T
HT 200
NB 2600
HTT 2400
Vcpu 1.45000
Vcpu/nb 1.25000
Vnb 1.20000
8/8/8/24/40/2T
DRAM 1.65
LLC 100% CPU and CPU/NB
All other bios settings Auto

I will post my results over in the AMD OC Club thread as well.




a c 103 à CPUs
a c 190 K Overclocking
January 24, 2013 5:29:53 PM

Glad you've hit stable man, although I know you'll be pushing for that extra .1 soon :) 
Moto
!