OC Log: Phenom II X4 965 BE, XFX Radeon HD7850D 2GB, Kingston HyperX Black 16MB 2X8

[amp281]

Hey guys, I am working on overclocking this system (signature). I ran prime 95 stock for 2 hours to determine stability. I am a rookie overclocker, my last time overclocking was my 939 Clawhammer back in 2008. If what I says is incorrect or doesn't make sense call me out on it because I really don't have a clue what I am doing.

edit: Found this guide that has helped me immensely http://www.overclockers.com/step-gui...ck-amd-phenom/

Heres my baseline (stock settings, no overclock) 3dmark11 score

Originally Posted by 3DMark11 all stock: FSB 200mhz, Multi 17x, 3.4ghz

3DMark Score
P4895
Graphics Score
5176
Physics Score
4251
Combined Score
4152
3DMark Score
4895 3DMarks
Graphics Score
5176
Physics Score
4251
Combined Score
4152
Graphics Test 1
20.88 FPS
Graphics Test 2
24.09 FPS
Graphics Test 3
33.4 FPS
Graphics Test 4
17.13 FPS
Physics Test
13.5 FPS
Combined Test
19.32 FPS

I then ran 20x multiplier (stock 17x) and 4ghz (stock 3.4ghz)
Core Temp 56-58 degrees (rule of thumb is do not exceed 62)
Stock voltage

6 hours into Prime 95 Small FFT (CPU Intensive). System randomly rebooted after that, not sure why as I was asleep.

I then ran a multiplier of 17.5x and fsb of 220 3.85ghz for 5 hours no issues small FFT.

I made some adjustments. I lowered the multiplier back to 17x and raised the front side bus from 200 mhz to 235mhz giving me a clock speed of 3.995ghz. Ram speed is overclocked by this to 470 (1:2 ratio,2x) so as I understand it my stock ram speed is 1332, my current ram speed is 1880mhz (470*2*2).

I ran memtest86 4 instances each with 4 gigs of ram (quad core *4 = 16gb total ram) for 20 mins with no errors shown. I guess the ram is stable. Its really overclocked because the timing is now 6-6-6-15-20 2T, 470mhz DRAM Frequency.

Now I am running the prime 95 blend torture test for an hour to see if it likes it. If so I'll try bumping the fsb again and see if it can go to 4.1ghz.

Interestingly enough it runs cooler with fsb bump rather than multiplier bump, I am at 55-56* C core (Down from 58* C). Not sure why, maybe thermal paste maturity, ambient room temp? I tried voltage adjustments but got worse performance with voltage +, I dont see a reason to volt down if the temp is still 6* under the recommended maximum.

[amp281]

Also, I should note that disabling thermal throttling seems to have resulted in more stable voltage under load and a cooler processor.

When I am finished experimenting with my overclock and decide to run it permanently, should I re-enable thermal throttling?

I have left cool and quiet enabled.

[Durvelle27]

is that score with the oc if so it seems kinda low as it should score higher than my FX 4100



http://www.3dmark.com/3dm11/4472200

[amp281]

Nope that score is the baseline, no overclock. Just put things in the box and fired it up

edit: still stable at 235 mhz front side bus, stock everything else (volts, multiplier) 4 hours in torture test. Probably bump the FSB to 240.

Here is updated score with the 235 mhz FSB

Originally Posted by 3Dmark11 with FSB of 235mhz, Multi 17x, 3.995ghz

3DMark Score
P5019
Graphics Score
5188
Physics Score
4639
Combined Score
4482
3DMark Score
5019 3DMarks
Graphics Score
5188
Physics Score
4639
Combined Score
4482
Graphics Test 1
20.94 FPS
Graphics Test 2
24.1 FPS
Graphics Test 3
33.53 FPS
Graphics Test 4
17.17 FPS
Physics Test
14.73 FPS
Combined Test
20.85 FPS

[amp281]

Originally Posted by Durvelle27

is that score with the oc if so it seems kinda low as it should score higher than my FX 4100



http://www.3dmark.com/3dm11/4472200

I compared our scores, I think you have me on the GPU, I haven't touched my GPU yet. Physics is about the same.

Okay going to try higher front side bus and see if stable.

[Durvelle27]

Originally Posted by amp281

I compared our scores, I think you have me on the GPU, I haven't touched my GPU yet. Physics is about the same.

Okay going to try higher front side bus and see if stable.

yea i had my card clocked pretty high

[3870x2]

You could probably run the tests for a shorter length (10-20 mins) until you find your theoretical maximum clock, then you would do the longer tests to determine long-term stability.

I guess if you have enough time you can run the tests for 4 hours though.

[amp281]

Originally Posted by 3870x2

You could probably run the tests for a shorter length (10-20 mins) until you find your theoretical maximum clock, then you would do the longer tests to determine long-term stability.

I guess if you have enough time you can run the tests for 4 hours though.

I tried running 245FSB but it failed after 2 minutes. Upping the voltage by .125 got it another 8 minutes of prime.

I am now running 238FSB which gets me just over 4ghz (4.046) with OCCT as I heard it is much faster at detecting problems than prime 95. Running stock voltage again.

Do I need to be concerned with how my ram is being overclocked?

[Durvelle27]

drop the FSB to 210 and raise the mutlplier if your ram is rated 1600MHz

[amp281]

Originally Posted by Durvelle27

drop the FSB to 210 and raise the mutlplier if your ram is rated 1600MHz

Do you feel that my Ram is unsafe at my current fsb speed? I don't know a whole lot about ram besides lower timings is faster and so forth.

[Durvelle27]

Most ppl who know there gonna oc usually get ram that has a High frequency like 1866MHz to give them headroom when FSB ocing as that raises your ram speed and if your raise it past what its rated most ram will no run right and cause errors.

what is your ram speed

[amp281]

Originally Posted by Durvelle27

Most ppl who know there gonna oc usually get ram that has a High frequency like 1866MHz to give them headroom when FSB ocing as that raises your ram speed and if your raise it past what its rated most ram will no run right and cause errors.

what is your ram speed

The stock ram is Kingston HyperX Black 16GB (2 x 8GB) 240-Pin DDR3 SDRAM DDR3 1600

[eidairaman1]

Please See the Phenom 2 OverClockers Club on the forums http://www.techpowerup.com/forums/sh...ad.php?t=79551

[amp281]

Okay guys I figured out the ram thing.



To get the ram to its rated 1600 mhz I had to bump the fsb up to 240mhz. Then the multiplier has to be reduced to keep it below 4ghz. I reduced the core voltage to keep temperatures low.

[amp281]

Overclocked my XFX 7850 2GB to max in catalyst control center. I tried Sapphire Trixx but didn't see myself going beyond ATI's recommended max in CCC so didn't see a reason to use Trixx. Stability tested using MSI Kombustor.

P5672!!!!

http://www.3dmark.com/3dm11/5271882

Originally Posted by 3DMark11: CPU 240x16.5, 1.4v. GPU 1050 core 1450 mem)

3DMark Score
P5672
Graphics Score
6197
Physics Score
4566
Combined Score
4460
3DMark Score
5672 3DMarks
Graphics Score
6197
Physics Score
4566
Combined Score
4460
Graphics Test 1
25.01 FPS
Graphics Test 2
28.81 FPS
Graphics Test 3
39.63 FPS
Graphics Test 4
20.61 FPS
Physics Test
14.5 FPS
Combined Test
20.75 FPS

[amp281]

Do I have a bad core or is there something else? Core 4 always fails first.
Everything else was stable for ~8 hours.
Core 4 failed after ~1 hour
Should I lower the multi or raise my voltage (from 1.4) or raise my NB voltage? Raise my ram voltage?
Was hoping to get 3.9996 ghz and have my ram run at the rated speed at 1600.
Specs are still same as post 14: http://i49.tinypic.com/23m8s29.jpg

fullsize:http://i.imgur.com/JP1O5.jpg

[Aquinus]

I would reduce the HT link speed and try to get it closer to 2Ghz as opposed to keeping it the same as the NB. You might save some heat and power and solidify your overclock without losing any performance.

If you haven't given the NB more voltage after upping it to 2400Mhz, I would try increasing that after dropping the HT speed.

It's also safe (if you're keeping the CPU cool,) to increase the core voltage. I used to run my Phenom II 940 at 1.450v for 3.7Ghz. I suspect your 965 can do better.

Also try a different stress test. Linpack-based benchmarks tend to give a good indication if your system is stable or not. There have been instance where P95 has been stable but OCCT/IntelBurnTest hasn't been. There have also been instances where OCCT/IntelBurnTest/P95 are stable, but games like SC2 are not.

[Tonim89]

This kind of stability test doesn't make much sense, since 99% of BSOD happens, at least with me, in lightly loaded applications like Web Browsers, email clients, or even with the system sitting at desktop.

I suggest you to leave the system running a loop of light CPU application and see what happens. Crysis 2 benchmark at max settings is a smart choice, since it represents a typical workload of a gaming PC.

[amp281]

2 hours in with the cpu northbridge voltage bumped from 1.0 to 1.125 and its still stable. If it makes it another 6 hours I'll say its good, if not I'll try bumping the ram voltage or low the ht link. Whats the recommended max NB voltage?

Yesterday I made it 2 hours with OCCT and no problems (and it didn't need the cpu nb volt+.125) so I'd like it to pass prime 95 torture test for 8 hours and then I'll christen this thing done!


edit: crashed right after I posted this! what are the odds LOL. It did a windows reboot crash not a BSOD or a core fail. My strategy now will be to bump the core volt a tad. Going to change the memory voltage from Auto to 1.5v to ensure its at its correct voltage. If still fails I'll try bumping that up 1 setting.

edit2: haven't tried prime yet but 1.4125 core volt 1.51 memory volt. Futuremark11 score 5700. Thats a bit higher than before so maybe more stable memory. I'll try prime now.
edit3: load temperature in prime is 50*C so no big deal with regard to the increased core volt bump.
edit4: prime95 torture is still going strong 4 hours. 2 hours to go for! I figure 6 hours is fine.

[amp281]

Okay I think I have a finalized setup:


Its a little slower than other profiles I've tried but its very stable. I even accidentally ran futuremark during a prime 95 10 hour session and it didn't care.

Ram seems stable with the 1333 frequency and 9-9-9-24 1T timing.

I'm happy with it now. Only thing I am wondering is why my futuremark is so low. I was willing to take a physics hit to have a more stable cpu so the score is lower than previous scores I have gotten. My ram is running slower but more stable, this could also do it. I think that my CPU is near its stable maximum. The graphics card should be scoring much higher. My guess is maybe the CPU is bottle necking the GPU. GPU is overclocked with catalyst so no overvolt and the clocks are pretty conservative. Other thought might be the VECC (catalyst) power control setting. Going to try bumping this up and see if it improves the graphics score.

Originally Posted by 3DMark11, Stable p95 OC 965BE: 3.9ghz, 2400nb,200fsb, 16gb1333 9-9-9-24 ram, 54*c load. gpu 1050core 1450mem

3DMark Score
P5745
Graphics Score
6205
Physics Score
4766
Combined Score
4604
3DMark Score
5745 3DMarks
Graphics Score
6205
Physics Score
4766
Combined Score
4604
Graphics Test 1
24.99 FPS
Graphics Test 2
28.88 FPS
Graphics Test 3
39.73 FPS
Graphics Test 4
20.64 FPS
Physics Test
15.13 FPS
Combined Test
21.42 FPS

I was hoping to cross 6000.

Heres my SuperPi 32M

and my 1M



Seems about average for mild overclocked Phenom II.

The overclock itch got me again. Bumped the multiplier from 19.5 to 20x and the core voltage from 1.4125 to 1.425, giving me 4ghz

[ThE_MaD_ShOt]

Glad to see you found your stable spot. I was going to chime in earlier but I figured I'd wait a tad and let you learn about it and come to the conclusion that the memory was going to give you issues. On Phenom rigs you always want to keep the memory as close to 1333 as possible as that is what the on chip memory controller is rated for. And Phenoms don't respond well to an oc'd memory controller.

[amp281]

Originally Posted by ThE_MaD_ShOt

Glad to see you found your stable spot. I was going to chime in earlier but I figured I'd wait a tad and let you learn about it and come to the conclusion that the memory was going to give you issues. On Phenom rigs you always want to keep the memory as close to 1333 as possible as that is what the on chip memory controller is rated for. And Phenoms don't respond well to an oc'd memory controller.

I appreciate the learning experience. The other big discovery I've made is it doesn't like running over 1.4125v, the temps get too close to 55* C and it runs slower in benchmarks even if its at a higher clock!. I am going to attach 2 fans to my CMV8 Heatsink as I have heard that can drop it 3-5*C, that might get me another 200 mhz.

[drdeathx]

If you get better cooling you will be able to overclock much higher if your 965 is a newer batch. New batches can get up to 4.5GHz.

[amp281]

Originally Posted by drdeathx

If you get better cooling you will be able to overclock much higher if your 965 is a newer batch. New batches can get up to 4.5GHz.

Just bought 3 more 120mm fans, going to put 2 on the cooler master sides and 1 on the side of the case.

Edit:
Gave up on the master cooler fan mod, it looks like a pain in the ass to get it all apart and I don't feel like removing my heatsink from the board. Instead I added more fans to the case, one behind the harddrive cage in line with the one on the outside of the cage and the other on the side panel. Both are blowing in.

Net result was 5 degree cooler cpu core temp! Hooray that gives me the head room for another 200+ mhz .

[amp281]

Something keeps killing my computer at and over 4.1 ghz. I have tried 4.1 and 4.2 with the multiplier only (21 and 22, no FSB overclock) and 1.4250 and 1.4375 and 1.50 core volts. No luck. I have had two reboots (just shuts off and reboots) and one BSOD. Temps have not exceeded 50*C during the test, it probably would have reached 55*C but it just shuts down well before. Basically starts prime 95 and crashes 2 minutes later.

Any ideas? I also tried using the front side bus 205 to get 4.1 ghz and it still killed it. Voltage doesn't help. It just doesn't seem to like being over 4.0ghz but it has no problem running 4.0 ghz in prime 95 at 50 degrees C. WTF!?

edit: would disabling spread spectrum help?