[Tig.]

I decided to try and unlock my cores on the Phenom II 550 BE today... Will 2 hours of stability testing will mean anything? Earlier, I did an hour and then stopped..

Also, I unlocked, so I don't know what the temperatures are, but I think the TMPIN1 is my core temperature. Is this is?

Is this stable?

 

[ChrisB17]

I'd go for 3 minimum on blend to test NB,CPU and RAM.

 

[Tig.]

Quote:

Originally Posted by ChrisB17 I'd go for 3 minimum on blend to test NB,CPU and RAM.

.. Dang I did the other one. So do I stop the test now and start the blend test?

Also thanks for the quick reply.

Edit: I changed it to blend testing. So far, its 2c lower than the first test I did.

 

[ChrisB17]

Just keep running the blend test for 3+ hours. Its testing a little bit of everything to make sure everything is running stable.

 

[raisethe3]

I know people say running it 24 hours will prove stability. For me, I usually test it 8 hours since I don't like leaving my computer on overnight. And another thing is, I don't use computer for more than 4 hours a day. Just not me.

 

[jimibgood]

3 hrs stresses NB to max at 512K(2-1/2 HRS)

 

[Tig.]

Umm, I wasn't here the whole day, but I left Prime95 on.

Almost 26 hours has gone without a problem on blend test with temperatures never going above 48c.

I guessing this is a stable unlock?

 

[awaizy]

Definitely. Start overclocking.

 

[Tig.]

Quote:

Originally Posted by awaizy Definitely. Start overclocking.

Lol. Easier said than done!

This will be my first time, so it's a bit scary..

What are the normal procedures I should do?

1. Go into Bios.
2. Change multiplier up in increments.
3. Restart and test for 10 minutes.
4. Repeat 2 and 3 until failure.
4. Get mad.
5. Go back into Bios and up the voltage.
6. See if it works.
7. Keep doing this until it goes to hot (70c right?) or it won't work anymore.
8. Come back to OCN and complain it doesn't work.

Is this correct?

 

[raxen]

Quote:

Originally Posted by Tig. Lol. Easier said than done! This will be my first time, so it's a bit scary.. What are the normal procedures I should do? 1. Go into Bios. 2. Change multiplier up in increments. 3. Restart and test for 10 minutes. 4. Repeat 2 and 3 until failure. 4. Get mad. 5. Go back into Bios and up the voltage. 6. See if it works. 7. Keep doing this until it goes to hot (70c right?) or it won't work anymore. 8. Come back to OCN and complain it doesn't work. Is this correct?

Heres what I do:
1. Go into BIOS.
2. Increase multiplier up by 1x up to 17x, then 0.5x afterwards.
3. Restart and run Linx @ max memory/20 passes.
3.5. If Fail, increase CPU voltage by one step (0.0125v) and try test again.
4. If OK, run Prime95 Blend test overnight (~10hrs minimum to cycle from 8k-4096k)
4.5. If Fail, increase CPU voltage by one step (0.0125v) and try test again.
5. Rinse and repeat over the entire week.
6. Record max multiplier @ max voltage.
7. Drop CPU multiplier back to stock (15.5x)
8. Increase CPU-NB Multiplier by 1x
9. Restart and run Prime95 custom test (FFT length 512k-1024k/10min) for an hour.
10. If Fail, increase CPU-Nb voltage by one step (0.0125v) and try again.
11. If OK, run Prime95 Blend test overnight.
12. Rinse and repeat for CPU-NB until you hit 13x, or 2600mhz.
13. Set max multiplier and max voltage for CPU and CPU-NB.
14. Run Linx max [email protected] passes.
15. If OK, run Prime95. If Fail, ragequit.

 

[Tig.]

Quote:

Originally Posted by raxen Heres what I do: 1. Go into BIOS. 2. Increase multiplier up by 1x up to 17x, then 0.5x afterwards. 3. Restart and run Linx @ max memory/20 passes. 3.5. If Fail, increase CPU voltage by one step (0.0125v) and try test again. 4. If OK, run Prime95 Blend test overnight (~10hrs minimum to cycle from 8k-4096k) 4.5. If Fail, increase CPU voltage by one step (0.0125v) and try test again. 5. Rinse and repeat over the entire week. 6. Record max multiplier @ max voltage. 7. Drop CPU multiplier back to stock (15.5x) 8. Increase CPU-NB Multiplier by 1x 9. Restart and run Prime95 custom test (FFT length 512k-1024k/10min) for an hour. 10. If Fail, increase CPU-Nb voltage by one step (0.0125v) and try again. 11. If OK, run Prime95 Blend test overnight. 12. Rinse and repeat for CPU-NB until you hit 13x, or 2600mhz. 13. Set max multiplier and max voltage for CPU and CPU-NB. 14. Run Linx max [email protected] passes. 15. If OK, run Prime95. If Fail, ragequit.

I like 15 better than all of them. Takes waaaay too long! Will overclocking the CPU-NB ( that's the north bridge right?) help a lot?

+rep btw.

 

[raxen]

Quote:

Originally Posted by Tig. I like 15 better than all of them. Takes waaaay too long! Will overclocking the CPU-NB ( that's the north bridge right?) help a lot? +rep btw.

I cheated a bit too when I started, jumping straight to 19x @ 1.5v and working my way downwards

However, that doesn't optimize all the parameters properly. Unnecessary voltage = unnecessary heat. Add into the mix of varying ram frequency, CPU-NB frequency, and all those voltages at the same time = impossible to identify your source of instability. Afterwards I learned my lesson and to go stepwise. Took me 2 months to iron out my final OC. Good thing my electricity bill is included in condo fees... haha...

Overclocking the CPU-NB does significantly increase your memory bandwidth and significantly affect superpi times. Whether you care about that is another story, but the effect is noticable jumping from 2000 to 2600mhz.REFERENCE

I guess a more important reason for increasing CPU-NB frequency is to maintain stability at higher CPU clocks > 3600mhz. IIRC the sweet spot for 3400mhz is around 2200mhz. 3800mhz is around 2400mhz. REFERENCE

 

[Tig.]

Okay so, after 10 hours of stressing, I think 3.4Ghz is stable with max temperature of 46c.

I think its stable.

Time to try 3.6 GHZ!!

You guys think I should change my timings from 9-9-9-24 to 8-8-8-21? The timings advertised of the RAM in my sig is 8-8-8-21, but I don't know why it is 9-9-9-24. So, do you guys think I should change that?

Also, do I just try to get the highest Ghz, then change the CPU-NB?
Can I change the CPU-NB to something like 2800Mhz or 3000Mhz without going in increments?

 

[Lord Xeb]

For me, if it is not at least 6 hours stable, then it is not stable

 

[almighty15]

Prime95 sucks, takes ages and even then it can fail on other programs.

Test using either Linx or Intel Burn test.

30 run's of one of those two with all ram tested takes half the the time and places the system under more load which is better

 

[Tig.]

Quote:

Originally Posted by almighty15 Prime95 sucks, takes ages and even then it can fail on other programs. Test using either Linx or Intel Burn test. 30 run's of one of those two with all ram tested takes half the the time and places the system under more load which is better

How do I test with all of the RAM? Do I just put it on the highest number possible?

Edit: Okay, so I just tried 3.8Ghz with 1.475V on Linx (Pic 1). It passed 10 runs without any errors and 53c was the highest temp it reached. But whenever I tried running Prime95, not even a minute passed, I get BSOD. Can anyone help me please?

Edit2: I just tried 3.6Ghz with 1.375V on Prime95. I don't think 10 minutes passed, and it BSOD. Help me please! First time overclocking and failure doesn't help.

This is hard, but I'm determined to get a minimum of 3.8Ghz with unlocked cores in less than 2 weeks!!

Edit3: Right now, I'm testing 3.6Ghz with 1.4V on Prime95 (Pic 2). Max temperature so far is 47c. 15 minutes has passed so far without BSOD. Wish me luck for the next 8 hours.

 

[raxen]

Quote:

Originally Posted by Tig. Okay so, after 10 hours of stressing, I think 3.4Ghz is stable with max temperature of 46c. I think its stable. Time to try 3.6 GHZ!! You guys think I should change my timings from 9-9-9-24 to 8-8-8-21? The timings advertised of the RAM in my sig is 8-8-8-21, but I don't know why it is 9-9-9-24. So, do you guys think I should change that? Also, do I just try to get the highest Ghz, then change the CPU-NB? Can I change the CPU-NB to something like 2800Mhz or 3000Mhz without going in increments?

Don't change your timings until you have verified everything else is stable. Change only one variable at one time.

I can guarantee you 2800/3000mhz will not work. Most chips will never reach that high without some really high volts. I suppose you can jump to 2600mhz and work your way down, but test this out at stock cpu multiplier.

Quote:

Originally Posted by almighty15 Prime95 sucks, takes ages and even then it can fail on other programs. Test using either Linx or Intel Burn test. 30 run's of one of those two with all ram tested takes half the the time and places the system under more load which is better

My personal experience says otherwise. Linx/IBT for 20 passes doesn't crash by computer, but Prime95 makes it BSOD 3-4 hours in while testing ffts between 512k-1024k. This means CPU-NB instability which Linx/IBT doesn't stress enough.

Quote:

Originally Posted by Tig. How do I test with all of the RAM? Do I just put it on the highest number possible? Edit: Okay, so I just tried 3.8Ghz with 1.475V on Linx (Pic 1). It passed 10 runs without any errors and 53c was the highest temp it reached. But whenever I tried running Prime95, not even a minute passed, I get BSOD. Can anyone help me please? Edit2: I just tried 3.6Ghz with 1.375V on Prime95. I don't think 10 minutes passed, and it BSOD. Help me please! First time overclocking and failure doesn't help. This is hard, but I'm determined to get a minimum of 3.8Ghz with unlocked cores in less than 2 weeks!! Edit3: Right now, I'm testing 3.6Ghz with 1.4V on Prime95 (Pic 2). Max temperature so far is 47c. 15 minutes has passed so far without BSOD. Wish me luck for the next 8 hours.

Like I said in my 15 point list, it is a really tedious process and should be done in a stepwise fashion. Too much voltage = too much heat = instability. You're at 53C with the socket temp probe, which means your core is at high 50s to low 60s. That's definitely enough heat to limit your OC potential.

 

[Tig.]

Quote:

Like I said in my 15 point list, it is a really tedious process and should be done in a stepwise fashion. Too much voltage = too much heat = instability. You're at 53C with the socket temp probe, which means your core is at high 50s to low 60s. That's definitely enough heat to limit your OC potential.

Ouch, that's bad... Is there any other way I can get to 3.8 without killing my CPU? I read that 60c should be the max temperature I should ever go with the Phenom II's. Really, 55c to me because it's just safer.

But good news!!!

3.6Ghz with 1.4V is stable for 6 hours on blend test!
...But, it ran at max of 48c. Maybe around 55c if it was correct, so I think I reached my limit on overclocking..
I can probably try 3.7Ghz, but I think I can't reach my goal of 3.8Ghz without upgrading anything else...

I haven't try CPU-NB yet, but if I do set it to.. say 2200, will I be able to lower the voltage?

Thanks a lot!
+rep

 

[raxen]

[email protected] is pretty darn good for an unlocked 550BE. As a reference, I've personally been stuck at 3.7ghz without going over 1.5v to reach 3.8ghz. IMO the extra 100mhz wasn't worth the extra voltage needed. I think 3.8ghz for you is a very realistic goal.

I'm offering the temperature as a possible explanation for your instabilities. The reason why I brought up temperatures is because Dolk always mentioned a "heat wall" for these chips. Essentially, no matter how much voltage you pump into the chip, you can never reach a certain frequency without bringing the chip down to temperatures lower than 40C.

Right now I would try to run 18.5x multiplier at 1.4v. Keep increasing your voltage until you get that stable. It is up to you whether you want to stay below 1.5v.

Afterwards, you can start to overclock your CPU-NB. Bring your chip back to stock voltage and multiplier, and increase CPU-NB multiplier and voltage. 2600mhz should be an attainable goal at 1.35v or less. As a reference, it takes my chip 1.31v to hit 2600.

Once you have max [email protected], combine the two overclock and stress again. If it passes, you can start to lower CPU Voltage to reduce heat.

Quote:

Originally Posted by Tig. Ouch, that's bad... Is there any other way I can get to 3.8 without killing my CPU? I read that 60c should be the max temperature I should ever go with the Phenom II's. Really, 55c to me because it's just safer. But good news!!! 3.6Ghz with 1.4V is stable for 6 hours on blend test! ...But, it ran at max of 48c. Maybe around 55c if it was correct, so I think I reached my limit on overclocking.. I can probably try 3.7Ghz, but I think I can't reach my goal of 3.8Ghz without upgrading anything else... I haven't try CPU-NB yet, but if I do set it to.. say 2200, will I be able to lower the voltage? Thanks a lot! +rep

 

[tK FuRY]

Anyone else realize this except me? His CPU temps aren't reading AT ALL?