Overclocking a Phenom, II 965 BE, How far can I go?

[Tech Tweaker]

Sounds like you got lucky and got a tougher one.

With my Phenom II 965 BE I hit instability around 55-60°C.

 

[poco242]

I will keep playing. I ran the passmark benchmark test on it. I used that because I could see what they say a stock CPU does with that test. They have it at 4290. Mine tested at just over 5100. According to their test it is equal to a stock 1065T. It may not mean much, but gives me somthing to go by.

 

[Mandrake4565]

You're probably better off running 2d benches off of Hwbot.org and then comparing yours to similar hardware. Here's some of mine with my 955be

 

[poco242]

OK. I did some more playing.

I did bump the CPUNB up to 2520. That was the closest choice to 2500. Everything was good. I then decided to bump the multiplier up to 20 for a 4.2ghz clock. I did not touch the VCore. It was at 1.4v. It ran prime 95 for about 10 min when it shut down. Temps were 53 at the time. No BSOD, just restarted on its own??

I now bumped the voltage up to 1.42. Running prime 95 now. Fingers crossed

 

[poco242]

This time it made it for 30 minutes before it shut down. This time I got a BSOD code 50 on start up. I am thinking this is caused buy the high clock rate I have on the RAM as was suggested, however, resetting the CPU back to 4ghz and 1.4 volts got it to start again. Will the ram frequency being to high show up more with a higher CPU clock?

 

[Mandrake4565]

No BSOD, just restarted on its own??

That happens, usually when you need a lot more Cpu V. You may be nearing the point where you're going to have to really pour the voltage to it to try and get it higher. You'll know you're close with the voltage, when you just lose a worker or two. Have you raised the Cpu NB Voltage at all?

Will the ram frequency being to high show up more with a higher CPU clock?

I'm not understanding the question.

 

[poco242]

The BSOD I got during start up after the last shutdown was a code 50. That has to do with RAM (from what I have read). It was suggested earlier that my RAM frequency was high. Would a higher CPU clock make that problem more apparent?

I need to get a second fan for my cooler before I give it any more voltage. I only have the one right now. I hit 58, but it did not shut down until well after that when the temp was back down to 55. I seem to get the highest temp on the second set of tests on prime95. It drops down 5 deg or so after that.

Forgot to say.I raised the CPUNB V from 1 to 1.025.

 

[poco242]

I had no luck last night with 4.2ghz, so I made a few changes. I dropped my memory down, dropped my HTT, raised the FSB, put the multiplier to 19.5 and raised the CoreV to 1.425. It held prime 95 for an hour, but my temps are high so I stopped after an hour. I will have the second fan for my 212+ cooler tomorrow so that should bring the temps back down I hope. If not, 4ghz will be my max.

 

[TransformedBG]

I wish i would have known then what i know now.. and got the 125w version...

 

[poco242]

Do you have the 140w version?

 

[TransformedBG]

Do you have the 140w version?

YES

 

[Tech Tweaker]

The BSOD I got during start up after the last shutdown was a code 50. That has to do with RAM (from what I have read). It was suggested earlier that my RAM frequency was high. Would a higher CPU clock make that problem more apparent?

Yes, on an AMD CPU it is possible that with a higher CPU clock frequency memory will become unstable sooner (and require setting memory to a lower clock frequency to remain stable).

That does sometimes happen, especially if the CPU is on the edge of what is considered "stable" and/or running especially hot under load. With all the strain on the processor the built-in memory controller sometimes just can't handle the additional strain of running high frequency memory (or in some cases memory with tighter timings).

 

[poco242]

OK. That is what I thought. I did drop the memory clock and it seems stable now. I need to install the second CPU cooler fan and one more case fan before I can give it a long Prime95 test. It was running around 58. It was working fine, but knowing what I know now, I do not want to run it that high for long. 30 min was as long as I was willing to try it for now. It ran P95 for 8 hours with no errors at 4ghz. I need more cooling before attempting that at 4.2ghz.

I must say that article by Dolk was VERY informative and explained things well. Still much to learn

 

[poco242]

I just got done with a little over 2 hours of Prime95 running at 4.2ghz. No errors. Max temp was 56. It mostly stayed around 51. I added a second fan to my cooler and a third to the case. I am calling this one good. I may do an 8 hour Prime 95 this weekend.

I am curious if it is normal for the FSB speed to fluctuate during Prime95? Mine did 5-10 mhz from time to time. That in turn would drop my CPU clock rate do 4.1 or so. A few times to 4.

Also my temps seemed to be the highest when it was running 8k - 14k FFT length tests. On the larger ones the tempos dropped 3-4 deg.


HWMonitor and CPU-Z. All BIOS settings unchanged.

 

[Mandrake4565]

Sure seems like a heck of a chip you have there poco, 4.2 on a Cm 212+ is not too shabby. What are your ambient temps, I'm curios?

I am curious if it is normal for the FSB speed to fluctuate during Prime95? Mine did 5-10 mhz from time to time. That in turn would drop my CPU clock rate do 4.1 or so. A few times to 4.

I've never seen it fluctuate 5-10 mhz but I have seen it fluctuate. I'm pretty sure it's just a software issue with Cpu-Z.

Also my temps seemed to be the highest when it was running 8k - 14k FFT length tests. On the larger ones the tempos dropped 3-4 deg.

If you're running Prime Blend some tests will stress the Cpu more then others. I have seen temps on certain test higher then others as well.

 

[poco242]

With the ghz cranked my idle temps are around 33. When I am not playing I load the stock profile and drop it back down to 3.4ghz and it runs around 27 idle, 41 on prime. I also think the fluctuation is in CPU-Z. In Core Temp it holds steady.

I just do this for fun. I have loved computers ever since High School when I took computer programming for 3 years.......on Apple IIe's. Just showed my age It is winter in Minnesota. Not much else to do after work.

 

[poco242]

Now I need some advise on how to go further. I just tried 4.3ghz. All I did is change the multiplier to 20. It just does not seem to like 20. It lasted about 10 seconds on prime and shut down. On startup I had a BSOD after that. I reset to stock and restarted. The BSOD code was 50 again. It started the first time fine.

I am guessing I need to up the Vcore. I will try 1.46. Should I also up anything else? Or maybe up the FSB freq and leave the multiplier at 19.5. It really seems to dislike 20

I will never run at these speeds. 4ghz is about it,but I want to see just how high I can go.

 

[poco242]

I increased the bus to 220mhz. I put the clock back to 19.5. Still 4.3ghz. It ran Prime for a few minute then shut down. This time BSOD code d1. That has to do with VTT from what I read. I then reset my memory to the next lowest clock since the new bus speed put it over 1600 again. Will try again.

Main question, will shutting down like this hurt anything? Just curious what the risk is outside the obvious. Temps or voltage too high. Neither are......yet.

Later-

Increased Vcore a tad, Ram voltage and NB. This time it went longer but shut down. BSOD code a. I read that means I need to increase QPI voltage. I think that is the HT, Correct?

 

[Johan45]

You should put up some CPU-z mem and spd shots. I have a feeling you NB might be too slow for that speed. Your voltage too maybe 4.2 @ 1.45v seems quite low to me.

 

[poco242]

I will do that this weekend. I know I am getting to the limit with air cooling. At least with this cooler.

This chip has been doing good with lower voltages. It ran Prime95 for over 2 hours at 4.2ghz at 1.44v with no errors. I am also going to run that a full 8 hours this weekend just to be sure.