Dolk's Guide to the Phenom II

WildWillie said:

I enjoy having a snappy computer. I hate having to wait for the dang things. I also enjoy playing some graphically intense games from time to time, but as I grow older there is not as much time for games as there used to be. What I do with my computer in my off time is donate computing time to the Folding@home project. I run four folding programs (one on each CPU) and one on the GPU (folds much faster than even four CPU's). A friend of mine has Alzheimers. It is too late for him unless a cure is found soon, but it would be great if a cure was found and others didn't have to go through it. So, I enjoy squeezing the most out of my computer to fold with a vengeance. I was hoping to get to 4.0Ghz (especially with the cooling system that I have), but I can't make it much past 3.6Ghz. I do realize the frame rate difference is minimal in any games (much better to work on the video card for that). So, next, I am buying the BFG GTX285. That should fold and play games just fine (almost twice as many SP's).

Click to expand...

yeah we all have different goals.

i am in to the gaming side but I dont like going over 1.50v's.
If you are comfortable with 1.50V's then keep going. i just have not seen any real facts that this is safe even if its cool?

I was not too happy with overclocking these phenom2's over 1.50 to get a decent overclock.

and yeah try NB speeds of 2 - 2.4Ghz with volts of 1.3 - 1.45

Latest attempt at 3.8Ghz

I had some time to try some things to attain stability at 3.8Ghz.
Things I tried:
Try #1: CPU-NB 1.8Ghz, CPU-NBv=Auto, CPUv=1.53V, HTT=1.8Ghz, crash
Try #2: CPU-NB 1.8Ghz, CPU-NBv=1.40V, CPUv=1.53V, HTT=1.8Ghz, crash
Try #3: CPU-NB 2.0Ghz, CPU-NBv=1.40V, CPUv=1.53V, HTT=1.8Ghz, crash
Try #4: CPU-NB 2.2Ghz, CPU-NBv=1.40V, CPUv=1.53V, HTT=1.8Ghz, crash
Try #5: CPU-NB 2.4Ghz, CPU-NBv=1.40V, CPUv=1.53V, HTT=1.8Ghz, crash
Try #6: CPU-NB 2.6Ghz, CPU-NBv=1.40V, CPUv=1.53V, HTT=1.8Ghz, ran for a while then crash.
Started to think regardless of CPU-NB frequency, stability could not be attained. Maybe I should have continued on with this progression and tried with the CPU-NB at 2.8Ghz, then 3.0Ghz, etc. Any thoughts?

Tried CPU voltage next:
Try#1: CPU-NB 1.8Ghz, CPU-NBv=Auto, CPUv=1.53V, HTT=1.8Ghz, crash
Try#1: CPU-NB 1.8Ghz, CPU-NBv=Auto, CPUv=1.55V, HTT=1.8Ghz, crash
Try#1: CPU-NB 1.8Ghz, CPU-NBv=Auto, CPUv=1.575V, HTT=1.8Ghz, crash
Try#1: CPU-NB 1.8Ghz, CPU-NBv=Auto, CPUv=1.600V, HTT=1.8Ghz, stable for a while. CPU-Z CPU volts read 1.50V, Asus Probe read 1.62 Volts, AMD Overdrive read 1.50 Volts. Temperatures were not bad, even with all four cores fully loaded (43 degrees C). Used new stability check program (CDEX). It is a CD ripping program that converts CDA files into MP3's. The program would crash within one minute if the system booted into Windows but was slightly unstable. Faster instability detection than Prime 95. I don't like running at 1.6V, even though the temperatures were not that bad.

I am still confused by the fact that I can run 100% stable at 3.6 Ghz (no crash for 1 week with all four cores fully loaded) at stock CPU Volts (1.35V), but I can't run with any voltage or CPU-NB frequency at 3.8Ghz for more than 1 minute. Something that I wanted to try but ran out of time for is to find the point of marginal stability for CPUv at 3.6Ghz. Maybe I'm right on the edge at 3.6Ghz and I really do need alot more CPUv at 3.8Ghz. If I am not right on the edge, then there is something else wrong. Dolk had said that he will be unavailable until thursday. I welcome any other thoughts.

You could try lowering your HTT down to 1600 and running the same set of NB tests again.

nzaneb, what did you have to do?

nzaneb, did you have to do anything other than mess with the multiplier or FSB to get 3.7 Ghz stable on your rig?
Thanks for the advice, I'll try it out when I get a chance.

WildWillie said:

nzaneb, did you have to do anything other than mess with the multiplier or FSB to get 3.7 Ghz stable on your rig?
Thanks for the advice, I'll try it out when I get a chance.

Click to expand...

That 3.7Ghz, is P95 small fft stable. It failed a blend test at about the 1 hour mark. I haven't gotten a chance to do much tweaking. My next step is going to be further analysis of the Memory settings, and the HT Link. I actually have it clocked back to 3.6 right now, until I can get back to it. My triple is needing A LOT more voltage than your quad, in order to get 3.6 stable. I'm at 1.5125V for 3.6, and at 3.7Ghz I was needing 1.5625. 3.6 is seeming like my sweet spot, but I still want to get 3.7 prime blend stable. I had to slow my memory down quite a bit for 3.7, but at 3.6 I can run full steam 1066 with 5-5-5-15 timings (maybe even tighter)
NB is at 2400mhz, with 1.4V for both 3.6 and 3.7. Further HT experimentation to come.

Try CPU-NB at 1.35 with the same tests. 1.4V may be too high and auto too low.

^^^+1. Try and keep your voltages as low as possible for stable clocks. The heat wall on the PII's is fast approaching.

Very low required voltage a4 3.6 Ghz

Last night I decided to try to see how close to the minimum required CPU volts I was at when overclocked at 3.6Ghz. I figured that this might justify the need for extreme voltage to run stable at 3.7 or 3.8Ghz. I ran 30 minutes of Prime 95 stable at 1.30V. Then I lowered to 1.25V and was still stable for 30 minutes. No boot at 1.20V. I know that a good test of stability takes more than 30 minutes, but thats all the time I had.

These results are telling me that I probably don't need much voltage at 3.7 or 3.8Ghz if I am able to run at 3.6Ghz at 1.25V (or less). So, I should be looking for another source of the instability when I am overclocked to 3.7Ghz or above. I like this idea because I don't like the idea of running at 1.600V. The highest CPU temperature that I saw was 43 degrees C. That didn't scare me, but I still didn't like running at 1.600V. If there was a cooling problem, I'm sure that the CPU temperature would have skyrocketed with 1.600V.

If you are doing that well at 3.6 I would personally stay there 24/7 and keep some nice cool temps.

ChanceCoats123 said:

If you are doing that well at 3.6 I would personally stay there 24/7 and keep some nice cool temps.

Click to expand...

I agree. Being stuck on mostly stock cooling atm, I am currently seeing how fast it can go undervolted. I'd be happy with [email protected], satisfied with 3.2!

That sounds good to me. I find my x2 fast at 3.125 although I would gladly bump it up to 3.3 or 3.25, it just takes way to much voltage to make me feel good running it almsot 24/7. I would siked with an undervolted 45nm quad at 3.2. Makin' me jealous.

Hey just a heads up AOD limits the voltage it can read. If I remember correctly you can change it in one of the .txt files. Do not use AOD for voltages as it will only show the maximum allowed by the software, not what is actually put into the BIOS. For example my AOD is limited to 1.45V in AOD while on the CPU im running 1.575V per bios set. Stable at 3.75V running P95 looking for options to get it stable at 3.8 but it takes way to much voltage for my Cooler Master V8

AOD has been known to do that. I like to simply use my bios. It really doesn't get any easier.

I hear AOD is nice but I don't even have it installed. I guess I am a traditionalist!

Change settings in bios, stress test, repeat. Once I find someplace I am comfortable speed heat & voltage wise, I stress test for 24h (prime in bground+normal computing+games&benchmarks).
I only just started flashing my bios from inside windows, awhile to go before I oc that way =)

I will probably install my other 2 sticks of dominator 1066 tonight, maybe I will get lucky and all 4 will work at that speed. If not, I will stick with 4gb@1066. I haven't filled it yet.

I have a question. Since I am multiplier OCing and not messing with the fsb, is there any performance reason to raise the HT multiplier(x9 default)? It is at 1.8ghz right now, is there a performance benefit to boosting it up to 2ghz or further?

Yes sir, it raises the speed of the on die memory controller so it should offer mild to wild benefits depending upon the speeds you run at.

So I'm sitting here thinking and thinking and I noticed something, there seems to be a similarity between your CPU top seed and the 720 Average top speed.

Also to note something. Your motherboard is designed around the AM3 platform minus the DDR3 architecture. I don't know why ASUS did this but they did. So basically you can call this an AM3 board. Which will help me look at this in a different way.

And no I didn't have a lot of time to write my report. Spring break was just way to relaxing

ChanceCoats123 said:

Yes sir, it raises the speed of the on die memory controller so it should offer mild to wild benefits depending upon the speeds you run at.

Click to expand...

Any idea what an average overclock of the HT bus is with these chips on 790gx/fx?

edit: Got my answer from Dolk's other threads actually ty

Malakai said:

I have a question. Since I am multiplier OCing and not messing with the fsb, is there any performance reason to raise the HT multiplier(x9 default)? It is at 1.8ghz right now, is there a performance benefit to boosting it up to 2ghz or further?

Click to expand...

It's not the HT Link you should be OC'ing but the NB. The HT Link has plenty of bandwidth for anything you want to throw at it.

It's the NB that increases the IMC, L3, and other data stream speeds ...

Mhm. There were couple of threads that showed the actual difference between nb speeds. It was pretty big in some cases.

It's so hot in my room right now I am hesitant to oc the nb, which has passive cooling only.