Some Phenom II Overclocking

[Baulten]

Okay since I am wrong tell me why noone seems to be able to hit the 4ghz mark with the new phenom II's. My math isn't great but that would give you a 1ghz overclock and it seems that most are only able to go from 3 to somewhere around 3.7 stable. So let's see I can go from 2.5 to 3.21 stable on my Phenom 9850 so show me where I am wrong. So I say again my math isn't great but they both seem to be able to give you a .775 overclock stable. Things that make ya go hmmmmmmm.

You had great Phenom chip. Most don't OC past 3.0 GHz. Usually 9950s could hit around 3.1-2 on good board with good air, but that was really though.

 

Almost all Phenom II's can hit 3.6 stable on air, most higher. I'm working on joining the ranks of "most higher".

[Nuclear]

You had great Phenom chip. Most don't OC past 3.0 GHz. Usually 9950s could hit around 3.1-2 on good board with good air, but that was really though.

 

Almost all Phenom II's can hit 3.6 stable on air, most higher. I'm working on joining the ranks of "most higher".

 

I would like to see stable screenshots of a 9850 BE at 3.2+Ghz on air. The last round of 9850s were great overclocking chips with the 790/750 chipset...but going past 3GHz stable was still quite a chore. Show me screens of OCCT after a few hours or multiple continuous loops of 3DMark. My 940 will boot and run at 4GHz but it's only stable enough to eek through one complete 3DMark test. Once I put it under water it;ll sit at 4GHz 24/7.

 

Baulten: Have you adjusted any of your secondary voltages? What NB and HT multi are you running?

Edited February 24, 2009 by Nuclear

[Baulten]

They're all running at stock voltage. Shouldn't need any changes if I'm only multiplier overclocking, should they? I never had too boost any secondary voltage while clocking my Athlon x2 via reference bus. I'd have to go into BIOS to get the voltages.

 

In fact, that reminds me, I was unsure about some voltage settings. I'll go get them and report back.

 

Oh, I'm running 200x11 on northbridge (2200 Mhz) and 200x10 on northbridge (2000).

 

Edit: Here's my voltage settings:

 

CPU VID Special Add: +125.0 mV

SB PLL 1.2V Voltage: 1.20V

DRAM Voltage Control: 1.925V

NB Core Voltage: 1.250V

NB PCIE Voltage: 1.120V

NB HT Voltage: 1.270V

 

I always thought NB Core was the actual, physical northbridge voltage. However, over on XtremeSystems someone was saying it's the memory controller voltage. I'm stumped now

Edited February 24, 2009 by Baulten

[Nuclear]

I've found that little secondary voltage bumps are almost always necessary even multi overclocking only. I would bump your NB voltage up to 1.3-1.4v (1.4v only if you have good cooling) and bring the NB up to at ~2.4GHz. If you can get the NB running around 2.6GHz on 1.4v you'll notice a pretty big gain in memory bandwidth and performance. You can leave the HT running around 2.2GHz.

 

My 940 hit a wall of 3.7GHz using a slight vcore bump. It took me bumping the secondary voltages up a little at a time to get it stable at 3.8GHz and with more vcore I can run benchmarks at ~4Ghz. This is all with the NB at 2.6GHz and HT at 2.2GHz. The bandwidth I gained in bring the NB up was impressive.

[Baulten]

I've got a Spirit II Extreme chipset cooler on my Northbridge. Keeps it running at under 40C under load. I'll try pushing the voltage a bit; Is NB Core Voltage the correct one?

[Nuclear]

Yes, that's the correct one. You should only need to push the NB HT voltage if you are trying to run a higher HT and that will net you very little gain in a small scale. Up the NB Core and try ti get it around 2.5Ghz to 2.6Ghz. If you have Everest or SiSandra installed try running a memory bandwidth benchmark with your settings now and then after upping the NB, you should notice a pretty good increase.

[Baulten]

All right, thought so. I had just heard some people saying NB Core was memory controller, but thinking back, I'm pretty sure it was CPU NB core that they said. My bad. I'm going to give upping the NB voltage a shot. Will standard stress tests like Prime 95 and OCCT do a decent job testing NB stability?

[Nuclear]

All right, thought so. I had just heard some people saying NB Core was memory controller, but thinking back, I'm pretty sure it was CPU NB core that they said. My bad. I'm going to give upping the NB voltage a shot. Will standard stress tests like Prime 95 and OCCT do a decent job testing NB stability?

 

Yes. I always use multiple instances of Prime95 to validate my 24/7 clocks. Here's a quick rundown of my stability process:

 

1. Run SuperPI 1M: If it fails then add appropriate voltage to either CPU or NB and retry. May also need to loosen RAM timings

2. Run 1 loop of 3DMark06 (video card at stock): If system crashes slightly add voltage and retry.

3. Run 1HR of Prime95 (multiple instances) and keep an eye on temps.

4. Run 12HR Prime95 (multiple instances)

 

Hope it helps

[Baulten]

I've been doing 1 hour OCCT followed by 2 hour Prime95. Once I get a solid setting I'll do an overnight run of 8-12 hours.

 

Thanks for the help, I'll report back once I try some Northbridge adjustments.

[Baulten]

Well, even at 1.37V I could not get system stable with 2.4 GHz NB.

[Nuclear]

That stinks. Maybe you've just found the limit of your chip. I'm on my 7th 940BE right now and I've had some that did worse. Let me think a bit and I'll post back in a while.

[Baulten]

Yeah, I think I may be. Which sucks, but 4 cores running at 3.6 GHz for under $300 was certainly worth it in any case. I'm setting up an area to do cold air overclocking Wednesday or Thursday night so I'll see if that lets me push it any further.