• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Dolk's Research

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Dolk

I once overclocked an Intel
Joined
Mar 3, 2008
It has been 3 weeks since I got my 940 BE and I have been busy. Not only Have I been trying to get above and beyond 4.0ghz, I have been doing a lot research as to see how the 940 works with the NB and Ram and vise verse for them all. My research has been very complicated and all over the place. I wish to share what I have learned so far, hopefully this may be able to help you with your OCing.

Once I am done with my research I will have a nice large paper for everyone to read and learn from. Already its 30pages long and includes information about AMD, the CPU, and how to Overclock. With your help I can get this paper done, and get back to my normal life. [This CPU will be the death of me].

Please note that I will be editing this post as I update my research.


Info you should know before reading:

I believe that there is two (possibly more) Overclocking zones. The 30% zone and the 30%+ zone. These zones apply to all Phenom II CPUs. The 30% represents the speed of the CPU. I will use my 940BE for example. As I approach the limit of 30% the CPU becomes more stubborn with overclocking. In fact once the 30% mark has been reached things become extremely difficult.

64bit vs 32bit:
A 32bit OS is more stable then a 64bit OS when it comes to Overclocking. Not as much demand on the CPU means more room for your CPU to gallop. If you like 64bit, then you should be able to go as high as 3.8ghz with a good W/C system. Same rule applies to 32bit as well though. But if you ever want to have some fun and get into the 4.0ghz range, I suggest you boot up into a 32bit OS. I do not believe it matters if your OS is XP or Vista. Not to sure about W7. I use it for my games and run my 940BE at 3.7ghz @ 1.53v.

NB research:
Most of you know that I have spent a lot of time working with the NB, maybe to much time. I believe that there is some linear relationship with the NB and the CPU as far as how much you should increase the Frequency and the voltage as you in crease the speed of your CPU. So far this theory has been proven while in the 30% range. I have not yet been able to test the 30%+ range.
My theory so far with the NB involves one important equation. The 3.15:2 ratio. The ratio is the CPU Frequency to NB Frequency. For example a 4000MHz 940BE would require a NB Frequency of 2539.683. Depending on the RAM you are using the Voltage needed could be from 1.25v - 1.35v. While inside the 30% range this ratio does not apply to much. I have seen people use 2.8ghz NB while the CPU frequency was around 3.6ghz. I would like to test this theory in the 30%+ zone but so far have not been able to get anything higher then 4.0ghz.
My other research with the NB is high voltage with a low Frequency. It seems that if you increase the voltage of the NB you can create a more stable environment, along with lossening your RAM timings. So far this has been true, especially in a 64bit environment. I have been able to run a 3.8ghz stable in W7 @ 1.55v using 1.35v NB @2.2ghz and my RAM at 5-7-7-27.
My research for the NB has come nearly to its end. I will be running a bench mark test from 1.8ghz to 2.8ghz and compare the difference. I'll even see how it helps with higher CPU clocks and HT clocks and different timings with the RAM. Graphs should be up soon for everyone to see.

NB Conclusion based on Theory:
After a couple of weeks of testing and benching, I have come up with the data in the next post. Most of the data shows what it should show. A higher NB yeilds a a better result in your benchmarks. Which is true, but not in all cases. The data shows that in some cases the respected NB frequency was the better choice. In most cases the Respected NB Frequency to the CPU Frequency was the cut off before a change in scores accured. Sadly I don't have enough data to show a more variety of scores. I would love to have 4.0ghz and higher but I am unable to do this as of now. I will conclude in saying that it is no doubtly true that if you raise the NB you will yield better scores.

CPU Research:
The 30%+ zone is not fun to try to get into. I really need to find a stress ball when working with 4.0ghz.
As far as the CPU goes, its pretty messed up. The amount of voltage needed for a specific speed will depend on the factors of your NB voltage and ram timings. If you can even out the load everything becomes smoother (<-- that sentance gave me some ideas). I have not tryied out a variety of different FSB speeds, but will sometime soon. I've heard that if you hit a wall with the multi then just go to the FSB and work with it.
FSB:
I would like to know what the FSB does to the NB, HT, and RAM. Does it help with the spread, or does it just complicate things? Does it complicate to help? Is there a god? Sorry getting out of hand here...

RAM Research:
One of the strangest things I learned and tested was the losser timings in a 3:8 ratio. Apparently if you have losser timings it creates a more stable environment. How this effects benchmarks, I'm not entierly sure. It could help because you now have a faster CPU and NB, but I am lead to believe that it will have a slight decrease in performance.
Most of the time I have spent in the 3:8 ratio, not really working in the 1:2 ratio to much. I believe that the 1:2 ratio should only be used once you start OCing the FSB.

HT Research:
Once, when I was trying to cold boot into 4.0ghz, I tried upping the HT link from 1.8ghz to 2.2ghz and instantly I got a stable platform (using an FSB of 200). I'm pretty sure something is here, but it could be a minor thing. If you experaince an unstable system and you have already increased NBv and lossened your timings, try increasing the HT link. As I continue with my research I'll be trying this same thing out to see if it does anything.
The HT Also has a voltage and can be accessed in some motherboards' BIOS. Will increasing this have any effect with the stability of a CPU. Do you need to any higher then 2.2 ghz at any time?

ACC:
Maybe? Could it do something? IDK I've never tried it out. Any info on this feature would be great.
 
Last edited:
Graphs and Tables

NB:


FSB = 200. CPU Multi and Voltage were increased as need.
NB Voltage is a an absolute minimum, but could also be close to how much you need depending on what your OC setup is.
96452459gw9.png


Going on with above, I conducted tests to see if my theory was correct. Report is in my post above.
geekbenchev3.jpg

3dmark06kj0.jpg

pcmark05ui5.jpg

1msuperpivx5.jpg

32mwprimewv5.jpg
 
Last edited:
Dolk,

Once again you show off why i come to this site. Knowledge, Knowledge, Knowledge. THe graph on the second post is exactly what i've been looking for. Maybe this will solve my problems above 3.2ghz of instability. Time to play
 
This is great, Dolk! I'm looking forward to what you can discover and I will try and help out where I can.

Question: There are 2 NB voltage settings in the bios. One is Processor -NB Voltage and the other says Northbridge Voltage, which has two sub menus, Hyper-Transport Voltage and Core/PCI Voltage. Which ones are you adjusting?

Thanks,

R7 :beer:
 
Adjust the CPU->NB Voltage R7. Good question too. I've been meaning to look into those voltages as well.
 
As for as raising the Buss clock (FSB is a misnomer) I don't think it put ths hardware in jepordy of instability as much as what the BIOS does to RAM timing and latency. I've gotten around some issues by kicking some of the timings around and getting stable. Some BIOS's have been known to make RAM perform way above it's rated limit.

One of the issues I've noticed it the inability to raise the 4th timing above 26 which is the holdover time for refresh. Many sticks I have require it to be 32 or higher when clocking above 900 or so. More VDimm helps but causes excess heat and possible lockup on some Phenome.

Good Work :thup:
 
I'm going to work on my FSB research right now. I'll be updating my main post along the night.
 
I got a quick question. So I'm doing my work on the NB, and I'm stuck trying to get above 3.0ghz on the NB. After trying to get it up in AOD, my screen goes red and then the system blacks out, no BSOD. Here is what my settings were:
FSB 215
Multi x18
CPUv: 1.58
NB Multi: x14
NBv: 1.4v
Ram: 2.2v 1:2 Ratio @ 5-5-5-18
HT: x9

So what makes the NB unstable when it has enough voltage, and the CPU is working at a high speed?
 
Sounds like the ceiling hit with the K8 family at around 3G or a little above.
I hope one of the engineers with AMD sees this, I think you are sitting on the dead sea scrolls of Phenom.
Getting a NB to run at 4-6G would own the comp in a more games and some of the benches.
 
People have been able to go way above it. The guys that did 6.5ghz used a 3.9ghz NB. What allows them to do this (minus the LHE). I'm pretty sure that 3ghz is duable using a regular Passive cooler + fans setup.
 
People have been able to go way above it. The guys that did 6.5ghz used a 3.9ghz NB. What allows them to do this (minus the LHE). I'm pretty sure that 3ghz is duable using a regular Passive cooler + fans setup.

Same as getting the Brisbanes above 3.2G with hyper cooling I would venture to guess. What blows my mind is how the Cores and crank it up but the NB lags.
 
I found someone on Xtreme systems forums with the exact same setup as mine. I'm going to talk to him how he did everything. Doesn't seem like he got above 3.0ghz with his NB as well.

I can say this even at 2.8ghz the NB makes a huge difference in games :D I'll setup a benchmark system for the NB and do that next week. I'll let you guys know by Friday.
 
"FIELD TRIP" (to XS) :D

This just gets better every day. I'm just waiting for folding to finish up this morning then I'll pump this sucker up.
 
So I found out that my NB Voltage is not cooled by the huge passive cooler on my Mobo :bang head
Gay design by Asus, my own fault for not realizing this ahead of time. This could be why the NB hates going anywhere near 3.0ghz. This could also be why my board hates 4.0ghz.

The top one has a smudge on the side of it, as you can see. :bang head Anyone know where I can get a waterblock for my Mobo :(

img7672hf4.jpg
 
Last edited:
You talking about the YAGEO, it's a choke coil which means it was defecto or just could not handle the heat. I don't think we can sink them :fzzt:
 
Do you think I could get an RMA? Or should I be worried? There is one more (the one right below the one that could be broken), and it is fine.
 
As for now, my research and testing with the Phenom II has come to a close. I am now working on benchmarks and finishing up my paper. If I find anything new I'll make sure to share with you guys.

Thanks for the feedback.
 
Back