EP45-DS3P/e5200/DDR2-800: OC of FSB,CPU,MCH,ratio,timing confused... please help

[vacapp]

I think (hope!) I am close () to solve the situation.
I found a couple of new graphs, equations. Now, because my situation /200MHz stock FSB clock/ is something different, I am trying to find/calculate my crossover point, although I have still some problems to determine stock vMCH vs. possible max. (OC) FSB setting (calculation?) vs. lowest tRD relation, analog to this -> img. -> whole guide .
But in the meantime I share below useful links in addition to the previous mentioned (sticky gigabyte threads of tweaktown forum, anandtech->(1.), xbitlabs->(2.), TTR->(3.)):
CAS - How memory frequency affects latency - Tighter timings vs. Higher speeds? - The Tech Repository Forums
Optimum OC Performance Points - TTR's Guide to Determining Optimum Overclocked System Performance Points - The Tech Repository Forums
Lifetime vs. OC - AnandTech: Intel's 45nm Dual-Core E8500: The Best Just Got Better

I hope I can finish my research soon.

If someone has an idea, please let me know! :)

[Lsdmeasap]

Sounds like you are diving in way deeper then you need to be for sure. Do you have a stable system now that you can use?

I hope so, and you are just doing all this for learning purposes right?

[vacapp]

Originally Posted by Lsdmeasap

Sounds like you are diving in way deeper then you need to be for sure. Do you have a stable system now that you can use?

I hope so, and you are just doing all this for learning purposes right?

Thanks for asking.

I think I have, because its still running on stock.

The reason for my "deep" research is (and however I did not find all answers I need, it seems I can soon try some settings) that, I have not very much time to make endless "brute force" oc sessions to find the right settings. I mean, I want to figure it out what could work and what's sure it wouldn't. So I have less to guess. It seemed there are a lot of equations that make much more easier to find the best performance points in specific cases, but the problem is, nothing fits enough to my config and although some results are reasonable in the articles, I dont think I could use them the same way.

Anyway at the moment I would like to know something about followings as a conclusion what is still missing, so I can get closer to the end :

Why I am thinking even about this, although I know every result/setting depends on each part of a config?

During the course of our testing we made a rather interesting discovery regarding 45nm silicon scaling: a window exists in which CPU frequency responds in a highly proportional manner. Calculating this value later tells us that between 3.0GHz and 4.0GHz our processor requires ~0.3mV (0.0003V) more Vcore for each one megahertz increase in core frequency.

How could they calculate this for example?

Increasing FSB is the main question, is there not even a close "calculation/equation/approximation" how much voltage is it needed for MCH/NB and/or if necessary CPU termination/VTT voltage and/or if necessary CPU core/VCC?
Let's see some considerations:
1.) if I do not want to set a really low tRD, compared to the given MCH strap default (as shown in the previous mentioned reviews/guides), it is not necessary to increase the voltage
2.) but if I increase compared to a given stock FSB (in my case 200MHz) how much percentage increase is equal raising the MCH/NB core voltage (how much V+?)
3.) If the allowed native MCH FSB of the motherboard is above the level of the increase of point 2. does it matter (native capable 1333 vs. 1280 in my case)? Does it "lower" then the necessary increase of MCH/NB voltage?
4.) Is the CPU termination/VTT voltage also relevant? Is there any relation if I set it lower/higher?
How is the MCH/NB core voltage affected considering these aspects together?

I was studying the well known guide of Anandtech which you also suggested elsewhere in previous threads AnandTech: ASUS ROG Rampage Formula: Why we were wrong about the Intel X48 but not enough information regarding my special situation...

[Lsdmeasap]

1. A lower tRD may not always be possibly on a given strap, even with massive voltages, it depends on many things.

2. I do not understand what you are asking

3. See above, as I think this applies to the answer of # you asked so I cannot answer here either.

4. CPU Vtt is relevant to many things, it affects the following >>>

CPU Termination (Vtt) Automatically changes the following if manually set >>>

CPU Reference Voltage (.63% of Default Vtt/1.20) .76 ET
CPU Reference2 Voltage (.67% of Default Vtt/1.20) .80 ET
MCH Reference Voltage (.63% of Default Vtt/1.20) .76 ET

DRAM Voltage Automatically changes the following if manually set >>>

MCH/DRAM Reference Voltage (.50% of Default Dram Voltage)
DRAM Termination Voltage (.50% of Default Dram Voltage)
Channel A Reference Voltage (.50% of Default Dram Voltage)
Channel B Reference Voltage (.50% of Default Dram Voltage)

(Please see the link in my signature about MCH Ref, I updated it tonight. Some info you may find useful as well)

[vacapp]

Thanks for your efforts to bring more light into the confusion.

Originally Posted by Lsdmeasap

1. A lower tRD may not always be possibly on a given strap, even with massive voltages, it depends on many things.

This part is clear (I think), I only noticed it, after reading the Anandtech review. :)

2. I do not understand what you are asking
3. See above, as I think this applies to the answer of # you asked so I cannot answer here either.

Sorry I try to clearify. Because you are familiar with Anandtech's review, I take some examples.
If I would like to increase my stock FSB 200MHz for example to 320MHz, how much should I raise MCH core voltage (and assumed I do not want extremly low tRD only similar decrease like Anandtech)?
Why would I like to know it and not try again and again? Anandtech was able (on stock MCH voltage) to go upto 425MHz FSB from 333MHz stock and tRD=6 by use of 400MHz MCH strap!
Is it equal when I raise from 200 upto 255 only (same percentage ca.28%)? Is it the maximum on stock voltage? Or MCH core voltage depends more on the value of the FSB itself? 320MHz is much more lower than 425MHz (although the chipset X48 supports as a default 1600MHz FSB vs. P45/1333MHz). On the otherhand lower FSB means lower tRD without voltage increase and what more they overclocked the FSB upto 1700MHz, but in my case after increasing it is still lower than the default specification of the chipset 1280 vs 1333.
I hope it make sense now, what I try to implement for my case.

4. CPU Vtt is relevant to many things, it affects the following >>>

CPU Termination (Vtt) Automatically changes the following if manually set >>>

CPU Reference Voltage (.63% of Default Vtt/1.20) .76 ET
CPU Reference2 Voltage (.67% of Default Vtt/1.20) .80 ET
MCH Reference Voltage (.63% of Default Vtt/1.20) .76 ET

DRAM Voltage Automatically changes the following if manually set >>>

MCH/DRAM Reference Voltage (.50% of Default Dram Voltage)
DRAM Termination Voltage (.50% of Default Dram Voltage)
Channel A Reference Voltage (.50% of Default Dram Voltage)
Channel B Reference Voltage (.50% of Default Dram Voltage)

(Please see the link in my signature about MCH Ref, I updated it tonight. Some info you may find useful as well)

I definitely will check it, as your other stickys, but how is this in relation increasing my stock CPU FSB from 200MHz upto the achievable goal of as less OC on voltages as possible, but maximum performance?
In other words - I can only refer again to Anandtech review of CPU core voltage vs. CPU speed: 0,0003V/MHz -, how could I approximately "calculate" if I should increase it or not? How much? This is in relation with question 2. and 3. Anyway does it depend on them at all? Or when is it necessary to think about an increase?

I hope I could explain in detail what I am looking for.

[Lsdmeasap]

If I would like to increase my stock FSB 200MHz for example to 320MHz, how much should I raise MCH core voltage (and assumed I do not want extremly low tRD only similar decrease like Anandtech)?

Maybe none, maybe a little? This all depends on what ram you are using, that means what Quality, overall final speed of this ram, and all other settings used.

Why would I like to know it and not try again and again? Anandtech was able (on stock MCH voltage) to go upto 425MHz FSB from 333MHz stock and tRD=6 by use of 400MHz MCH strap!

You cannot base your attempts/future attempts solely on one review. And do keep in mind they were using a different board with a different chipset, and different ram as well. P45 is very ticky about this, and depending on the ram you use, and all other settings. I do not see anything under 7/8 working on 400 Strap in P45, and 7 is in extreme cases with other settings loose, or high voltages on certain things. 8-9 is the normal Amount I see and have been able to easily use on 400 Strap with various quality rams.

Now the MCH voltage amount for that 333 to 425 FSB Jump all depends again on the ram used, the amount of the ram, the overall speed of the ram, the quality, and all other settings.

Is it equal when I raise from 200 upto 255 only (same percentage ca.28%)? Is it the maximum on stock voltage? Or MCH core voltage depends more on the value of the FSB itself?

Kind of yes, but the amount of this increase without extra voltages is up to many things not limited to but including your board, CPU, ram and ram amount and speed, ram quality and again all other settings used.

MCH Core voltage needed depends on the amount of strain you put on your NB, that could be with tight timings, or low tRD, large amounts of ram, very high speed on the ram, or very high FSB.

Vtt amount is also a factor in how much MCH may or may not be needed, which also controls MCH Ref voltage, which may also factor into things depending on what you are running.

You do not need equations and such to overclock and know when you need more volts. Sure you could spend days trying to figure that out on paper, but I bet you less time would be spent by actually physically doing the overclock.

All you need to do is push the FSB 5-10 at a time, when it fails raise Vcore a bit and go at it some more. Or you can just plug in a higher decided voltage you think you may need at your final desired speed and set that speed, then adjust accordingly if it fails or if it passes with ease you could slowly lower it over a short period of time until it fails again then you would know you lowered it to far.

There are far to many variables at play to try to think about what voltage you need at a said speed for you to figure out on paper based on some others findings. Sure you may be able to get close or some idea, but without actually going at the overclock you are not really much closer in the end. I say this because of all those variables are in play, and differ per board/CPU/ and ram. And all the other settings you set in the BIOS may all affect another setting you are not even thinking about in any of this.

You need a voltage increase when it fails, there really is not going to be any equation that will tell you for sure what speed you need what voltage. You just have to try it first hand as all chips/boards/ram/user variables and boards many variables are all just that.... To many variables for any equation to aptly apply to

And wow, if you are really this concerned about voltages and lowering them how are you going to overclock or why? I mean I just wondered after all this why you do not leave C1E and EIST Enabled and not overclock. No offense meant there, but wow you are putting in some major thoughts and time about nothing really. You wont save that much energy, and your temps are not going to be that much lower if you use .25 less volts anyway

[vacapp]

First of all thanks for your detailed comments.
You are right I ve spent a lot of hours (days) to study as many information as I could find regarding this issue. It was worth, although I was hoping to come much closer, anyway I think I am prepared to avoid some mistakes which I had definitely made without this knowledge what we discussed in this thread and in the thread of the other guy who is looking for somehow similar goals.

Originally Posted by Lsdmeasap

And wow, if you are really this concerned about voltages and lowering them how are you going to overclock or why? I mean I just wondered after all this why you do not leave C1E and EIST Enabled and not overclock. No offense meant there, but wow you are putting in some major thoughts and time about nothing really. You wont save that much energy, and your temps are not going to be that much lower if you use .25 less volts anyway

I see what you mean, but the reason of such deep research (I can only repeat some already mentioned thoughts) was, to decrease the time of endless guessing and figure it out before starting OC, what final achievements should/could I set considered followings:
1.) to keep stock cooling
2.) to keep the stock voltages if possible or if I can get a much better result with some addition voltage, then just definitely upto the maximum allowed specification from the manufacturer -> in another aspect, what maximum voltages can work considering No.1 and heat
3.) to achieve the best overall performance considered above conditions (I thought if I can implement all results/rules/facts/equations I came close to the estimated crossoverpoint - in the meantime it turned out too hard only alongside the theory...)

These were my initial purposes. After all, I am considering now to make a concept how to process the overclocking, implementing all the results of my research.
Therefore I post later so we could discuss the way or (independently from my ideas) how would you start to get the best results.
I will post the bios default datasheet as usual also and all details of my config again.


P.S.: I wanted to write a short reply first to answer your comment.

[Lsdmeasap]

A short reply? Uhh oh

I hope I did not make you mad, my comment was not meant to offend you at all

[vacapp]

Originally Posted by Lsdmeasap

A short reply? Uhh oh

I hope I did not make you mad, my comment was not meant to offend you at all

Of course not. But I would like to reply to any comment in a short period. This is not always possible as you see, but I have very big problems at the moment and I try to fix them, maybe you could tell me, where should I post such thread? I could not even start overclocking as I mentioned you previous, to test some oc settings. I am really pissed off.

First of all I have some problems with my ATI graphic card (as I ve read there are a lot of people who have trouble...) now I have installed the newest ATI catalyst software we will see.

The other problem (which I can not solve) is my internet connection () . One problem is not enough ...

--- This issue has been moved to following thread ---

[Lsdmeasap]

ATI works great for me!

I have replied to your other topic, just making my rounds is why you se me here