Cheers Drums! That adds some more flexibility to the solution.
I also found mentioned elsewhere an indication that my E6700's 'locked' x12 multiplier is only locked upwards, but not downwards. That could bolster my plans even more.
If that is so, it shapes my strategy greatly.
Here is what I have deduced so far from reading other THW fora discussions on the topic.
To optimize performance using the components at hand.
Modest increase in performance, with adjustments to ensure better system stability.
No voltage increase on CPU, and no timing changes to RAM, to ensure stability and long life.
Stock E6700 has a Bus speed of 266MHz (rated FSB of 1066) with a multiplier of 12 (266MHz x 12 = 3192MHz or 3.2GHz).
My G.Skills RAM have a frequency of 400MHz (rated as 800MHz as it's DDR2).
To get a 1:1 ratio between the 400MHz RAM and the CPU, I adjust the CPU bus speed to 400MHz.
At a multiplier of x12 that gives a rating of 4800MHz (400MHz x 12 = 4800MHz = 4.8GHz). That is too high for my Zalman AM air cooling.
I drop the multiplier to 9, to achieve the 1:1 ratio at 400MHz, with a rating of 3600MHz (400MHz x 9 = 3600MHz = 3.6GHz).
Meanwhile, my changing the CPU FSB frequency from 266MHz to 400MHz changes the overall BUS rating from 1066Mhz to 1600MHz FSB. (This 1600MHz FSB freq. is supported by the EP45-UD3R motherboard.)
IF my E6700 can handle this overclock, the CPU clock speed will be increased by 11%, the system FSB will be increased by 33%, the RAM will be running synchronously at it's designed speed, and the overall system performance should be optimized.
Most say that the CPU voltage will not need to be increased to support this type of overclock. If so, then no serious heat increases should occur, and a strong after market cooler can help ensure best performance.
you dont need to have a 1:1 ratio between ram and cpu...what you should do is leave your cpu multiplier where it is and increase the fsb until you get your desired speed.....so you would increase your fsb to 300...that would force your ram up to 902mhz or so....which in my opinion is going to be completely stable you could leave it at that (most likely)....but if you dont want to run your ram any higher than 800mhz you could drop your ram ratio to the next step down which should be around or lower than 800mhz....does this make sense? your oc will probably not be stable with a fsb of 400 without increasing voltage....to get the easiest oc you want to increase your fsb as little as possible
you dont need to have a 1:1 ratio between ram and cpu...
So you are one of those who feel the 1:1 fixation I've heard about is over-rated?
what you should do is leave your cpu multiplier where it is and increase the fsb until you get your desired speed.....so you would increase your fsb to 300...
I have come to understand some of the best overall performance can come from opening up the bus speed, and felt sure 400MHz would be more effective than 300MHz. But I assumed that the clock was what determined the need for voltage.
that would force your ram up to 902mhz or so....which in my opinion is going to be completely stable you could leave it at that (most likely)....but if you dont want to run your ram any higher than 800mhz you could drop your ram ratio to the next step down which should be around or lower than 800mhz....does this make sense?
I only don't fully comprehend the tolerances involved. I was leery of doing anything in excess of the RAMs native settings, assuming that could shorten their life, and/or cause errors. I was also leery of manipulating any ratios because I imagined that could cause unwanted latencies.
your oc will probably not be stable with a fsb of 400 without increasing voltage....to get the easiest oc you want to increase your fsb as little as possible
A few examples of other overclockers' attempts with similar CPUs (wolfdales) seemed to indicate that I might get away with that 400MHz FSB without taking the voltage out of stock setting. With my Gigabyte and Zalman brand equipment at hand I was hoping that might be so.
So in my first draft I proposed dropping the multiplier to dump a 33% increase on the FSB with only an 11% change to the clock (and keep my 1:1 with RAM), but YOU advocate keeping the multiplier high so that I'm only increasing both clock and bus equally at only 11%, and for me to stop caring so much about any asynchronicity or possible latency factors regarding the memory?
Asynchronicity. Latency. Damage to memory modules. These are all 'boogie-men' for me that you seem to indicate I wouldn't be as concerned about if I had more first-hand knowledge.
Thank you for your time in this. I hope this convo might be help to others too.
the thing about overclocking is the less you gotta change to get to the speed you want the better...if you have the ability to change your fsb only a little bit vs. alot you should try to go the little bit route....the more stuff you change and the greater you change it the higher the chance of you have stability issues...increasing the ram speed up 100mhz really is not gonna cause any issues or shorten the life span of the ram...believe it or not I think you are over analyzing it (not a bad thing).....I would just start off by increasing the fsb until you get to the cpu speed you want (leaving the voltage stock)...then test the cpu for stability using prime95 and test the ram for stability using memtest...if they both pass then you are good to go
Just letting you know and for other peoples research later on...
I took your advice to heart about the smaller moves thing. And I know that you ARE right.
But given the dual BIOS feature on this Gigabyte mobo and the kinds of trends I was seeing reported about this particular chips family (wolfdale's),v I saw little harm in giving my 400MHz bus idea a go, first off.
I fully expected your prediction to be vindicated within minutes if not moments. And I know I was being very pushy doing such a first OC without any incremental exploration.
Well, so far so good.
I'll admit to only Prim95'ing the new OC for just 6 hours without errors, but I have been playing games etc. etc. for a full day without incidents.
So for the record, AT THIS TIME, it seems like an OC on a E6700 3.2 GHz with 266MHz FSB to a 3.6GHz with a 400 MHz FSB and no voltage change 'can' be stable.
I'll hit back here the day this OC tits-up and I have to dial it back.
*Funny thing is, my PC profile under computer properties in Win 7 reads the OC as 4.8GHz. I had a laugh showing that to a pc geek friend and making her believe it was true.