HOWTO: Overclock C2Q (Quads) and C2D (Duals) - A Guide v1.7

[TheDrake]

Originally posted by: cEvin Ki
@TheDrake:

based on what those articles indicate, i really do think that is what's happening. now, i do not know for sure, as i am not an IT tech.

but it would make sense, if behind the scenes, the memory was really running at that actual NB frequency, and therefore was entirely too overclocked. also, since you have the Sandra suite, one of those posts indicated that Sandra gave the true FSB speed in it's system specs display. look at that above highlighted link, post #2, at his screenshots. (the "mainboard" tab in Sandra, as compared to what CPU-Z displays.) i'm curious as to what your's reads.

i'd do it too, but since i rebuilt my system late 07, i've not played with Sandra at all.

i ran some numbers last night too. if you do use frequencies between say around 350 to 399 for the FSB, it would seem to work mathematically if you used a lower multiplier than 9. for example, 9/8 X 366 = 411.75 Mhz. could this perhaps be why a system using this FSB will work, but at a multiplier of *9*, NOTHING will make mine POST at 366?

i am itching to experiment!

Yep, your deduction was right! in Sandra it was reading the FSB at the higher rated than what CPU-Z was reading, very interesting indeed. In my attempt to lower the FSB and up the multiplier I really didn't have much luck. It seems as though I can get it to post between 370-390 when the memory is at 1:1, but if I try to up it to 5:4 then no go at all even though its still well below what the DIMMs are rated at. I can certainly get it to post above 400 but then I get into CPU stability issues. I can get 3.9 GHz to post ~ 433 with 9x multiplier but thats about it, lol. It is no where near stable. I can drop it down to 410 ~ 3.7Ghz, but I think this processor will only do 3.6Ghz stable within safe temperatures on air cooled. So my best bet seems to be the 450 FSB 8x multiplier with memory at 1:1. I was hoping I could get some more out of this memory but thats alright, it was cheap so you get what you pay for

This is very interesting behavior that I havent seen before! Its always fun to learn though! Thanks again graysky for this great thread and help. Thanks cEvin Ki as well for pointing out what I didnt put together, this would have driven me crazy if I couldnt figure it out, lol.

 

[cEvin Ki]

well, i'm still researching this. thanks for letting me know how Sandra was reading yours, TheDrake. it seemed all too probable!

ok, two things.....

first, i'm reading this mind-blistering article on RAM timings, specifically tRD (read delay) and how it has more effect on system memory performance than any other timings alone or in concert. --> http://www.anandtech.com/mb/showdoc.aspx?i=3208&p=1 give this a very close peruse. not only does this explain the reasons for setting this manually in bios, but more importantly, i see nothing at all indicating what happens when a different CPU multiplier is used in regards to the FSB, or the memory speed. maybe it only affects the internal clock of the NB itself? i am going to read it several more times and get back on this before i rush to any conclusions.

second: if you read graysky's article near the end, (where he touches on the subject of "what is the better overclock", a high divider/ lower FSB vs. low divider/ highest FSB) it is clearly argued and empirically proven, that at a given CPU core frequency, there is no real difference in real world applications. i firmly believe this to be so. follow the links he cites to the x264 benchmark and the results pages. so i was wondering, since your (TheDrake's) CPU runs at 3.6 Ghz (450x8), have you ever tried it at exactly 400x9? it SHOULD theoretically do that at your present vCore, SHOULD give you the same core temps, and would NOT be superclocking your NB. actually, you may find a divider that would run your 1000 Mhz RAM at specified speed and timings.

 

[TheDrake]

hmm, thats something strange that I have just noticed, I wonder if I tested it out before or not. But I actually cant get it to post doing 400 FSB 9x multiplier with 5:4 memory no matter how I set my memory timings, I put them as lax as I can get and still no post. I am begining to wonder if this memory will even post at 1000Mhz... When I have some time I am going to drop the processor to default settings for it and just see if I can get 1k memory to post at all. Might explain some overclocking problems I have been having too.


Edit: Alright I had some time tonight to mess with it and it seems when I had the memory timings set to manual it wasnt posting, as soon as I did 400 FSB and did auto for the timings everything came up, running some tests now to check for stability. Checked Sandra real quick and bandwidth jumped to 7.45GB/s

 

[cEvin Ki]

@TheDrake:

nice! which Sandra suite, and which tests, BTW? it's been SO long since i used that. i may give it a go, just for the heck of it. (the entire point of this thread is overall system stability in an overclocked state, and real-world use. so, bottom line: if you can get your newest settings to remain stable, and your RAM running up to rated speed, i'd say you did well. pat yourself on the back and call it a day!)

 

[TheDrake]

Originally posted by: cEvin Ki
@TheDrake:

nice! which Sandra suite, and which tests, BTW? it's been SO long since i used that. i may give it a go, just for the heck of it. (the entire point of this thread is overall system stability in an overclocked state, and real-world use. so, bottom line: if you can get your newest settings to remain stable, and your RAM running up to rated speed, i'd say you did well. pat yourself on the back and call it a day!)

Its Sandra Lite (the free one) version 2008.1.13.12 and I am running Vista 64-bit (if that matters) and its the Memory bandwidth Benchmark. I just like having a number to reference even though its most likely negligible in real world performance.

Everything appears to be all stable and good, running close to 24 hour burn in prime95 and everything is great. temps are a little higher than I wanted and its hitting 55C at max.... I am not worried about it though. Memory times when on auto are actually at 5-5-5-13 for 1000Mhz. Pretty darn good for a Q6600 though, didnt think I would hit 3.6Ghz stable.

 

[cEvin Ki]

@TheDrake:

again, NICE! i've been too chicken to try 3.6. what vCore and what is your VID on Coretemp (if you use that to measure temps that is)? also, what is your voltage showing as in CPU-Z at idle and under load? oh, and thanks on that Sandra info, BTW. i'll try tonite and get back.

 

[TheDrake]

VCore is set to 1.45 in the BIOS, when its idle it hovers around 1.42V and on full load its ~1.36V. VID according to Core Temp is 1.2625v, but I like SpeedFan a little better than Core Temp personally.

I am going to load up a few games and and see how everything goes now. I only have HL2, battlefier 2142 and oblivion on hand right now to test with.

 

[cEvin Ki]

ok, for posterity...

my machine runs stable at 1.24375v set in bios for vCore, ddr2 800 running 1:1 @667 3-3-3-8, at 333x9 (q6600). i decided to attempt some overclocking even further.

to even get into windows for a split second, (at 400x8) i had to raise the vCore six notches, to my chips VID of 1.2750. just to rule it all out, other voltages such as FSBT, PLL, etc. i set to auto. and we know these overvolt all too often. memory voltage i raised to 2.25 (per hardware monitor -- i also have the same RAM graysky has, Crucial Ballistix ddr2 800 4x1GB.) as i said above, for an insignificant gain of 200 Mhz of core speed, and an also yawn-enducing gain of RAM speed, i am FAR from stable even with a drastic increase so far in voltages. my last attempt, once i got into windows, gave me a warning that "serious errors" had occured. SO!..... i KNOW i would have to raise voltages even further, and/or loosen my RAM timings to be anywhere near stable. this is counter-intuitive since such a meager increase in RAM speed would be negated by looser timings. and the obligatory increase in temps would exist.

now, i STILL have not tried 400x9, which MAY be more reasonable as it is the default multiplier..... but, so far, in my opinion it's pointless and ridiculous for me personally to overclock further, for a few percentage points of performance gain i MAY achieve.

@TheDrake: how did the gaming go? kudos on that 400x9, BTW. oh, and Sandra reported a 6115 for my memory at the above settings. 114 percent efficiency.

EDIT: ok, just for fun.... set the "AI Clock Twister" option in bios from "auto" to "strong." now the bandwidth is : 6351 Int. and 6362 Float, with a bandwidth efficiency of 118.84 and 119.1 percent respectively.

 

[TheDrake]

Yeah, everything is great on it, played a few games and had them running for a few hours and everything is great. I am going to take the safe route and drop it down to 380x9 @ 1.325 Vcore just so I can keep temps down cause when summer rolls around I dont like to have the A/C running much and I dont want these temps going any higher than what they already are. But its certainly nice to know I can push it successfully to 400x9 if I ever want to

 

[Nishua]

Look like a great guide. I haven't even read the whole thnig, but when I buy my new PC when it comes to overclocking it I will refer to this guide. I'm a total newbie so I appreciated teh amount of detail that went into the guide. Hopefully, I won't mess things up. I don't play on overclocking my system when I initially get it because tiw ill be update but maybe a year later I will. Is it worth doing it out the box ?

Anyhow, thank you for taking the time in typing out this guide.

 

[graysky]

I don't think there's a reason to wait really. If it's stable @ the stock settings, computers aren't like internal combustion engines that needs some break-in time to allow parts to mate together. I'd say give it a few days running prime95 to verify stability, then start overclocking... I should say if you're getting a PC from a major retailer like Dell or HP, odds are you won't be overclocking it period due to crippled BIOS options on propitiatory motherboards.

 

[rocketmen2k5]

How does O/Cing a CPU and or other components effect their life span?
In other words does say O/Cing my CPU reduce it's lifespan?

 

[graysky]

@rocket - there are a number of threads discussing this. IMO, as long as you're keeping the vcore within specs, you are stressing things that much. I have no real long-term data to back up my statement though. My Q6600 has been running 9x333 for nearly a year now with no problems. I had my older 3200+ overclocked for years with no problems either.

 

[robebdr]

graysky,

Fantastic OC reference. I've built my own systems since the late 90's, but never OC'd one. Didn't feel comfortable doing it. This thread convinced me to try and I am impressed. I'm running a Q6600 with an asus p5kc (bought them a week ago), crucial xms2 DDR2 800 memory, antec 900 case and a coolmaster hyper 212 cpu cooler. Running at 3 gig with no effort at all. Core temps at or below 54 degrees C with prime 95 running. An item of interest, I set the cpu voltage to auto as you recommend for the initial start up and cpu-z is reporting core voltage around 1.24. Not sure if I'll change that, but do need to set the memory to a more aggressive setting. Thanks again for the guide.

 

[andrei3333]

hey guys have read this article a while ago but in a nutshell to get my e8400 from 3.0 to 3.6 all i have to do is make the FSB go from 333 to 400, disable that intel thing that slows down the cpu when its at idle and increase the vcore but by how much ?

with my sig below will i see a noticible difference in games or no ? its pretty fast as it is btw

 

[graysky]

andreei - there isn't an easy answer to that question. I'd say just set up your overclock and start out low since failure will occur more rapidly. Document what you do on a piece of paper so you know and start inching up. Also know that you might need to bump some mb voltages as well.

 

[brjoon1021]

so, when running Prime95 for a dual or quad core CPU - there should be a check mark next to the "round off checking" option in the advanced tab, right?

Thanks,

B

 

[graysky]

right... you must enable round off checking or else errors aren't reported

 

[OLpal]

You'll probably be able to run stock volts, just keep an eye on your core temps [just to error on side of caution]. Makes for 1 easy , smooth, overclock !!
Ol'Pal

Originally posted by: andrei3333
hey guys have read this article a while ago but in a nutshell to get my e8400 from 3.0 to 3.6 all i have to do is make the FSB go from 333 to 400, disable that intel thing that slows down the cpu when its at idle and increase the vcore but by how much ?

with my sig below will i see a noticible difference in games or no ? its pretty fast as it is btw

 

[graysky]

OK, just got finished upgrading the guide to version 1.6!

? Added a detailed section to help you find the minimum stable CPU and MB vcore settings. Check it out (near the bottom of the guide entitled, ?Stress Testing and Minimizing Your Vcores?)!
? Updated the CPU table
? Provided a less than $5 method you can use to shave off some NB load temps (in the thermal management section at the end of the guide).

 

[OLpal]

Good Job Graysky, i'm sure everyone appreciates your guide, i know i do !!
Thanks for your effort !! Ol'Pal

Originally posted by: graysky
OK, just got finished upgrading the guide to version 1.6!

? Added a detailed section to help you find the minimum stable CPU and MB vcore settings. Check it out (near the bottom of the guide entitled, ?Stress Testing and Minimizing Your Vcores?)!
? Updated the CPU table
? Provided a less than $5 method you can use to shave off some NB load temps (in the thermal management section at the end of the guide).

 

[graysky]

Thanks for the kind words, dude. I'm glad to put the time and effort into it so long as I know people are getting something out of it!

 

[bryanW1995]

great update graysky! however, I do have one problem with it: you listed the Q9550 as a currently available intel processor. What a cruel, cruel joke!

 

[graysky]

@bryan - haha! If you read the history from the online inventory locater for the q9550, PCConnection did have some:

+April 24, 2008, 12:40:38 PM EST - Intel Q9550 Boxed was in stock at PCConnection

+April 23, 2008, 2:59:33 AM EST - Intel Q9550 Boxed was in stock at PCConnection

+April 21, 2008, 11:14:42 PM EST - Intel Q9550 Boxed was in stock at PCConnection

...why wait when you can get a X3360 right now?

 

[BTRY B 529th FA BN]

Wouldn't it be if it's quad core pumped, DDR3 1600 a ratio of 1:1 for a quad core with a fsb of 1600 like the QX9770; and in a dual core system, if a dual core (dual pumped) used 1600fsb using DDR3 1600, a ratio of 1:2?

And also in this statement:

The 667, 800, or 1066 refer to the max speed (front side bus in MHz x the divider) for which it's rated. It's a little more complicated... these are DDR2 type so you divide those numbers by 2 (dual channel). So you'll get 667/2 = 333 MHz; 800/2 = 400 MHz; and 1066/2 = 533 MHz. The CPU : DRAM divider is discussed later in this document. If you want to run a FSB of 400 MHz then you?ll need AT LEAST DDR2-800. The same rule applies for the newer DDR3 memory, divide the 2nd number by 2 since they too are dual channel DIMMs. For example, DDR3-1333 and DDR3-1600. 1333/2 = 666 MHz and 1600/2 = 800 MHz.

Is it NOT because of dual channel but DDR (double data rate) as apposed to single data rate (single sided ram). Before you were saying the divider was 3 for DDR3 as if it has 3 sides to a dimm???

just wondering

Thanks