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

[graysky]

@dragon5152 - you're running 9x400 and want the RAM to go faster than 1:1? By your sig, you only have DDR2-800 so you're at the max for the memory right now. You would need faster DDR2 to run it any faster unless your memory o/c's really well. For reference:

FSB DRAM Ratio
400 800 1/1
400 1,000 4/5
400 1,200 2/3
400 1,333 3/5
400 1,600 1/2

If you had DDR2-1066 you could likely run @ a 4/5 divider for example.

 

[Germonicus]

Thanks for all your help and advice graysky.Your guide has been my first time OC manual and gospel

As I say,it's my first ever successful OC,I could never get any stability out of my old AMD CPU's but I'm now running my E8400 @ 3654.4 Mhz,FSB 406x9,RAM 4.0-4-4-4 4:5 2T and it's Prime stable for over 3 hours.
I even felt safe in increasing the Vcore,which scared the bejeezus outta me previously

 

[TheRyuu]

Originally posted by: graysky
@dragon5152 - you're running 9x400 and want the RAM to go faster than 1:1? By your sig, you only have DDR2-800 so you're at the max for the memory right now. You would need faster DDR2 to run it any faster unless your memory o/c's really well. For reference:

FSB DRAM Ratio
400 800 1/1
400 1,000 4/5
400 1,200 2/3
400 1,333 3/5
400 1,600 1/2

If you had DDR2-1066 you could likely run @ a 4/5 divider for example.

I do have DDR21000 (1066) ram, it just won't run at it's rated speeds without giving errors. I really don't care though as I believe the CPU overclock is much more important then getting my ram to blast at DDR2-1000 or something.

I'm not going to worry about it now, I like what I have (3.6ghz on quad core? what more could you want...)
Really, really, really happy with my purchase of a q6600.

And I believe my ram is erroring probably because of something to do with the northbridge. I don't want to raise the NB voltage anymore as I'm already at 1.44, so for now, I think I'm going to leave it.

 

[Heller]

seems i thiought oiverclocking was just upping the volts and changing the frequencies, i did'nt relize that my ram might be afftecing the o/c

so now im running 3.6ghz with 1:1 ram ddr800 and 1.475 volts.


lets see if she holds up.


assuming she does what should i set my ram back to? i've not real good with latencies..

 

[graysky]

Originally posted by: Dragon5152
I'm not going to worry about it now, I like what I have (3.6ghz on quad core? what more could you want...)
Really, really, really happy with my purchase of a q6600.

And I believe my ram is erroring probably because of something to do with the northbridge. I don't want to raise the NB voltage anymore as I'm already at 1.44, so for now, I think I'm going to leave it.

I think you're right. I just saw that you're running 4 DIMMS. I've read that folks with similar setups (4 DIMMS) have also had issues running at higher multipliers. Just to scratch the itch, you might wanna up the NB, IHC, and FSB vcore and see if memtest passes. No biggie either way though.

 

[Heller]

what other factors am i missing?

i can't get my q6600 to be stable @ anything less then 1.53v.

I've also loosened my ram to 5 5 5 15.


go gentile on me, im an o/c nub.

 

[TheRyuu]

Originally posted by: graysky

Originally posted by: Dragon5152
I'm not going to worry about it now, I like what I have (3.6ghz on quad core? what more could you want...)
Really, really, really happy with my purchase of a q6600.

And I believe my ram is erroring probably because of something to do with the northbridge. I don't want to raise the NB voltage anymore as I'm already at 1.44, so for now, I think I'm going to leave it.

I think you're right. I just saw that you're running 4 DIMMS. I've read that folks with similar setups (4 DIMMS) have also had issues running at higher multipliers. Just to scratch the itch, you might wanna up the NB, IHC, and FSB vcore and see if memtest passes. No biggie either way though.

Those voltages are already up there
And I only have 2 sticks of ram. But as we've said, no biggy

I have one notch of room on the ICH and fsb voltage so I think I'll up those and see if it's stable. I won't even have to run memtest because it just crashes prime95, if that doesn't crash, then I'll move on to memtest.

 

[graysky]

Originally posted by: Heller
what other factors am i missing?

i can't get my q6600 to be stable @ anything less then 1.53v.

I've also loosened my ram to 5 5 5 15.


go gentile on me, im an o/c nub.

Sorry dude, either your chip isn't gonna tolerate it or your MB chipset isn't gonna. I believe you have reached the limit.

 

[TheRyuu]

Well, final overclock for me.
Was fun

Prime stable, 12 hours tested (really closer to 16 or more, I forget), also folded some crap in rosetta with no problem for 24 hours as well as OCCT stable.

3.6ghz (400x9) @ 1.45 vcore
MCH 1.44v
ICH up two notches
vtt up one notch
ram @ 2.0v @ 400mhz (ddr2-800) @ 4-4-4-12 2T

Stable and HOPEFULLY safe

Thanks graysky.

*updates sig*

 

[graysky]

Cool man, glad you were successful... 9x400 is nothing to sneeze at for a 65 nm quad.

 

[blanketyblank]

apparently ram timings make a huge difference. Before I had them set to auto and I couldn't get stable at all without a vcore above 1.35. Now I can do a vcore of 1.26 with my ram at 5-5-5-15 with a fsb of 333 mhz and I can do 3.0ghz.
btw when do you have to raise the north bridge voltage? Do you just do this when an overclocking attempt fails? I want to try to get up to 400 mhz/3.6 ghz since that is supposed to be around the limit for my motherboard.
I've been able to boot into windows at this setting with 1.38 vcore, but only once before it restarted so I'm guessing it's not stable. Should I try increasing my n/b and what level is safe? By default it's at 1.25 and I can up in .25 increments.

 

[graysky]

Yeah, you can start to bump up the motherboard vcores if you can't boot or remain stable and your cpu vcore is maxed out for example. Try raising your mb vcores to stabilize the o/c... there are a lot of variables. Use prime95 v25 for stability testing btw.

 

[TheRyuu]

Originally posted by: graysky
Yeah, you can start to bump up the motherboard vcores if you can't boot or remain stable and your cpu vcore is maxed out for example. Try raising your mb vcores to stabilize the o/c... there are a lot of variables. Use prime95 v25 for stability testing btw.

Also, I'd run that 30min OCCT stability test as well. It seems to give a GENERAL idea of stability or not. It is NOT a replacement for prime95 testing though. I would still recommend running prime95 for 12-24 hours after you've tested it with OCCT, that way you don't have to wait so long sometimes (OCCT will normally fail within 3-5 minutes, prime95 could take hours)

 

[blanketyblank]

What's OCCT? is that orthos?

 

[cEvin Ki]

graysky, your guide is magnificent. like you, my q6600 runs at 333x9 at 1.2625v. i left a *cough* long post over at your thread at tomshardware forum about my success with the vdroop mod. guys, and gals, it simply works!

got a question.... again, like you i've elected to run identical ballistix memory at 1:1. (and yes, i've scoured the threads on better overclock, and memory dividers. lovely work!) i got to reading at a few threads over at the ballistix club at overclock.net. some of those guys with the same ram, cpu, and boards, get theirs to run at 3-3-3-8 2t. tried it, and yep, it works on mine! however, i also ran across another thread.....

http://www.thetechrepository.com/showthread.php?t=195

what do you make of this? it does make mathematical sense, but to me, seems to indicate that most of us are running our memory way under potential, based solely on calculated actual real world latencies. i'd just like a second opinion, and your's i respect. and NO, i don't care a can of beans about benchmark fluff. i just want a lean mean beast.

ok, i will stop. i apologize if this post is another tedious long one.

 

[graysky]

@cevinki - Thanks for the kind words. That article is pretty detailed and I don't actually have the time to dig into it. What I do know is that you can test your timings with superpi mod 1.5 which is pretty sensitive to changes in memory and see if there are any differences, then repeat the tests using applications you actually use and preform the same analysis.

BTW, I think the formula for the first five memory timings are:

x-x-x-3x-(x+1)

So if x=3 the first five timings should be: 3-3-3-9-4

So, set up 3-3-3-9 and run the 16 or 32 M calculation (better the 32 M) in superpi. Record the time and repeat at least 3 times. Average the results. Reboot and setup 4-4-4-12, run the same test (16 or 32 M) a total of 3 times and compare the average results you got. If there are differences, go ahead and repeat the experiment using some app or game you use that has the ability to record a benchmark - some games have timed demos that'll run and report a fps score, other apps such as winrar or x264 report benchmarks, etc. If you do a bit of video encoding, try my x264 benchmark that'll run the same encode 5 times for you and report all the data in a nice text file. Whatever you use, make sure you run it at least 3 times averaging the results at the various timings and see if there are tangible differences.

I haven't tried it on my quad, but there is a guy who uploaded the x264 benchmark test using his E6400 at three difference o/c levels each with three different timings.

Here are the results

 

[graysky]

Originally posted by: blanketyblank
What's OCCT? is that orthos?

It's another stability testing program.

http://www.ocbase.com/perestroika_en/index.php?Download

 

[cEvin Ki]

thanks graysky. will give it a go as soon as possible. and yeah, i meant to do that benchmark and post my results. you've got quite the list going there.

and blanketyblank, trust me, it does test well. if you are unstable, it'll poof quickly.

 

[graysky]

@cevin ki - cool lemme know if your results are in line with that I posted. Also remember that x264 is only one measure -- there are others.

 

[Vishuss]

First, thanks for the awesome guide, graysky. For someone who had no clue about OC'ing coming into this, it's been really informative and helpful.

I've run into some issues OC'ing my system.

This is my setup:

Core 2 Duo E6750 CPU
4x1gb 667mhz DDR2 RAM
GeForce 8800 GTS 640MB
Microsoft Windows Vista 32 bit
Gigabyte GA-P35-DS3L Motherboard

Normally the settings for my CPU are (I'm probably mixing up the labels but the numbers are right):

8.0x multiplier (this is limited to a 6.0-8.0 range)
333 mhz FSB
FSBRAM ratio is 1:1
PCI-E frequency set to 100mhz
5-5-5-15 RAM timings

I've tried to overclock this deal by increasing the FSB mhz, but I can only get things stable at 370 mhz and below. If I increase the Vcore a little I can get it to 380 mhz, but that's it and the system can't run Prime95 without errors above 370 mhz.

My computer won't even POST above 380 mhz. Is there something I'm doing wrong or some other major item that I should check for? Thanks.

 

[graysky]

Originally posted by: VishussCore 2 Duo E6750 CPU
4x1gb 667mhz DDR2 RAM
GeForce 8800 GTS 640MB
Microsoft Windows Vista 32 bit
Gigabyte GA-P35-DS3L Motherboard

That RAM is only rated to a FSB of 333 so you're pretty lucky if you can hit anything higher... best advice is to get yourself some DDR2-800 or -1066 DIMMS.

 

[Vishuss]

Alright, thanks for the info

 

[schizoid77]

Hey guys, which CPU Vcore value am I to trust, these programs are all over the place....

Asus Probe = 1.33
CoreTemp = 1.3000
CPU-Z = 1.248
CPUID Hardware Monitor = 1.25
In the BIOS reports about the same (usually higher) as the ASUS Probe....

I'm thinking CPU-Z and CPUID HM are giving me the most consistent results, but WTF?

What else can I run to give me accurate temps?

Also, I'm running idle at 3.0Ghz with the following temps;
Core 0 = 28C
Core 1 = 28C
Core 2 = 30C
Core 3 = 27C

Overall Temp = 23C
Ambient = 34C

All according to Asus Probe.....these are obviously good temps but only if they are 'true'.

CPUID Hardware Monitor reports same Core Temps as Probe....btw.

 

[markreflex]

@graysky
well done on the article.

maybe you can do an update on memory sub-timings like trd and trc which have shown to have a bearing and big involvement on performance.

 

[graysky]

@schizoid - About your vcores: there are two factors in play here that make the values lower than what you dial-in via your BIOS:

1) Vdrop
2) Vdroop

Vdrop is the difference between what you set in the BIOS and what you get in your O/S
Vdroop is the difference between an idle and load state

Some boards are better about both factors than others. Read the temperature managment section of the guide for a link to a pencil mod for vdroop on a P5B-Deluxe (you can pretty much google the terms "pencil mod" and your MB model or "vdroop mod" and your MB model).

As to your temps, since Intel never officially published what Tjmax is for our chips, coretemp, hwmonitor, and others assumed 100 for q6600s but unclewebb has recently written a program he calls realtemp that uses a Tjmax of 85. Who is right? No one knows for sure. Also, idle temps don't mean much... what are you load temps?