Q6600 OC Results / Questions

[Bman.]

Evening guys.

I just finished installing my new cooler (Xigmatek vs AC7). I managed to knock my core temps down to 45 / 40 idle which is currently clocked at

FSB = 370 x 9 (3.330)
vCore = 1.3875
NB = 1.25
Ram = Auto
Ram Volt = 2.1

I want to push it more but I have run out of combinations based on other thread that I have read here. Perhaps I have hit the end of the oc for this chip - how do I know?

I have tried the following combos with no complete boot:

380 fsb 1.3875v NB 1.25
380 fsb 1.425v NB 1.4

I have tried all of the combination between those two hence the large jump. Here are my specs

ASUS P5K
Q6600 (VID 1.275 - G0 Stepping)
Xigmatek S-1283
4 x1GB GeiL (Orange)

Windows XP + SP3

The RAM has is at 4:5

I have read that to get up in the 3.6 range that NB will need at least v1.4. However, I only have passive cooling on the NB. Is it safe to run v1.4 with passive?

Should I keep pushing the voltage on 380 to see if it kicks in?

Hmm 3.3 may be the top end.. at least temps are much nicer!


Benjamin

 

[d3v0]

Hmm NB temps shouldnt be much higher than your full load CPU temps (what is your cpu temp @ full load P95 small fft? Load is much more important than idle haha.)

But yes, 1.4-1.5v is needed for much above 3.3ghz on Q6600s in my experience. Also make sure your ram is not limiting you. I would suggest setting FSB:Ram settings to ratios that dont take the ram much over stock settings when testing overclocks for stability. My Q6600 is a GO stepping with a 1.2725vid as well and it will do 3.6ghz with no sweat on a $65 mobo. Your P5k is a much more capable motherboard and I wouldnt be afraid to push it a bit farther! Best of luck!

 

[Bman.]

Alright, I upped the FSB to 380 at 1.45v - currently running prime95 (blend) at 60c/55c

...I don't see any ratio setting for the ram. I only see Ram Frequency...and it gives me a list ...I selected 766 (for I have DDR2 800)

I will be back for one more step up.

 

[d3v0]

Alright, I would suggest checking cpu-z memory tab to see what your ram is actually running.

 

[Bman.]

Ok, back... 390 (at 1.45) = 3.510 temps... 48/44

I note there is "DRAM FREQUENCY" and it is simply giving me preset options which I believe to be 1:1...4:5 etc.

For example:

780mhz
975
1300

etc - those are all the ratios already done for me. So, if I am sure to set it to 1:1 then I should be ok until I go over 400 fsb, right?

I don't pass prime95 however. I will have one core give out almost right away on blend. I know this is a measure of stability but is it safe to try and run anyway. Last time I ran Prime95 test I failed but ran for months with no issues. If even one core fails that constitutes failure / unstable correct?

Thanks

 

[d3v0]

Yes it does mean failure. The memory tab should show a frequency around 400mhz or so at that setting.

The samw core failing generally means not enough voltage. Did you increase the NB as well? I don't use a pc without at least 12hrs p95 stable.

 

[SkiBum1207]

1.4 is plenty for the NB, im running a 1800mhz FSB no problem with it.

Voltage seems a bit high though, are those values for BIOS or actual?

By they way, the P5K vanilla (i assume thats what you have) has absoltuely horrible vdrop/vdroop. There is a pencil mod that fixes this and makes this board into a absolutely rock solid overclocking board. PM me if you want the diagram (it takes only about ten minuets)

also while trying to find your cpu max, drop the Ram to the lowest speed and up their timings, once you find your max OC, then you can up the speed and drop the timings. The reason to do this is just to remove one of the many varibles when overclocking lol

 

[Bman.]

The BIOS is set to 1.465 (running at 900 mhz x 9) However, it's no where even close to stable. Actual, which I presume to be the reading from CPUz is 1.3xx. I am at the office so I am going from memory.

Ok, I will up the NB to 1.4 as it is currently 1.2 right now

Regarding the ram timings, I don't have all of those option in my BIOS or I am looking at the wrong location. The I currently have the setting at the lowest which is 1:1 default setting. The timings are currently auto. If I keep the timings as specific of 4 4 4 12 is won't get past the windows boot logo.

Pencil mod - sure I am game for it. Yes, P5K... vanilla colored.

 

[Bman.]

Question about the vDrop - can I push the vCore above 1.5v with the vDrop in mind?

For example, right now I am running in BIOS - 1.465. From CPU-z
Core Voltage = 1.400v...1.382v...1.336v under Orthos and Prime95

What is the critical vCore to watch? The BIOS setting or what I am reading from CPU-Z?

UPDATE:

I have increased the vCore in the BIOS a few more notches above v1.465 and my stability didn't get any better. I would still fail Prime95 on one core nearly right out of the gate. I have started backing everything down trying to at least fine something that will pass Prime95 for at least a while.

Benjamin

Currently I am at 370x9 = 3.329 at 1.425 (1.384 indicated through CPU-z). Temps are at 45c and we will just see how it does.

Considering my previous OC of 3.1 with the AC7 ran for nearly 6 months before I slapped in the Xigmatek last week I suppose I will consider 3.3 an ok increase. I am not sure what else I can do to achieve something that is stable. I will check out some other settings on the forums here to see what people have done. I was hoping that 3.6 would be my mark but I don't see that happening. I can get it to boot at 3.6 just not pass any stressing.

 

[Kazappin]

I'm using an ACF7Pro and an Asus P5Q along with my Q6600. I have a stable but odd OC @ 3.2 GHZ (400FSB x 8)

Running with the NB and SB at only 1.1v, cpu voltage of 1.3625, RAM @ 1:1 4-4-4-12, I hit load temps of 55c after 8 Hrs of prime.

What are your idle/load temps? You might want to try upping the Vcc a bit, and try to hit 3.4 GHz. or you can go all out on 3.6GHz, whatever you want I tried 3.6 but even with ridiculously high volts on everything and my ram running at high timings, couldn't get it stable for more than 2 mins. I have a VID of 1.325 though, so it's to be expected I guess.

Good luck with whatever you decide to do next, and nice OC

 

[baditude_df]

Question about the vDrop - can I push the vCore above 1.5v with the vDrop in mind?

It's not really of a question of if you can, it's a question of how much it's gonna take. In other words, you HAVE to keep your Vdrop in mind when you set your BIOS Vcore. If you know you need at least 1.4V to keep a 3.6GHz overclock stable, and you know you have a VDrop of .065v then you need to set at least 1.465v in BIOS. Then the Vdroop after you apply load may sink the core voltage lower than what is required and you will loose stability again. The problem here is, not only are you loosing stability because you are trying to overcome a nasty Vdroop, but the increased voltage you are feeding is causing more heat and further instability.

If you want more stability with that board, you will most likely need to do the pencil mod and or invest in some water cooling. Not really practical or cost effective, but those two things will help you immensely.

What is the critical vCore to watch? The BIOS setting or what I am reading from CPU-Z?

If you're looking to do some quick benches you can really feed the voltage. It is a 65nm after all. I've done 1.7v for 4.1GHz on my q6600. But for 24/7 I kept that chip at 1.4v BIOS/1.375v Actual for 3.6GHz.
Of course I have a rather good Load Line Calibration on my motherboard.

The other option, and my guess, is that if you upgraded your motherboard, you could see a better overclock. Even the P5E-VM-HDMI that I tested out for my folding rig hit a 1.4v 3.6GHz no problem with a Thermaltake MaxOrb.

Perhaps you could make a list of your BIOS settings and what you have changed?

Have you disabled your CPU parameters? ie: C1E, Speedstep, etc.

 

[Bman.]

Thanks for the response Baditude. You are exactly right - I will be chasing my tailing pumping more vCore on account of vDroop only to cause more instability.

I have landed at 1.4375 at 3.5 (390 x 9). That appears to be stable and the temps are ok. Idle 46 / 42 and under max load it will hit 71c. A bit warm for sure but it's only for 2 to 3 seconds during the heavy load.

I think you are right, in the end I either need more cooling or a difference mobo. At this point the instability is heat based. The tests are killing me out on CPU heat warning levels.

On my next machine I would investigate water cooling. I don't know anything about it but will read up prior to my next purchase. Right now, not worth the potential .2 or .3 I might attain.

Benjamin

 

[baditude_df]

TBH you shouldn't really need watercooling to keep a 3.6GHz Q66 Stable or terribly cool. I'd like to see your temperatures down in the low to mid 60's with a good air cooler and good case ventilation, and although your cooler should be sufficient, I've only ever done an air job with a TRUE.
Try playing with your BIOS parameters some more to see what you can do in terms of running that CPU stable at a lower voltage. I know on my X48, giving slight adjustments on GTLREF values, then the CPU and MCH Skews will allow the CPU to run with less voltage and run more stable.

Not sure if you have read these articles, but they are quite useful when you have exhausted your brute force options:

http://www.thetechrepository.com/showthread.php?t=87
http://www.edgeofstability.com/articles/dfi_p35/gtl/gtl1.html
http://www.edgeofstability.com/articles/dfi_p35/default_gtl.html

http://www.ocztechnologyforum.com/forum/showthread.php?p=285270

 

[Bman.]

I will review the articles. Currently brute force is all I really know so it's not surprising Im not truly stable at 3.6ghz.

Whoa...what happened... I think I just fell asleep reading the first article.. ha. Just kidding but that is a heavy read!

 

[baditude_df]

Those are very meaty articles and not for the feint of heart. I think though, that what separates good overclockers from mediocre ones is the ability to understand and apply that kind of knowledge. I, for one, am only part way there. Especially when you consider that each motherboard(even boards from the same manufacturer) have a completely different set of BIOS parameters and you almost start from scratch each time you switch.

And BTW, to get to 3.6GHz with a Q6600 shouldn't even take things to the level of brute force. I've done 3.6GHz on 3 different VIDs (1.2625v, 1.3250v and 1.2000v) All did the job no problem with 1.4v. Both on water and with a stock TRUEs. I still have the 1.200VID chip and the other two are currently in friends machines overclocked to that same frequency under those stock TRUEs in well ventilated and very well wire managed Antec 300s.

 

[Bman.]

Yes, I suppose I am missing the whole level of dedication to understanding. I mean, I do want to understand but it's a whole level of language and computing that would take a lot of time and question nagging to get there. Not that I don't want to know - I mean, it's a wonder I have come this far. In the end, I OC not for the sake of OC itself but rather to get the most performance I can for the simulator.

I don't have any issues getting to 3.6... just the stable part (prime95 / OCCT etc)

Here are the images you requested (see below). Also I have a few questions.

1) I see that my BIOS is no longer current (1201 is out) - I have 1006. Should I upgrade. The mobo itself and hardware components seem to run fine with the mobo as is.

2) FSB Strap to North Bridge - should I change this setting?

3) Transaction Booster - should I change this setting?

 

[jason4207]

Leave strap on auto, but you can try changing Transaction Booster to relax level 0 or 1. This will relax your Performance Level (tRD). If you find it helps, then go back later and try to use Boost level 1, and work your way up. This affects the bandwidth of the NB, and is pretty important for performance, but a setting that is a bit tight can cause instability.

Go ahead and disable DRAM Static Read Control. This also affects tRD and some other sub-timings as well I think.

Clock Overvoltage control helped on some of the quads I had. You can put it up to 1.00v, and see if that helps.

Set LLC to enabled.

CPU PLL 1.6v. You can try 1.7v as well to see if it helps.

FSB Term. Start at 1.2v, and work your way up if it helps.

Playing w/ the GTLs might help as well.

You might want to try setting the multi to 8, and play w/ FSB Term, NB voltage until you can pass the OCCT2.0 custom 2hr RAM test.

If that works up the multi to 9, and use vcore and GTL adjustments to stabilize w/ P95 small fft.

You don't want to keep voltages on auto, though. It's best to pick the min voltage for each setting, and work your way up 1 at a time until you figure out what helps, what hurts, and what does nothing. You can consider the Auto setting to be the 'min' setting w/ GTLs, though, as they aren't voltages like the rest...sometimes Auto is best on GTL.

 

[Bman.]

Thanks - I will give some of those settings ago. Wow, so many variables....

Regarding mulitplier... I thought I was stuck on 9. Where do I change it?

 

[baditude_df]

Under AI Overclock Tuner

CPU Ratio Setting

 

[SkiBum1207]

Hey sorry bout the late reply, but heres the diagram for the p5k VANILLA



Sorry for the text being upside down, but it makes the resistor easier to find.
Do the Vdoop mod, the Vcore mod is only if the boards max Vcore is not enough