Advice on hitting 3.6GHZ

[DrewJW]

Hey all running a G0 Stepping Q6600 on stock intel air, in an Antec 900, I've currently got it on 3.4ghz ( 378mhz x9, ram 1:1:20 DDR2 907 ) with CPU Core at 1.44v and ram at 2.1v. This runs stable, when I up to 400mhz X9 and I drop the ram to 1:1:00 800mhz DDR2, and up the CPU Core to 1.47 the system will boot but fail a 3dMark06 test on the cpu section, and won't run any game stable for more than ten minutes.

Coretemp shows the idle temp to be around 40c ( helps having the Antec 900 ) and max temperature on the logs was 61c.

Running this on an Abit IP35 Pro, powered by a Corsair HX620. And the ram is Corsair XMS2 DDR 800mhz 2x2048MB.

Any tips on how I can obtain 3.6ghz?

[gwai lo]

Are those your BIOS vcores?

Have you tried bumping the NB voltage?

[DrewJW]

I haven't no sorry.

And what do you mean Bios Vcores? I have the ram at 2.1v and the cpu core to 1.47 when I tired for 3.6before running at 3.4now I have cpu core at 1.44v.

The rest are stock.

[gwai lo]

Well, the vcores set in the BIOS will not represent true voltages because of vdroop. If your IP35 Pro has massive vdroop that could affect your stability, but I think the vNB is slightly more important for now.

[DrewJW]

Well, the vcores set in the BIOS will not represent true voltages because of vdroop. If your IP35 Pro has massive vdroop that could affect your stability, but I think the vNB is slightly more important for now.

Right I am looking at the bios on my gaming system now.

CPU Core Voltage is at 1.44v

DDR2 Voltage is at 2.100v

CPU VTT 1.2V Voltage is at 1.20v

MCH 1.25V is at 1.25V

ICH 1.05V is at 1.05V

ICHIO is at 1.5v

DDR2 Reference Voltage is at 0%

CPU GTLREF0&2 67%

CPU GTLREF 1&3 67%

Don't know if most of that means anything to you?

[gwai lo]

Try giving the MCH a bump in voltage (that's your northbridge's voltage setting). Quads are pretty hard on the memory controller, so that's why a lot of them have been FSB limited. Bumping the voltage to the northbridge assists in pushing the FSB further.

You can really look up what they mean in your motherboard manual (save the last three things, those kind of require googling).

[DrewJW]

Ok I have the MCH on 1.29 now, I have the FSB at 400x9 = 3600mhz ram is at 1:1:00 DDR2-800 cpu core is at 1.47v.

Will boot it up now

[DrewJW]

Got a BSOD during testing :/

edit: core temp logs show a massive 74c high across the cores... so dropping back to 3.4 for now.

[khaydin]

Got a BSOD during testing :/

edit: core temp logs show a massive 74c high across the cores... so dropping back to 3.4 for now.

You may want to get an aftermarket cooler if you're going to overclock a quad like that. Try something like the thermalright Ultra 120 Extreme or the Xigmatech s1283. To go higher you need to figure out why it's unstable. First thing to do is take the cpu out of the equation and lower your cpu multiplier and set your fsb to 400 and see if it's stable. If it is stable your cpu didn't have enough voltage or its just getting way too hot. I'd go with it being too hot since you're using the stock cooler and pumping a lot of extra voltage into it.

[DrewJW]

I agree that should be my next move, next question is on my bios it reads 378mhz x9 = 3402mhz yet cpuz reads 2200mhz 378mhz x6, so does coretemp..

Strange?

[gwai lo]

First thing to do is take the cpu out of the equation and lower your cpu multiplier and set your fsb to 400 and see if it's stable. If it is stable your cpu didn't have enough voltage or its just getting way too hot. I'd go with it being too hot since you're using the stock cooler and pumping a lot of extra voltage into it.

Oh duh, how did I forget about this? Do this, I do it while I test for overclocks...yet I always fail to mention it to other people. -_-

As for the clocks looking funny, CPU-Z might be reading Speedstep's dropping of your multiplier.

[DrewJW]

Right I realized something important, I had NOT turned off the damn speedstep in the bios, and upon booting up at 3.4ghz without the speedstep I was pushing 54c IDLE!, I've dropped back to 3.2ghz which I know runs a 1.33v, and will sort out a much better cooler.

Whats easy to fit as in the likes of the stock one?

[gwai lo]

If you're doing heavy overclocking with a quad, you'll want something beefier than a simple drop in. With that said, the Xigmatek S1283 (as mentioned by khaydin) comes close to the TRUE in performance while maintain the push pin design and being cheaper (37 + 7 dollars for retention bracket). Most people buy the retention bracket because they don't like the push pin design, but the retention bracket isn't required. If you don't want to get it, then you can get it for 32 at mwave.

If you can't get your hands on the S1283, then I'd look at the Zerotherm NV120 and Noctua NH-U12P with the TRUE not really being in play because of its price.

edit: errr, disregard the prices, I forgot not everyone's from the US. -_-

[utdarkviper]

Also, you are only testing this in 3dMark06? This program tests the performance of your system, not it's stability. I would use ORTHOS or Prime95 to test your processor for an extended period of time (usually 8 hours is recommended). I don't think you are going to be as stable as you think because your processor will be getting hotter than with 3dMark06.

[DrewJW]

Well it performed a orthos test on 3.2 when i first hit this OC so I know I am stable, just been lazy since :p but I felt a run through Crysis to the snow level near the end was enough to give a system a test.

[utdarkviper]

I was just checking :p I'm jealous by the way as I have the B3 Q6600 and I barely got 3.0GHz with the NV120 cooler. I got stuck with a picky one unfortunately.

[Jason S. Global Moderator]

what does the '20' mean when you type 1:1:20??? :huh:

[utdarkviper]

lol I didn't even notice that... if he means 1 : 1.20 then that would equal 5:6 ratio. Maybe that's what he means?

[Jason S. Global Moderator]

i suggest he does a 1:1 ratio

[utdarkviper]

I did the math 907/756 ~=1.2 (RAM/FSBx2). He is indeed running 5:6 rather than 1:1. We suggest you change the ratio slightly Mr. OP :) I don't know how much difference it will make though... I just know it's better.

[Jason S. Global Moderator]

in my case, i changed from a 5:6 ratio to 1:1 and got a nice performance boost in my synthetic testing. i'll admit you will not notice a difference otherwise.

[DrewJW]

I plan to give it more of a OC from the 3.2GHZ it's on now, when I have fitted a superior cooler.

[+shift. MVC]

1:1 ratio is ideal, but in real world performance you won't notice a difference. In a benchmark you might see slight gains in running a 1:1 instead of 5:6.

Most of the people around here run a 1:1 though (just take a look @ the Overclocking Database thread)

https://www.neowin.net/forum/index.php?showtopic=642085

[Ci7]

1:1 ratio is ideal, but in real world performance you won't notice a difference. In a benchmark you might see slight gains in running a 1:1 instead of 5:6.

Most of the people around here run a 1:1 though (just take a look @ the Overclocking Database thread)

https://www.neowin.net/forum/index.php?showtopic=642085

just because of easier time with OCing in 1:1 as you dont have to worry that it is possibly that the ram crush the pc cause of overclocking

[+shift. MVC]

And a rule of thumb is to overclock CPU and get that stable before you work on RAM. So maybe loosen timings for now? And see what are the tightest timings you can run stable.