Just got a Q6700

[error8]

But I'm not even hitting 3.5 ghz Linpack stable. I have a VID of 1.23 V.
I hit 74 C full load under Linpack,
Vcore is at 1.41 V in bios
Vtt 1.37 V ( can I go higher? )
MCH 1.4 ghz
Ram is underclocked

I expected a 3.6 ghz out of this ah heck and with lower voltage then what I'm using. I wonder what am I missing here!? Does anyone has experience with a similar setup as mine?
Might be related with the fact that I have all my motherboard memory slots occupied?

 

[Idontcare]

Are your IHS and HSF lapped?

You running 10x350 for FSB with your 8GB of ram clocked synchronously at 750MHz speeds?

Step one is to find your ram clockspeed limit. Set your CPU's Vcc to its default VID and then manually set the multiplier to 6x (the lowest it can go). Set your FSB to the default for your chip (which is 267MHz, i.e. 1066MHz quad-pump). Then start changing ram multipliers and find out where your ram starts dying from too much clockspeed, raise Vdimm if you have to but only within your comfort limits.

Use memtest as well as Large FFT in prime95 to determine whether the ram is stable.

Step two is to find your NB/FSB clockspeed limit. Set your CPU's Vcc to its default VID and then manually set the multiplier to 6x (the lowest it can go). Set the ram multiplier to synch (some BIOS's call this 1:1, some all it 1:2, and others call it simply "synch")...basically for any given FSB clockspeed you want your ram to be clocked the lowest speed it can be for that FSB to ensure any instability you encounter is NB related and not ram or CPU related.

Now start pushing your FSB to the limits, overvolt the MCH and FSB/NB as needed as you push above 350MHz FSB. Test stability with Large FFT in prime95.

Now that you know whereabouts your ram will limit your OC, and whereabouts your NB/FSB will limit your OC, the third step is to find your max CPU OC.

Again set your ram to its lowest multiplier (1:1 or sync), set your CPU multiplier to 10x (the highest allowed for Q6700) and start pushing up that FSB and raising Vcore as needed to maintain stability (as well as bumping up Vdimm and/or NB voltages as needed based on your prior testing in steps one and two) and test stability using Small FFT in prime 95.

Once you have this data, then you can surmise whether your 3.5GHz OC is limited by CPU (Vcc and Temp limited), by NB/FSB (more active cooling), or the ram itself can't go higher (need Vdimm or better ram).

 

[alyarb]

or you can get a Q9550 and be done with it. i had a 6700 and was in your same situation. thank god some poor sap on ebay gave me $200 for it. I'm at 3.70 1.22 vcc. Did you want higher? I don't have a E0 step.

 

[error8]

Originally posted by: Idontcare
Are your IHS and HSF lapped?

You running 10x350 for FSB with your 8GB of ram clocked synchronously at 750MHz speeds?

Step one is to find your ram clockspeed limit. Set your CPU's Vcc to its default VID and then manually set the multiplier to 6x (the lowest it can go). Set your FSB to the default for your chip (which is 267MHz, i.e. 1066MHz quad-pump). Then start changing ram multipliers and find out where your ram starts dying from too much clockspeed, raise Vdimm if you have to but only within your comfort limits.

Use memtest as well as Large FFT in prime95 to determine whether the ram is stable.

Step two is to find your NB/FSB clockspeed limit. Set your CPU's Vcc to its default VID and then manually set the multiplier to 6x (the lowest it can go). Set the ram multiplier to synch (some BIOS's call this 1:1, some all it 1:2, and others call it simply "synch")...basically for any given FSB clockspeed you want your ram to be clocked the lowest speed it can be for that FSB to ensure any instability you encounter is NB related and not ram or CPU related.

Now start pushing your FSB to the limits, overvolt the MCH and FSB/NB as needed as you push above 350MHz FSB. Test stability with Large FFT in prime95.

Now that you know whereabouts your ram will limit your OC, and whereabouts your NB/FSB will limit your OC, the third step is to find your max CPU OC.

Again set your ram to its lowest multiplier (1:1 or sync), set your CPU multiplier to 10x (the highest allowed for Q6700) and start pushing up that FSB and raising Vcore as needed to maintain stability (as well as bumping up Vdimm and/or NB voltages as needed based on your prior testing in steps one and two) and test stability using Small FFT in prime 95.

Once you have this data, then you can surmise whether your 3.5GHz OC is limited by CPU (Vcc and Temp limited), by NB/FSB (more active cooling), or the ram itself can't go higher (need Vdimm or better ram).

Well I really don't want to overclock my memory, I just want to keep it around 800mhz as much as I can. If I have to run it bellow that, I don't mind. So I was really trying to skip this part since it takes years to have just one pass at memtest for all these 8gb.

So I'll probably try to see if it's the NB keeping me from overclocking. But from what I know, this particular mobo shouldn't have any problems with quad cores at 350-360 FSB. And the 1.23 VID I have for the cpu should give me a good overclock, isn't it?

And, Idontcare, don't you think LinX (Linpack) is a better stress tester then Prime95? It takes less time to find any weaknesses in the system.

 

[error8]

Oh and I don't have any lapping done. It's the Xigmatek S1283 with AS5 and nothing more.

 

[Idontcare]

Originally posted by: error8
And, Idontcare, don't you think LinX (Linpack) is a better stress tester then Prime95? It takes less time to find any weaknesses in the system.

I like to use Prime95 simply for the fact the large vs. small FFT tests so neatly deconvolve the underlying hardware which as the source of the instability.

I have nothing against linpack or OCCT or any of the other faster-to-BSOD stability testers, just haven't figured out a way to leverage them as much as I have Prime95 in generating the info I need to optimize an OC.

Definitely substitute your stability programs of choice in place of any I recommend if you have experience/comfort levels with making them do the job that needs to be done. (And share your secrets too so I'm not the only one left around here using Prime95 in 2015 :laugh

 

[error8]

I now seem to have reached some sort of stability ( didn't tested too much, but no more BSODs in Linpack ) at 3.4 ghz. Vcore is set at 1.42 V in bios, but I only have 1.33-1.36 V under load in Linpack, so that looks like a horrible vdrop for this mobo. Maybe this was the problem all along: I thought I was overvolting the cpu, when it's clearly that I wasn't.

 

[Rubycon]

Various encoders were my poison with 45nm yorkdale overclocking. I hit 4.4GHz on a QX9650 Prime/OCCT/Linpack stable for a week! Yet I had to back down to 4GHz to be artifact free on my encodes up to 3 min and 3.6GHz to be artifact free on batch jobs taking over six hours. CPU temps in the 60's under workloads with a machined flat TRUE with a 135CFM Adda screamer. :Q

EDIT:

RE: Vdroop...

Yep if it drops, droops, etc. that much you either have to overvolt higher (1.5+) OR use regulation if your board supports it. Regulation is preferred only because if the processor is idle most of the time that high voltage isn't needed nor helpful.

 

[error8]

Originally posted by: Rubycon
Regulation is preferred only because if the processor is idle most of the time that high voltage isn't needed nor helpful.

And how is that done?

 

[heyheybooboo]

You may want to run the OCCT3 vdroop to see if there are any serious swings, spikes or drops in the cpu voltage (beyond what you have identified) ....

 

[alyarb]

Originally posted by: error8

Originally posted by: Rubycon
Regulation is preferred only because if the processor is idle most of the time that high voltage isn't needed nor helpful.

And how is that done?

in gigabyte BIOSes it's referred to as load line calibration. look for something like that in the CPU voltage config area. it may be referred to as LLC.

 

[error8]

Originally posted by: heyheybooboo
You may want to run the OCCT3 vdroop to see if there are any serious swings, spikes or drops in the cpu voltage (beyond what you have identified) ....

It doesn't look like I have serious voltage swings. I can post an OCCT3 vcore picture if you'd like.

 

[error8]

Originally posted by: alyarb

Originally posted by: error8

Originally posted by: Rubycon
Regulation is preferred only because if the processor is idle most of the time that high voltage isn't needed nor helpful.

And how is that done?

in gigabyte BIOSes it's referred to as load line calibration. look for something like that in the CPU voltage config area. it may be referred to as LLC.

I don't have anything like that in my BIOS.

 

[error8]

I also have 10 C difference between the two pairs of cores. What is that all about?

 

[n7]

Originally posted by: error8
Vcore is set at 1.42 V in bios, but I only have 1.33-1.36 V under load in Linpack, so that looks like a horrible vdrop for this mobo. Maybe this was the problem all along: I thought I was overvolting the cpu, when it's clearly that I wasn't.

Bingo.

I needed around 1.4v for 3.5 GHz with my Q6700.
1.4v under load i mean.

With vdrop & droop taken into account, you'll need high vcore in the BIOS...

 

[Rubycon]

It's not abnormal to see differences in core temps. The lower the better and if they change after a remount than you know where the problem lies. If the difference stays the same it's internal and stays that way.

 

[error8]

Originally posted by: n7


I needed around 1.4v for 3.5 GHz with my Q6700.
1.4v under load i mean.

With vdrop & droop taken into account, you'll need high vcore in the BIOS...

Good to hear that. So high voltage is all this sucker needs. Too bad I'm already starting to approach 80 C.

screenshot

There I have 1.43 V in bios and 1.36 V under load. And the massive delta between the pairs of cores is also visible: 11 C!!! I know I have my cooler mounted right, since I'm using the backplate/screw system.

 

[error8]

OK, stable at 3.5 ghz.

How does this sound to you?
389X9 mhz
vcore:1.46 V in bios, between 1.37-1.39 V under load
VTT 1.35V
SB at 1.8V
NB at 1.45V
ram is underclocked

Temps are: 82,82,72,72 under Linpack
and 71,71,60,60 under Prime95

I'll probably add a 60mm fan on the NB heatsink, since it looks like it heats up like hell. But, does this seem "safe" to you? Can I run it 24/7 like this?
I would also want to use EIST on, to keep the clocks and voltage down under idle, but it's only using 10X multiplier so I can only keep it at 3.89ghz with my current FSB. Any idea about that?

 

[cubeless]

that's a bit warm... i have a very similar setup 3.5ghz q6700 under a tuniq and always try to keep c2's under 70c...

and the nb's definitely get hot when pushed, along with the vrm's... u need air blowing down on them directly... i like the antec 300 case for this...

 

[error8]

All five fans on my Antec are on high rpm. There couldn't be any better airflow then this. So do you think my temps are on the high side? What voltage are you using for 3.5ghz?

 

[Idontcare]

Originally posted by: error8
All five fans on my Antec are on high rpm. There couldn't be any better airflow then this. So do you think my temps are on the high side? What voltage are you using for 3.5ghz?

What TJmax are you using for the temp measurements? TJmax is 90C for Q6700.

 

[error8]

Really?I thought it was 100C.

 

[Idontcare]

See slide 15 of this official Intel document regarding DTS and TJmax.

From this thread.

Q6700 (and Q6600 TJmax is 90C.

 

[error8]

Thanks for that. Before I got it overclocked I did a google search and everyone was saying that 100 is the Tjunction for Q6XXX cpus. It looks that they were all wrong. And because I've set it in RealTemp, now I noticed that the last two cores were set at 95 Tjunction value and the first two were set at 100 C. So this is where my fantastic 10 C difference came from. Now it's only 5 C difference.

And with Tjunction set at 90 C, I now have a sexy 60 C maximum temperature in Prime95. Now I feel I can push it further.

 

[Tempered81]

So are you going to push it into the sky, until it lights on fire or whatever... or are you going to respect your 60C threshold?

Here's to catching your mobo on fire :beer: