Q6600 & DS3L - How can I get that last .1Ghz

[althius]

My goal was always to get to 3.6Ghz on air, and I'm so close.

I can get to 3.5Ghz (9*389), but taking it to 400 gives me a BSOD about 5 mintues into the PRIME. I pushed by Vcore in the BIOS to 1.5V. I've given the FSB +.1 and the GMCH +.1.

Any suggestions from here? Should I push the Vcore some more? At 1.5VCore I'm getting 1.456 in CPUZ, and 1.392 under load.

Can I safely push it a little farther, or should I just be happy with what I got and chalk it up to a weaker chip?

My temps seem fine (around 30C idle/55C load at 1.5v/3.6Ghz)

Thanks in advance.

PS - I hope at this point you know I've read the hell out of all the guides. If I'm doing something stupid, or if I've forgotten anything, please know its just stupididy and not laziness.

PPS - I'm going to try and run 400*8 at 1.5V to see if that works - EDIT: It was stable... what does that mean? Also, if it matters I've got my timings relaxed to 5-5-5-15 and added .4v to the DDR2.

 

[CompuTronix]

althius, it means that you simply need to increase Vcore. If your ambient is near 22c, and the temp at 55c is the hottest Core (and it's correct), then since you have a TRU120E, you have ample headroom. Increase Vcore BIOS to 1.525, but don't be concerned about this value. In CPU-Z, Vcore Idle is important, however, Vcore Load is the primary consideration. The Delta between the two values is called Vdroop.

Below are my settings for comparison:

Vcore BIOS 1.475
Vcore Idle 1.456
Vcore Load 1.440
Vdroop 0.016

Q6600 G0 @ 3.6 - 12 Hrs Prime95 Stable (Small FFT's & Blend)
Ambient 22c - CPU 60c - Hottest Core 72c
ZeroTherm BTF90 - X23 Thermal Interface Material

Asus P5K Deluxe @ 450 x 8
Corsair 800 C4 @ 900 - 1:1 @ 4-4-4-12
4Gb - 4 x 1024 - Vdimm 2.3


Also, from the Temp Guide:

Section 12: Overclocking

Every CPU is unique in it's overclock potential, voltage tolerance, and thermal behavior. If the maximum stable overclock is known at 1.35 Vcore, then each increase of 0.05 volts will typically allow a stable increase of ~ 100 Mhz, and will result in a corresponding increase in CPU temperatures of 3 to 4c. Ambient and Vcore are the most dominant Variables affecting temperatures.


Please post your Ambient, and define your temperatures as well as which test was used.

Comp

 

[cnumartyr]

As Comp said, keep pumping vCore into it.

If it was stable at 400x8 that means you just need more vCore!

 

[althius]

Vcore BIOS 1.5625
Vcore Idle 1.504
Vcore Load 1.440
Vdroop 0.064

Q6600 G0 @ 3.6 - 1 Hr Prime95 Stable (Small FFT's)
Ambient 21c - CPU ??c - Hottest Core 58c
TRUE - X-2 Thermal Interface Material

GA-P35-DS3L @ 400 x 9
Patriot Extreme @ 800 - 1:1 @ 5-5-5-15 (rated 4-4-4-12)
4Gb - 2 x 2028 - Vdimm 2.2

-----------------------

Compu, thanks for taking the time to talk to someone about this for the 1,000,000th time. I have my CPU temp as a ??, becuase I can't get Speedfan to work properly under VISTA64. I believe that is what I need for that reading.

Also I small note on my V-Droop. I'm seeing a 'double droop', and by that I mean when I first start out, I have a Vcore Load of 1.456V (0.048 droop), and then after 2 to 3 mintues, I get the second droop down to 1.440V. I see this at any Vcore.

I really had to go a long way to get from 3.5Ghz to 3.6Ghz. Here is my info for 3.5Ghz

------------------------

Vcore BIOS 1.475
Vcore Idle 1.424
Vcore Load 1.360
Vdroop 0.064

Q6600 G0 @ 3.5 - 2Hr Prime95 Stable (Small FFT's)
Ambient 21c - CPU ??c - Hottest Core 50c
TRUE - X-2 Thermal Interface Material

GA-P35-DS3L @ 389 x 9
Patriot @ 800 - 1:1 @ 5-5-5-15
4Gb - 2 x 2028 - Vdimm 2.2

--------------------------

I'd love your feedback on this. I am almost of the opinion I'm just better running at 3.5Ghz, since I'm pushing so hard to get to 3.6Ghz (And I still haven't had a chance to run a long blend... I'll do it overnight tonight).

Also, when I go and 'retighten' my timings... will I need to adjust any Voltages? (I've already got +.4V for the RAM) Or do you think I'll need to leave them loose to maintain this stability?

Thank you very much for your time.

EDIT: Something else I just noticed. Under the BLEND test, my Vcore is different than the Small FFT test. It is also my fluid, bouncing between 1.472, 1.488, 1.456 altough not in that order. I haven't read about that anywhere else... is that known and normal?

 

[cnumartyr]

The bouncing is normal, different load demands on the power circuit.

If the RAM was stable at 800 MHz 4-4-4-12 than there is no reason you shouldn't be able to go back to stock factory settings.

 

[CompuTronix]

As cnumartyr said, since 3.6 = 9 x 400, which yields DDR2 800 at a perfect 1:1 ratio, there's no reason your RAM shouldn't function properly at it's rated specifications.

Concerning Vdroop, one of the primary reason that cnumartyr and I purchased these Asus P5K series motherboards is that they have a BIOS feature known as "Voltage Damper" which is also called "Line Load Compensation" on more recent versions. The purpose of this feature is specifically designed to minimize Vdroop, and as you can see, it works extremely well. This feature was also highlited in a Tom's Hardware motherboard article written this past fall.

Vcore is created on the motherboard by voltage regulators which are fed from a +12 volt rail in the power supply. Again, as cnumartyr said, it's typical for any power source to to sag or "droop" as load increases. This is quite normal and expected, although 64mV is a considerable amount of Vdroop.

Since Prime95 "Blend" represents a fluctuating Load due to the cyclic nature of each test series, the fluctuating Vcore which you describe is simply the motherboard's voltage regulator circuits responding to a fluctuating Load, and can be graphically observed using SpeedFan "Charts" to display Vcore.

Have you tried SpeedFan 4.34 BETA 38? I haven't recommended it in the Temp Guide because it requires users to jump through some registration hoops, which many users may find too complicated or discouraging, as the BETA expires every 30 days, and requires the latest download to overwrite the executable.

Also, what is the CFM rating of the 120mm fan on your TRU120E, are you testing at 100% RPM, and which utility and version are you using to display your temperature?

Comp

 

[cnumartyr]

I only disagree with you on one point Comp:

The 64mV vDroop.. I think this is normal all things considered. He has a 4 Phase vReg without heatsinks, not sure if it is analog or digital, however my Blood Iron had a 4 Phase Digital vReg and was (at 3.6 GHz) drooping as much as 120mV.

I agree 64mV is quite a bit, but considering the board and the power demands of the Quad, I would be quite happy.

The other reason I chose the P5K-E is it is one of the cheapest solutions out there with true 8 Phase vRegs.

 

[CompuTronix]

Understood cnumartyr. We only differ in perspective. If I recall correctly, the older Asus A8E boards were notorious for excessive Vdroop, and my old MSI K8N Neo 4 Platinum had an 88 millivolt Vdroop, which made it a bitchy balancing act to get my Opty 170 stable at 3.0Ghz, and resulted in a high Idle temperature.