Seeking a second opinion overclocking some dated hardware

[error8]

Originally posted by: Shaq

Originally posted by: error8

Who says you can only go 0.2 V over your VID? You can go as far as 1.55V, as Intel states for a 65 nm chip, as long as you can keep it cool enough.

I meant that as a rule of thumb. If you have a 1.2 VID CPU you won't be able to go over 1.5v on air without it being too hot. .2v is just a rough estimate for overclocking on air.

Well, even if the VID would be at 1.35V, the chip will still heat up like crazy at more then 1.5V, on air. In my opinion, as long as you can keep it cold enough, even 1.55V shouldn't be considered high, for a 65 nm chip, no matter the VID it has.

 

[imported_Shaq]

Originally posted by: error8

Originally posted by: Shaq

Originally posted by: error8

Who says you can only go 0.2 V over your VID? You can go as far as 1.55V, as Intel states for a 65 nm chip, as long as you can keep it cool enough.

I meant that as a rule of thumb. If you have a 1.2 VID CPU you won't be able to go over 1.5v on air without it being too hot. .2v is just a rough estimate for overclocking on air.

Well, even if the VID would be at 1.35V, the chip will still heat up like crazy at more then 1.5V, on air. In my opinion, as long as you can keep it cold enough, even 1.55V shouldn't be considered high, for a 65 nm chip, no matter the VID it has.

In other words...take a 1.2v and a 1.35v VID CPU and put them both at 1.5v and 3.2Ghz. Do you think they would both have the same temps? All else being equal..same cooler, TIM, case, all other voltages the same etc.

 

[error8]

Originally posted by: Shaq

In other words...take a 1.2v and a 1.35v VID CPU and put them both at 1.5v and 3.2Ghz. Do you think they would both have the same temps? All else being equal..same cooler, TIM, case, all other voltages the same etc.

This is what I'm thinking, but I don't have how to test this theory though. It's the same chip after all, the difference is that one is stable at 1.2V and the other one at 1.35V. Once you up the voltage, they'll both heat up the same.

 

[BonzaiDuck]

Originally posted by: firewolfsm
You should stick to a more modest OC and you'll be able to lower your voltages dramatically.

He's right. The relationships for the general Intel CPU technology between voltage, temperature and "scalability" in over-clocking involves ascending, exponential curves or patterns. Each family of chips -- differentiated by stepping as well -- has their own pattern or profile. You eventually hit a point where increased speed requires progressively disproportionate increases in voltage (and therefore heat dissipation). And your choice of over-clocks should be disciplined by voltages that do not much stray above the Intel spec "safe-voltage" upper boundary. It's been my experience that when you have approached this boundary anyway, you're already getting less and less over-clock mileage in terms of speed for successive voltage increases.

 

[Blitz KriegeR]

Originally posted by error8
Just set the Tjunction manually and see what happens. Idle won't go bellow ambient for god sake and even if it does, idle is not something that you want to care about. You are only interested in full load temps. My idle temps now are at 33 C with Tjunction set at 90C. You haven't even tried it and your so scared about it. Tjunction set at 90C will give you the real temperature, 10 C lower then what you're having now, with it at 100C. If intel states that the Tjuncion value is 90C, why would you use another wrong Tjunction value????

I have set it manually and checked the results. I'm not afraid of anything; it's just that as much as I'd like to believe that my CPU is idling 28-27-24-22 at 3.2Ghz 1.5v Vcore in a room that is 26'C ambient, It's just *not possible*; it's against the laws of physics. An air cooler *cannot* cool below the ambient temp of the air it's throwing about.

Yes, you're absolutely right; it's the load temps that matter. However, if I'm getting idle readings that are against the laws of physics, thereby we must assume they are inaccurate, it is logical to assume that the load temps generated under the same circumstances are also inaccurate. In either case, I'd rather have the temps being reported inaccurately high (curbing my OC) then inaccurately low resulting in me unknowingly pushing my chip above safe limits.

I'll be keeping my TJ max at 100'C. Oh and I've read about a few other forums regarding that spec sheet you posted, and a lot of folks seem to be finding fault with it the same way I am. (Their chips running below ambient on air cooling using the TJ max in the spec sheet; impossible!)

Originally posted by: BonzaiDuck

Originally posted by: firewolfsm
You should stick to a more modest OC and you'll be able to lower your voltages dramatically.

He's right. The relationships for the general Intel CPU technology between voltage, temperature and "scalability" in over-clocking involves ascending, exponential curves or patterns. Each family of chips -- differentiated by stepping as well -- has their own pattern or profile. You eventually hit a point where increased speed requires progressively disproportionate increases in voltage (and therefore heat dissipation). And your choice of over-clocks should be disciplined by voltages that do not much stray above the Intel spec "safe-voltage" upper boundary. It's been my experience that when you have approached this boundary anyway, you're already getting less and less over-clock mileage in terms of speed for successive voltage increases.

I'm aware of this. Just wanted to find the upper-end of what my chip is capable of and bench it. I plan on dropping to 3.0 or 2.8 for day-to-day use. Hell a 200 mV increase was required to boost my CPU from 2.8Ghz (1.3v) to 3.2Ghz (1.5v). I think that's enough of an unreasonable jolt right there.

 

[imported_Shaq]

Originally posted by: Blitz KriegeR

Originally posted by error8
Just set the Tjunction manually and see what happens. Idle won't go bellow ambient for god sake and even if it does, idle is not something that you want to care about. You are only interested in full load temps. My idle temps now are at 33 C with Tjunction set at 90C. You haven't even tried it and your so scared about it. Tjunction set at 90C will give you the real temperature, 10 C lower then what you're having now, with it at 100C. If intel states that the Tjuncion value is 90C, why would you use another wrong Tjunction value????

I have set it manually and checked the results. I'm not afraid of anything; it's just that as much as I'd like to believe that my CPU is idling 28-27-24-22 at 3.2Ghz 1.5v Vcore in a room that is 26'C ambient, It's just *not possible*; it's against the laws of physics. An air cooler *cannot* cool below the ambient temp of the air it's throwing about.

Yes, you're absolutely right; it's the load temps that matter. However, if I'm getting idle readings that are against the laws of physics, thereby we must assume they are inaccurate, it is logical to assume that the load temps generated under the same circumstances are also inaccurate. In either case, I'd rather have the temps being reported inaccurately high (curbing my OC) then inaccurately low resulting in me unknowingly pushing my chip above safe limits.

I'll be keeping my TJ max at 100'C. Oh and I've read about a few other forums regarding that spec sheet you posted, and a lot of folks seem to be finding fault with it the same way I am. (Their chips running below ambient on air cooling using the TJ max in the spec sheet; impossible!)

Originally posted by: BonzaiDuck

Originally posted by: firewolfsm
You should stick to a more modest OC and you'll be able to lower your voltages dramatically.

He's right. The relationships for the general Intel CPU technology between voltage, temperature and "scalability" in over-clocking involves ascending, exponential curves or patterns. Each family of chips -- differentiated by stepping as well -- has their own pattern or profile. You eventually hit a point where increased speed requires progressively disproportionate increases in voltage (and therefore heat dissipation). And your choice of over-clocks should be disciplined by voltages that do not much stray above the Intel spec "safe-voltage" upper boundary. It's been my experience that when you have approached this boundary anyway, you're already getting less and less over-clock mileage in terms of speed for successive voltage increases.

I'm aware of this. Just wanted to find the upper-end of what my chip is capable of and bench it. I plan on dropping to 3.0 or 2.8 for day-to-day use. Hell a 200 mV increase was required to boost my CPU from 2.8Ghz (1.3v) to 3.2Ghz (1.5v). I think that's enough of an unreasonable jolt right there.

awww come on. Do you want that CPU to last forever? lol I had mine at 1.54v all the time for a year and it is still okay. It was at 1.4v for 6 months before that. If something were to happen to it you'll have a reason to upgrade.

 

[Idontcare]

Originally posted by: Blitz KriegeR

Originally posted by error8
Just set the Tjunction manually and see what happens. Idle won't go bellow ambient for god sake and even if it does, idle is not something that you want to care about. You are only interested in full load temps. My idle temps now are at 33 C with Tjunction set at 90C. You haven't even tried it and your so scared about it. Tjunction set at 90C will give you the real temperature, 10 C lower then what you're having now, with it at 100C. If intel states that the Tjuncion value is 90C, why would you use another wrong Tjunction value????

I have set it manually and checked the results. I'm not afraid of anything; it's just that as much as I'd like to believe that my CPU is idling 28-27-24-22 at 3.2Ghz 1.5v Vcore in a room that is 26'C ambient, It's just *not possible*; it's against the laws of physics. An air cooler *cannot* cool below the ambient temp of the air it's throwing about.

Yes, you're absolutely right; it's the load temps that matter. However, if I'm getting idle readings that are against the laws of physics, thereby we must assume they are inaccurate, it is logical to assume that the load temps generated under the same circumstances are also inaccurate. In either case, I'd rather have the temps being reported inaccurately high (curbing my OC) then inaccurately low resulting in me unknowingly pushing my chip above safe limits.

I'll be keeping my TJ max at 100'C. Oh and I've read about a few other forums regarding that spec sheet you posted, and a lot of folks seem to be finding fault with it the same way I am. (Their chips running below ambient on air cooling using the TJ max in the spec sheet; impossible!)

This is where you have to decide who knows more about their chips, Intel or a bunch of random people on the internet.

Intel's documents state G0 TJmax is 90C. End of game.

Now why is your idle temp reported by some program below ambient when you use the correct TJmax? Because of well known and well document error between the temperature reported by the CPU's DTS versus actual CPU temperature for any temperature below TJmax.

CPU's DTS is only calibrated to correctly report CPU temperature when the CPU temperature is at TJmax. When your CPU's actual temperature is below TJmax the DTS will erroneously report a falsely higher or falsely lower CPU temp.

http://www.techpowerup.com/realtemp/docs.php

As the magnitude of the error decreases as the actual CPU temperature increases, using the correct TJmax will enable your software temperature programs to report your loaded temps more accurately (less error) versus you setting your TJmax to some fictitious (wrong) value just so your idle temps "look" right.

TJmax for G0 Q6600 is 90C. CPU actual temp cannot be below ambient temps if solely air-cooled by the ambient temperature air. If CPU temperature is reported to be below ambient error then you have observed some of the unavoidable error in the DTS measurement itself. Do not try and remove this error by recalibrating the interpretation of TJmax by your temperature measurement program.

 

[Idontcare]

Originally posted by: BonzaiDuck

Originally posted by: firewolfsm
You should stick to a more modest OC and you'll be able to lower your voltages dramatically.

He's right. The relationships for the general Intel CPU technology between voltage, temperature and "scalability" in over-clocking involves ascending, exponential curves or patterns. Each family of chips -- differentiated by stepping as well -- has their own pattern or profile. You eventually hit a point where increased speed requires progressively disproportionate increases in voltage (and therefore heat dissipation). And your choice of over-clocks should be disciplined by voltages that do not much stray above the Intel spec "safe-voltage" upper boundary. It's been my experience that when you have approached this boundary anyway, you're already getting less and less over-clock mileage in terms of speed for successive voltage increases.

The old-school textbooks tell you its quadratic, but we had a thread here a while back where we were chiding some website (forget which) about their claiming it was actually a cubic dependence. So we generated the data and lo and behold the website was right, with these modern IC's we've cross a threshold where now the power consumption dependence on Vcc is a cubic function and no longer a simple quadratic.

Vcore versus Power Consumption at fixed Clockspeed

None of that speaks to stability though, for the data there its all stable. For minimum stable Vcc at a given clockspeed you can see the dependence is quite tight too, a quadratic for my chip.

Minimum Vcore versus Clockspeed at fixed Temperature (phase cooling)

 

[Shmee]

I happen to have a pretty good Q6600, got it to 3.4 Ghz on my 780i FTW (yes, an nvidia board) but its VID was only 1.25. Trouble is, when running linx, it got up 80 C or more under my tuniq. Ill try it with my new Darknight soon, see if i can get it higher.

Of course, this is my second computer

 

[error8]

Originally posted by: Shaq
awww come on. Do you want that CPU to last forever? lol I had mine at 1.54v all the time for a year and it is still okay. It was at 1.4v for 6 months before that. If something were to happen to it you'll have a reason to upgrade.

Yes, I agree with you on this. :laugh:

 

[Assimilator1]

Originally posted by: Blitz KriegeR
.........................

I'm a bit confused about what you said about the PPL voltage? How can I "drop to 1 above the base manual setting"? It's 1.5v by default, up to 3.0v in 20 mV increments. Are you suggesting I increase it one step to 1.52v? Not sure how that will help cut heat output; won't that just do the opposite?

Anyway, off to drop my tRD down to 8 and rerun all my tests. It's doing wonders for memory read speeds and access latency.

Got my tRD down to 7 .

Re voltages, read at least my 1st post in my thread I linked, although it might not applie to your mbrd. You say you've manually(?) set CPU PLL to 1.5v?, does the h/w mon confirm this?
On P5Q Pros auto CPU PLL & manual setting 1.5v actually gives 1.82v over 333 FSB!:Q, manually setting 1.52 gives 1.545v (measured with a DMM). Hence 1 higher setting gives lower voltage! lol

Originally posted by: Shaq

Originally posted by: VirtualLarry
Wow, those later date Q6600s just don't clock like they used to. Mine is an older one, and I got to 3.6Ghz at 1.325v (IIRC). LLC enabled, on a DFI X48 board. Problem is temps, not voltage, for me. Temps get to like 85C under OCCT: Linpack.

It depends on your VID. Mine was 1.325. It is the highest or second highest value. You should be able to go .2 volts over your VID when you OC. You probably had a 1.2 VID chip, or less, which is quite good. Your chip would burn up before you got to 1.5 volts. lol

Not according to Intel - Q6600 G0, safe to 1.5v, temps allowing. (saw your qualifying statement in a latter post but left the link in for info ).

 

[error8]

Originally posted by: Assimilator1

Not according to Intel - Q6600 G0, safe to 1.5v, temps allowing. (saw your qualifying statement in a latter post but left the link in for info ).

According to this it's at 1.55 V. I'm feeling happy at 1.52 V in bios and 1.45 V under load with my q6700.

 

[Idontcare]

Originally posted by: error8

Originally posted by: Assimilator1

Not according to Intel - Q6600 G0, safe to 1.5v, temps allowing. (saw your qualifying statement in a latter post but left the link in for info ).

According to this it's at 1.55 V. I'm feeling happy at 1.52 V in bios and 1.45 V under load with my q6700.

Yep, the max VID is 1.5V, that doesn't mean the CPU cannot operate on higher Vcc and still be OK and within warranty/spec.

Max VID just means Intel won't sell a G0 with a default Vcc requirement under load that exceeds 1.50V. VID range != spec limits.

 

[imported_Shaq]

Originally posted by: Idontcare

Originally posted by: error8

Originally posted by: Assimilator1

Not according to Intel - Q6600 G0, safe to 1.5v, temps allowing. (saw your qualifying statement in a latter post but left the link in for info ).

According to this it's at 1.55 V. I'm feeling happy at 1.52 V in bios and 1.45 V under load with my q6700.

Yep, the max VID is 1.5V, that doesn't mean the CPU cannot operate on higher Vcc and still be OK and within warranty/spec.

Max VID just means Intel won't sell a G0 with a default Vcc requirement under load that exceeds 1.50V. VID range != spec limits.

Does that mean i7 is okay up to 1.6v? Check out page 15 on the link.

http://download.intel.com/desi...or/datashts/320834.pdf

 

[Assimilator1]

Originally posted by: error8

Originally posted by: Assimilator1

Not according to Intel - Q6600 G0, safe to 1.5v, temps allowing. (saw your qualifying statement in a latter post but left the link in for info ).

According to this it's at 1.55 V. I'm feeling happy at 1.52 V in bios and 1.45 V under load with my q6700.

Hmm, wierd, don't know which is max then but I'd say 1.5v is more than enough anyway . What temps are you getting at that voltage?

 

[Idontcare]

Originally posted by: Shaq

Originally posted by: Idontcare

Originally posted by: error8

Originally posted by: Assimilator1

Not according to Intel - Q6600 G0, safe to 1.5v, temps allowing. (saw your qualifying statement in a latter post but left the link in for info ).

According to this it's at 1.55 V. I'm feeling happy at 1.52 V in bios and 1.45 V under load with my q6700.

Yep, the max VID is 1.5V, that doesn't mean the CPU cannot operate on higher Vcc and still be OK and within warranty/spec.

Max VID just means Intel won't sell a G0 with a default Vcc requirement under load that exceeds 1.50V. VID range != spec limits.

Does that mean i7 is okay up to 1.6v? Check out page 15 on the link.

http://download.intel.com/desi...or/datashts/320834.pdf

No, that simply lists the addressable range of VID settings which Intel has programmed, it doesn't mean it actually intends to use the entire range or that the processor line is compatible with the entire range of values represented in that table.

However the max Vcc is stated in that document, top of page 22, and it is 1.55V for the i7.

Further, for the i7 the VID range is 0.8V to 1.375V...see top of page 23.

And in standard operating conditions the maximum allowed Vcc overshoot is VID + 0.05V (happens when the processor goes from loaded to unloaded condition)...see page 29.

 

[error8]

So the i7 chip can be pushed to 1.55 V? I like that on a cpu.

 

[imported_Shaq]

Originally posted by: error8
So the i7 chip can be pushed to 1.55 V? I like that on a cpu.

LOL...I'm having bad thoughts now. Maybe I can get 4Ghz.

Originally posted by: Shaq

Originally posted by: Idontcare

Originally posted by: error8

Originally posted by: Assimilator1

Not according to Intel - Q6600 G0, safe to 1.5v, temps allowing. (saw your qualifying statement in a latter post but left the link in for info ).

According to this it's at 1.55 V. I'm feeling happy at 1.52 V in bios and 1.45 V under load with my q6700.

Yep, the max VID is 1.5V, that doesn't mean the CPU cannot operate on higher Vcc and still be OK and within warranty/spec.

Max VID just means Intel won't sell a G0 with a default Vcc requirement under load that exceeds 1.50V. VID range != spec limits.

Does that mean i7 is okay up to 1.6v? Check out page 15 on the link.

http://download.intel.com/desi...or/datashts/320834.pdf

No, that simply lists the addressable range of VID settings which Intel has programmed, it doesn't mean it actually intends to use the entire range or that the processor line is compatible with the entire range of values represented in that table.

However the max Vcc is stated in that document, top of page 22, and it is 1.55V for the i7.

Further, for the i7 the VID range is 0.8V to 1.375V...see top of page 23.

And in standard operating conditions the maximum allowed Vcc overshoot is VID + 0.05V (happens when the processor goes from loaded to unloaded condition)...see page 29.

And it looks like the max memory voltage is 1.875v and not 1.65v that you hear all the time?

 

[Blitz KriegeR]

Originally posted by: Assimilator1
Re voltages, read at least my 1st post in my thread I linked, although it might not applie to your mbrd. You say you've manually(?) set CPU PLL to 1.5v?, does the h/w mon confirm this?
On P5Q Pros auto CPU PLL & manual setting 1.5v actually gives 1.82v over 333 FSB!:Q, manually setting 1.52 gives 1.545v (measured with a DMM). Hence 1 higher setting gives lower voltage! lol

Hmm read your post, and my board doesn't seem to do that. It under-volts the CPU slightly (about 4% it seems, more under load with vdroop obviously) an over-volts everything else. These other voltages however do not increase/decrease with my overclock, they are statically 20-85 mV above whatever manual setting is applied to them. Here are the settings; I think I've finally ironed out every last rinkle!

3.2Ghz (400x8), 333Mhz strap to NB running 6/5 divider, DDR2-960 @ 5-5-5-15-2N tRD 7. DRAM throughput is 8227MB/s Read, 7259MB/s Write, 7935MB/s Copy, 61.8ns Latency.

CPU Vcore: 1.5125v BIOS, 1.456v idle, 1.405v load
*CPU PLL: 1.5v BIOS, 1.526v actual
*{As you suggested, I tried increasing the setting one step to 1.52v and checked the result; 1.552v actual}
FSB Term: 1.20v, 1.22v actual
MCH: 1.25v, 1.245v actual
ICH: 1.05v, 1.056v actual
ICH PLL: 1.5v, 1.507v actual
DRAM: 2.00v, 2.036v actual
LLC: Disabled
C1E: Enabled

Temps:
{CPU} 66-65-59-57'C max load OCCT / Prime95, 76-76-68-67'C LinX / Intel Burn Test
{MCH} 40'C max load

Random thoughts:
Try as I may I can't get the bloody tRD down to 6! I've boosted MCH to 1.55v and DRAM to 2.15v and the board still fails to post at tRD6.
Tried to change to 5/6 divider on 266Mhz FSB strap; works ok but is flaky at times. Might be able to stabilize with more effort but I'm happy for the time being.
Other than CPU Vcore which is through the roof, I've actually been able to keep all the voltages *surprising* low and my rig is still super stable. Longest uptime since last bios tweak; 3d 14hrs.
So, after my adventure, I'm thinking of picking up a 9550 and doing it all over again! ^.^

 

[BonzaiDuck]

Blitz Krieger --

Is there an X48 Rampage BIOS that precedes the release of 45nm Penryn processors -- perhaps dated before end of 2007?

If not, then -- yes -- you should sell the Q6600 and pick up a Yorkie quad (if quad is your desire).

The BIOS consideration may or may not apply in your situation. But with motherboards preceding release of 1333 FSB cores like E6850 (and I think QX6850), the Q6600 cores ran better on the BIOS version that preceded a fix-list including "addresses improved over-clocking with CPUs having 1333 FSB."

 

[error8]

Blitz KriegeR, check the FSB termination voltage. Increasing that might give you better stability. For example my Q6700 is unstable at 3.6 ghz and FSB termination ( VTT ) voltage under 1.4 V. Once I increased that at 1.4 V, Linpack passes all tests. 1.2 V that you're using now it's very, very small for a quad.
Oh and I'm using 1.52 V in bios for 3.6 ghz, which gives 1.4 V idle ( EIST on) and 1.43 V under load. I have a massive 0.1 V vdroop.

 

[Blitz KriegeR]

Bonzai: Not sure now. Anand's review is from Jan 25th, and I bought the board in early March. Thinking back I could have sworn the original BIOS was dated with something like 12/29/07, but now the oldest BIOS on ASUS's site is 04/14/08. Even the release BIOS on this board supported 400Mhz FSB CPUs out of the box though.

error8: I'm not sure what I would achieve by increasing VTT, since as I said the system has been stable at these settings for up to 3 days. Passes 4+ hours of OCCT/Prime and passed LinX 15 passes. The only thing I can think of is if increasing the VTT will allow me to decrease my Vcore slightly, but I doubt it.
Random thought: 1.52v BIOS = 1.4v idle and 1.43v load? Isn't load voltage always lower than idle since as Amps increase Voltage decreases slightly to compensate? Sounds like you *must* have LLC enabled, or you've got those two figured mixed up?

Thanks for all of the thoughts / comments / supports and criticisms. Much appreciated. :beer:

 

[error8]

Originally posted by: Blitz KriegeR


error8: I'm not sure what I would achieve by increasing VTT, since as I said the system has been stable at these settings for up to 3 days. Passes 4+ hours of OCCT/Prime and passed LinX 15 passes. The only thing I can think of is if increasing the VTT will allow me to decrease my Vcore slightly, but I doubt it.

Yes, it can probably let you use less vcore on the current clock, or higher clocks for the current vcore. Either way, increasing VTT gives you only benefits and it's recommended.

Originally posted by: Blitz KriegeR
Random thought: 1.52v BIOS = 1.4v idle and 1.43v load? Isn't load voltage always lower than idle since as Amps increase Voltage decreases slightly to compensate? Sounds like you *must* have LLC enabled, or you've got those two figured mixed up?

Like I've said: I have EIST enabled from bios, which underclocks and undervolts my cpu in idle,with the power plan profile in Vista put on "balanced".
So, I'm ending up with a higher vcore under load, then what I have it in idle. Weird, huh? Of course that when the cpu is slightly used, I have spikes as high as 1.48 V. But when it does absolutely nothing, it's at 1.4 V.

 

[California Roll]

I'm late to this thread, but I have a late model Q6600 (1.325 vid) stable at 334x9 at 1.25V in bios (Z-cpu shows voltage lower in windows). Ram is at stock settings at 1.8v.

I could not get it stable at 3.2ghz, no matter what I tried, 400x8, 356x9, etc. Actually, I tried only up to 1.4v in bios. It just got too dang hot and noisy. 3.2g might be doable for me with a better cooler, but I just decided to settle on 3.0.

 

[imported_Shaq]

Originally posted by: California Roll
I'm late to this thread, but I have a late model Q6600 (1.325 vid) stable at 334x9 at 1.25V in bios (Z-cpu shows voltage lower in windows). Ram is at stock settings at 1.8v.

I could not get it stable at 3.2ghz, no matter what I tried, 400x8, 356x9, etc. Actually, I tried only up to 1.4v in bios. It just got too dang hot and noisy. 3.2g might be doable for me with a better cooler, but I just decided to settle on 3.0.

You're right. I had to go over 1.4v, 1.54v to get 3.3 stable. It was only hot when checking for stability. Stayed under 65C for everything else.