3 GHz is very safe. I think I settled at 3.2 GHz with my G0 stepping, while it did do more with higher voltages. You can probably get more out of it, but the key thing is you must do stability testing with a burning program for a couple of hours.

Safely depends on 3 things
1: your systems cooling
2: your motherboards vrm
3: your psu

If you have good cooling, 8pin cpu cable and strong vrm along with a psu that can handle the load 4ghz would be my limit for daily usage.
But 3ghz is the highest oc most can hit and 3.2ghz on a g0

Jade Falcon wrote:

Safely depends on 3 things 1: your systems cooling 2: your motherboards vrm 3: your psu […]
Show full quote

Safely depends on 3 things
1: your systems cooling
2: your motherboards vrm
3: your psu

If you have good cooling, 8pin cpu cable and strong vrm along with a psu that can handle the load 4ghz would be my limit for daily usage.
But 3ghz is the highest oc most can hit and 3.2ghz on a g0

I posted the specs above. The PSU is an 80+ bronze 730w. The mosfets suck, as on most Asus boars (sans the expensive maximus / rampage boards, those have great fets). This motherboard has been repaired (twice) and now uses much better mosfets off a dead DFI lanparty. It won't read CPU voltage correctly because of this (it shows Vcore as 0.003v in bios) but it reads independent core voltages correctly, and I tested the CPU voltage with a voltmeter just to be safe. The cooler is a gigantic Tuniq Tower 120. It barely fits in the case and it does a great job at cooling the Q6600 at 3ghz - it gets 61C tops in World of Tanks after a 4-5 hour gaming session.

Now I found some modded LGA 771 Xeon CPUs for 10-15$ locally - I can spare that much. The guy is listing several 3.0, 3.2 X5482 and even a 3.4 GHz X5492 witch seem to be based off the Q9xxx series CPUs, as they have 12MB of cache. Would these clock any higher?

Short answer : 3,6-3,8GHz 24/7 should be doable with good enough cooling (you have it).
I clocked B3 revision of Kentsfield (in form of QX6700), at 3,62GHz and ~1,45V Vcore : LINK
Key thing to stability is Vcore + Temperature, and next in line is VRM temp.
Q6xx0's love Vcore (and 1,2-1,3V VTT), nothing else is needed for high OC (like PLL/GTL/etc.).
Basicly : If Vcore is high enough, it will work fine (at least all 4 of my samples are/were like that 😉).
Oh and don't worry about Vcore too much - you can't kill those CPU's with even 1,7V (I tried 😁).

With Q6700 G0 (same thing as Q6600, just with x10 multi available), I got 4GHz on P5B "Vannila" MB, proof : SuperPi 32M - LINK, Fire Strike : LINK
It was bench stable, but using 9 MOSFET's in total (on 3 non-heatsink'ed phases), is REALLY pushing what VRM can do 😀
Side notes : I wasn't using a case, and I had to actively cool VRM (duh 😁).

As to 45nm Xeon's :
They will clock higher at lower Vcore, but you need to mod socket (or buy one with LGA 775 holes cut already). 12MB of L2 is really nice to have, if you can get one cheaply - I say go for it.

I asked the same question about a year ago - refer Intel Core 2 Duo (E8400) versus Intel Core 2 Quad (Q6600).
I've been running my Q6600 (G0 stepping) at 3 GHz with no issues. Everything else is running at "stock" speeds.

That is a mean looking PC case 🤣

I have a Q6600 that'll do 3GHz at stock voltage. My Q6700 does 4GHz/1600FSB quite easily on a P45 motherboard. Both are G0 chips.
3GHz is absolutely nothing for those old quads.

MMaximus wrote:

That is a mean looking PC case 🤣

Yup - I've allways loved the Z11+. It's one of the better looking "mean" cases in my opinion 😀

agent_x007 wrote:

Short answer : 3,6-3,8GHz 24/7 should be doable with good enough cooling (you have it). I clocked B3 revision of Kentsfield (in […]
Show full quote

Short answer : 3,6-3,8GHz 24/7 should be doable with good enough cooling (you have it).
I clocked B3 revision of Kentsfield (in form of QX6700), at 3,62GHz and ~1,45V Vcore : LINK
Key thing to stability is Vcore + Temperature, and next in line is VRM temp.
Q6xx0's love Vcore (and 1,2-1,3V VTT), nothing else is needed for high OC (like PLL/GTL/etc.).
Basicly : If Vcore is high enough, it will work fine (at least all 4 of my samples are/were like that 😉).
Oh and don't worry about Vcore too much - you can't kill those CPU's with even 1,7V (I tried 😁).

With Q6700 G0 (same thing as Q6600, just with x10 multi available), I got 4GHz on P5B "Vannila" MB, proof : SuperPi 32M - LINK, Fire Strike : LINK
It was bench stable, but using 9 MOSFET's in total (on 3 non-heatsink'ed phases), is REALLY pushing what VRM can do 😀
Side notes : I wasn't using a case, and I had to actively cool VRM (duh 😁).

As to 45nm Xeon's :
They will clock higher at lower Vcore, but you need to mod socket (or buy one with LGA 775 holes cut already). 12MB of L2 is really nice to have, if you can get one cheaply - I say go for it.

Thanks for the tips! I'll have a go at 366x9 @ 1.35v, and then try 400x9 and see what voltage it needs at that speed and how hot it gets.

Thanks everyone for your input! I'll report back after some experimentation.

I have a Xeon E5450 running at 3600MHz. (FSB 1600) @ 1.25v.
Probably the cpu can go higher but it is already very hot!!

Case: Fractal Design Define R4 Arctic
Cooler: Thermalright HR-01 Macho

Arctic wrote:

I have a Xeon E5450 running at 3600MHz. (FSB 1600) @ 1.25v. Probably the cpu can go higher but it is already very hot!! […]
Show full quote

I have a Xeon E5450 running at 3600MHz. (FSB 1600) @ 1.25v.
Probably the cpu can go higher but it is already very hot!!

Case: Fractal Design Define R4 Arctic
Cooler: Thermalright HR-01 Macho

I recommend checking Tj. Max setting (Xeon E54x0 have 85C not 100C, source : LINK) or HR-01 mounting system (because HR-01 and Vcore 1.25V cannot get hot, or are you passive cooling that bad boy 😁).

@agent_x007
So all my tools are reporting the real temperatures + 15°C ?

Thanks a lot! I set the Tj max offset in the core temp ini to -15 on all four cores and now I supposedly have much lower temperatures.
I am running some prime95 to test it now for the "real" temperatures.

The Thermalright Macho is by far the best air cooler I've ever used. The only downside is that it's MASSIVE.

Arctic wrote:

@agent_x007
So all my tools are reporting the real temperatures + 15°C ?

No... well, not all.
Each core temperature is calculated by subtracting sensor value from tj max.
Problem is that value (tj max), can be manipulated.
AIDA64 can display 85C as tjmax, real temp can show 90C, and Coretemp : 100C.
Because of that, they will show different temperatute values across all cores.

Only "CPU" sensor is actually measuring temperature.

Source : LINK

The q6600, i've used that one for years.. bought it brand new when it was just out and straight out of the box i ran it at 3ghz for a while. Then i bought a Tuniq 120 and clocked that bay to 3,6ghz with 1,45 v. Perfectly stable. That was the B3 stepping, later i went for the G0 stepping and that one could do 4ghz but it wasn't perfectly stable with 1,5v and i didn't want to go any higher voltage wise.

All this was done on an Asus X48 Rampage Formula.

So i settled with the G0 again at 3,6ghz. It could do it with 1,35-1,4v. Nice cpu and i had it a long time. Temps were okay, 100% loaded at 4 cores did 68 degrees celcius, tested with OCCT.

if its a GO, set voltage to 1.45, and set FSB to 1600, havn't meet a Q6600 GO that can't do 3.6

Just keep an eye on those vrm temps and it should be bullet proof like the old tier 9 IS 4 wink wink 😉

candle_86 wrote:

.... havn't meet a Q6600 GO that can't do 3.6

Then consider youself lucky 🤣
http://hwbot.org/submission/2372173_havli_cin … 600_3.63_points
and back in the day when C2Q was still relatively new I had another one which did 3.5 GHz top at 1.5V.

havli wrote:

Then consider youself lucky :lol: http://hwbot.org/submission/2372173_havli_cin … 600_3.63_points and back in the day when C2Q […]
Show full quote

candle_86 wrote:

.... havn't meet a Q6600 GO that can't do 3.6

Then consider youself lucky 🤣
http://hwbot.org/submission/2372173_havli_cin … 600_3.63_points
and back in the day when C2Q was still relatively new I had another one which did 3.5 GHz top at 1.5V.

Isn't P5WDG2 kinda old for pushing Quad Core's to 1600MHz FSB (which is needed for 3,6GHz) ?
Have you tried that CPU in something newer (like P965 or P35) ?

Old... maybe but this board can do 450 MHz FSB with dual cores, so 400 MHz @ C2Q should be doable as well. Anyway I think GA-P35-DS3 wasn't any better running this particular Q6600 last time i tried.
The other Q6600 (3.5GHz one) was running X38 Maximus Formula board. Actually most of my 775 CPUs are sub-average overclockers... except maybe 3.8 GHz E6600. Just a bad luck 😵