Safe Overclock of a Intel Q6600

[Ext User(Matt)]

Hey guys.

I'm considering going for a Q6600 quad core CPU at the moment over an
E6850 (see a separate discussion topic on this :) ). What's been holding
me back over the Q6600 is, aside from the lower clock speed (which as
more applications become multi-threaded won't be an issue) is the lower
FSB speed it supports (1066 vs. 1333 MHz).

Since I'm not keen on overclocking a CPU (due to the shorter life and
the noise a good cooling system creates), is it possible to get a
motherboard and memory that fully supports 1333MHz, set this CPU's FSB
to 1333MHz and then just reduce the multipliers so its clock is running
close to 2.4GHz again?

Kind Regards,

Matt

[Ext User(Matt)]

> In real-world benchmarks, the quad-core 2.4GHz 6600 is maybe only
> 10-15% faster than the dual-core 6850 at 3GHz.

I still fancy getting hold of that 10%-15%, especially with the move ot
multi-threaded applications in the future.

You cannot change the
> multiplier of any Intel CPU except the "Extreme" versions, such as
> QX6xxx Series.

Fair point. Would an overclock to a 1333MHz FSB (taking it up to 3GHz)
for a Q6600 be feasible without a monstrous cooling system? I'm not at
all keen about shortening the life of the CPU.

> And you will pay a heavy price for them. 1.06GHz FSB vs
> 1.33GHz is only an increase of 25%, however, there is very little
> performance advantage unless you look very deeply due to memory speed
> and so on - what I would do is go with the E6850, and spend the extra
> cash saved on a good m\board supporting 1.33GHz FSB.

I'm certainly not discounting the E6850 at the moment, but if I can get
a quad core up to 3GHZ with a 1333MHz FSB that has 2 extra cores, then
I'm clearly onto a winner :)

Kind Regards,

Matt

[Ext User(Howard Goldstein)]

On Sat, 05 Jan 2008 23:56:24 +0000, Matt <m.p.boulton@durham.ac.uk> wrote:
: Fair point. Would an overclock to a 1333MHz FSB (taking it up to 3GHz)
: for a Q6600 be feasible without a monstrous cooling system? I'm not at
: all keen about shortening the life of the CPU.

With the right motherboard, yes, absolutely you can. You can even
cool it with a cheap Arctic Cooling Freezer 7 air cooler as I used to
do here in Florida when the ambient temp in my office was around 30C.
As for going over 3ghz, it seems a lot of folks can, I'm not stable
anywhere over 3ghz :(

[Ext User(~misfit~)]

Somewhere on teh intarweb "AdenOne" typed:
> On Jan 6, 5:18 am, Matt <m.p.boul...@durham.ac.uk> wrote:
>> Hey guys.
>>
>> I'm considering going for a Q6600 quad core CPU at the moment over an
>> E6850 (see a separate discussion topic on this :) ). What's been
>> holding me back over the Q6600 is, aside from the lower clock speed
>> (which as more applications become multi-threaded won't be an issue)
>> is the lower FSB speed it supports (1066 vs. 1333 MHz).
>>
>> Since I'm not keen on overclocking a CPU (due to the shorter life and
>> the noise a good cooling system creates), is it possible to get a
>> motherboard and memory that fully supports 1333MHz, set this CPU's
>> FSB to 1333MHz and then just reduce the multipliers so its clock is
>> running close to 2.4GHz again?
>>
>> Kind Regards,
>>
>> Matt
>
> In real-world benchmarks, the quad-core 2.4GHz 6600 is maybe only
> 10-15% faster than the dual-core 6850 at 3GHz. You cannot change the
> multiplier of any Intel CPU except the "Extreme" versions, such as
> QX6xxx Series.

Wrong! You can reduce them *down* within certain parameters with the right
motherboard. For instance, my E4500 can be set to anything between 6x and
11x. I have it set to 8 x 413MHz FSB, memory on 1:1 ratio. (That's 1.652GHz
FSB for those who buy into the "quad-pumped" thing) I have left Speed-Step
enabled so it still drops to 6x when it's not under load.

> And you will pay a heavy price for them. 1.06GHz FSB vs
> 1.33GHz is only an increase of 25%, however, there is very little
> performance advantage unless you look very deeply due to memory speed
> and so on - what I would do is go with the E6850, and spend the extra
> cash saved on a good m\board supporting 1.33GHz FSB.

The rest of this "info" is bogus as it's based on a false premise.
--
Shaun.

[Ext User(Fred)]

Matt wrote:
> Hey guys.
>
> I'm considering going for a Q6600 quad core CPU at the moment over an
> E6850 (see a separate discussion topic on this :) ). What's been
> holding me back over the Q6600 is, aside from the lower clock speed
> (which as more applications become multi-threaded won't be an issue)
> is the lower FSB speed it supports (1066 vs. 1333 MHz).
>
> Since I'm not keen on overclocking a CPU (due to the shorter life and
> the noise a good cooling system creates), is it possible to get a
> motherboard and memory that fully supports 1333MHz, set this CPU's FSB
> to 1333MHz and then just reduce the multipliers so its clock is
> running close to 2.4GHz again?

To answer the question in the title of your post.
The Q6600 GO stepping is a real easy overclocker.
The original heatsink and fan are more than adequate to run this cpu at
3.0GHZ although it will run hotter it won't be an issue as it will be well
within specs.
So IMHO 3.0GHz is a perfectly sahe overclock.

You could take it even higher but to do so it would be prudent to purchase
expensive RAM and aftermarket cooling.
Also the BIOS of some modern motherboards are designed to fascilitate
overclocking.

An example I recently assembled for a family member using a basic X38
chipset board.
Asus P5E motherboard
2.4GHz Q6600 cpu overclocked 25% to 3.0GHz
Four 1GB sticks of cheap A-DATA DDR2-800 SDRAM running at 5-5-5-18
timings.
The memory bus is running at 400 MHz

I am using the auto overclocking feature of the motherboard rather than the
manual settings.
With auto you just nominate a frequency to run the cpu front side bus at and
the BIOS takes care of the rest of the settings.
Now the Q6600 has a cpu multiplier of 9.0x so to run it at 3.0GHZ is just a
matter of setting cpu FSB at 334.
In auto overclock the BIOS takes care of the DRAM:FSB ratio and sets a ratio
of 12:10 running the RAM at it's maximum rated speed.

As a side note the core processors use SpeedStep technology that dynamically
varies the cpu multiplier depending on load.
In the case of the Q6600 the multiplier drops back to 6.0x and the cpu core
voltage drops marginally to reduce noise, temperature and power usage during
low demand periods.
SpeedStep can be disabled in the BIOS.
My graphics card does something similar.
It uses ATI PowerPlay 8.0 to dynamically vary gpu frequency and graphics
memory speed depending on load.

[Ext User(AdenOne)]

On Jan 6, 2:06 pm, "~misfit~" <misfit6...@yahoot.co.nz> wrote:
> Somewhere on teh intarweb "AdenOne" typed:
>
>
>
> > On Jan 6, 5:18 am, Matt <m.p.boul...@durham.ac.uk> wrote:
> >> Hey guys.
>
> >> I'm considering going for a Q6600 quad core CPU at the moment over an
> >> E6850 (see a separate discussion topic on this :) ). What's been
> >> holding me back over the Q6600 is, aside from the lower clock speed
> >> (which as more applications become multi-threaded won't be an issue)
> >> is the lower FSB speed it supports (1066 vs. 1333 MHz).
>
> >> Since I'm not keen on overclocking a CPU (due to the shorter life and
> >> the noise a good cooling system creates), is it possible to get a
> >> motherboard and memory that fully supports 1333MHz, set this CPU's
> >> FSB to 1333MHz and then just reduce the multipliers so its clock is
> >> running close to 2.4GHz again?
>
> >> Kind Regards,
>
> >> Matt
>
> > In real-world benchmarks, the quad-core 2.4GHz 6600 is maybe only
> > 10-15% faster than the dual-core 6850 at 3GHz. You cannot change the
> > multiplier of any Intel CPU except the "Extreme" versions, such as
> > QX6xxx Series.
>
> Wrong! You can reduce them *down* within certain parameters with the right
> motherboard. For instance, my E4500 can be set to anything between 6x and
> 11x. I have it set to 8 x 413MHz FSB, memory on 1:1 ratio. (That's 1.652GHz
> FSB for those who buy into the "quad-pumped" thing) I have left Speed-Step
> enabled so it still drops to 6x when it's not under load.
>
> > And you will pay a heavy price for them. 1.06GHz FSB vs
> > 1.33GHz is only an increase of 25%, however, there is very little
> > performance advantage unless you look very deeply due to memory speed
> > and so on - what I would do is go with the E6850, and spend the extra
> > cash saved on a good m\board supporting 1.33GHz FSB.
>
> The rest of this "info" is bogus as it's based on a false premise.
> --
> Shaun.

According to Intel this is considered a form of core frequency
adjustment that will void the warranty as you are operating the
processor outside of the tested limits of its core design. I know
SpeedStep technology uses multiplier downshifting, however hijacking
this technology to get a higher bus speed with the same core speed is
a form of over-clocking, and will put extra stress on the core. The
thread started stated he is not too happy to overclock, and bus
frequency increases are just as much overclocking as core frequency
increases. If you buy into the whole "more is better" philosophy, then
sure go for a quad-core QX6850 and at least you won't have to worry
about something being faster. Point is, 1.33GHz FSB is going the be
the standard for Penryn, Wolfdale and Yorkfield, as well as Nehalem,
so you might as well get a 1.33GHz m\board now. So if choosing between
a 1.33GHz FSB dual-core, and a 1.06GHz Quad-Core, the 1.33GHz dual-
core wins, whether you like it or not. Remember, each core has to
share the FSB, so 1066 divided by 4 = 266MHz bandwidth per core,
whereas 1333 divided by 2 gives 666MHz per core of bus bandwidth. The
Conroe core is crippled when its bandwidth on the FSB is reduced too
much, just look at the low end 800MHz models - most of their
performance hit is due to the cores having to wait for the FSB - and a
quad-core with 266MHz per core is less per core than even a Celeron or
Pentium Dual Core, which have 400MHz per core.

Then again, don't take my word for it, its a personal choice, but
given the inefficient way Vista handles thread allocation on multicore
machines, most of the extra boost from the quad-core is lost, sadly.

[Ext User(AdenOne)]

On Jan 6, 5:18 am, Matt <m.p.boul...@durham.ac.uk> wrote:
> Hey guys.
>
> I'm considering going for a Q6600 quad core CPU at the moment over an
> E6850 (see a separate discussion topic on this :) ). What's been holding
> me back over the Q6600 is, aside from the lower clock speed (which as
> more applications become multi-threaded won't be an issue) is the lower
> FSB speed it supports (1066 vs. 1333 MHz).
>
> Since I'm not keen on overclocking a CPU (due to the shorter life and
> the noise a good cooling system creates), is it possible to get a
> motherboard and memory that fully supports 1333MHz, set this CPU's FSB
> to 1333MHz and then just reduce the multipliers so its clock is running
> close to 2.4GHz again?
>
> Kind Regards,
>
> Matt

In real-world benchmarks, the quad-core 2.4GHz 6600 is maybe only
10-15% faster than the dual-core 6850 at 3GHz. You cannot change the
multiplier of any Intel CPU except the "Extreme" versions, such as
QX6xxx Series. And you will pay a heavy price for them. 1.06GHz FSB vs
1.33GHz is only an increase of 25%, however, there is very little
performance advantage unless you look very deeply due to memory speed
and so on - what I would do is go with the E6850, and spend the extra
cash saved on a good m\board supporting 1.33GHz FSB.

[Ext User(kony)]

On Sat, 05 Jan 2008 23:56:24 +0000, Matt
<m.p.boulton@durham.ac.uk> wrote:

>> In real-world benchmarks, the quad-core 2.4GHz 6600 is maybe only
>> 10-15% faster than the dual-core 6850 at 3GHz.
>
>I still fancy getting hold of that 10%-15%, especially with the move ot
>multi-threaded applications in the future.

In real world benchmarks the quad at 2.4GHz is slower than
the dual at 3GHz, on average. Such benchmarks must be
picked to show quad core winning and you have made no
mention of specific apps you use or plan to purchase to reap
the gain.

>
>You cannot change the
>> multiplier of any Intel CPU except the "Extreme" versions, such as
>> QX6xxx Series.
>
>Fair point. Would an overclock to a 1333MHz FSB (taking it up to 3GHz)
>for a Q6600 be feasible without a monstrous cooling system? I'm not at
>all keen about shortening the life of the CPU.


Yes it's a good goal, but as misfit noted you can use lower
multipliers. It is not going to significantly shorten the
life of the CPU, the CPU will still outlive the motherboard,
power, video, etc if past history is any evidence. I'm not
sure where you got the idea that overclocking (when done in
moderation) has any significant effect on processor life.
Reducing life of a part that would already last multiple
times as long as the other parts has to be seen in context.


>
>> And you will pay a heavy price for them. 1.06GHz FSB vs
>> 1.33GHz is only an increase of 25%, however, there is very little
>> performance advantage unless you look very deeply due to memory speed
>> and so on - what I would do is go with the E6850, and spend the extra
>> cash saved on a good m\board supporting 1.33GHz FSB.
>
>I'm certainly not discounting the E6850 at the moment, but if I can get
>a quad core up to 3GHZ with a 1333MHz FSB that has 2 extra cores, then
>I'm clearly onto a winner :)


Unless the lower heat of a dual core allows overclocking it
higher using the same heatsink and/or less noise from lower
fan RPM. Any time you were actually getting significant
gain in performance from using more than 2 cores, so too do
the 3rd and 4th core start producing a significant amount
more heat. Heat is, unless one is using exotic cooling
methods, generally the first reason one finds their
overclocking limitation unless they were only misfortunate
enough to have a board with chipset and/or bios limitations
where they just couldnt' set the speed as high as they'd
wanted.

Yes if you can get the quad core up to the same clockspeed
it is a good performance option.

[Ext User(Matt)]

> In real-world benchmarks, the quad-core 2.4GHz 6600 is maybe only
> 10-15% faster than the dual-core 6850 at 3GHz.

I still fancy getting hold of that 10%-15%, especially with the move ot
multi-threaded applications in the future.

You cannot change the
> multiplier of any Intel CPU except the "Extreme" versions, such as
> QX6xxx Series.

Fair point. Would an overclock to a 1333MHz FSB (taking it up to 3GHz)
for a Q6600 be feasible without a monstrous cooling system? I'm not at
all keen about shortening the life of the CPU.

> And you will pay a heavy price for them. 1.06GHz FSB vs
> 1.33GHz is only an increase of 25%, however, there is very little
> performance advantage unless you look very deeply due to memory speed
> and so on - what I would do is go with the E6850, and spend the extra
> cash saved on a good m\board supporting 1.33GHz FSB.

I'm certainly not discounting the E6850 at the moment, but if I can get
a quad core up to 3GHZ with a 1333MHz FSB that has 2 extra cores, then
I'm clearly onto a winner :)

Kind Regards,

Matt

[Ext User(kony)]

On Sat, 5 Jan 2008 19:35:13 -0800 (PST), AdenOne
<pacific-one@hotmail.com> wrote:


>According to Intel this is considered a form of core frequency
>adjustment that will void the warranty as you are operating the
>processor outside of the tested limits of its core design.

Of course the warranty is void, but mostly for two reasons:

1) They may not have tested it nor tested a resulting large
enough stability margin at higher clock frequency (or they
did, but #2 was more significant).

2) They are making more money for the higher clocked parts,
that is why they set it to a lower speed to match the lower
price and you aren't necessarily exceeding the limits of the
core design.


>I know
>SpeedStep technology uses multiplier downshifting, however hijacking
>this technology to get a higher bus speed with the same core speed is
>a form of over-clocking, and will put extra stress on the core.

That's like saying that driving your car at 50 MPH puts more
stress on it than 30MPH. Of course it does, but not an
excessive amount of stress.


>The thread started stated he is not too happy to overclock,

based on false premise, the two main factors are to keep it
cool enough and verify stability.


>and bus
>frequency increases are just as much overclocking as core frequency
>increases. If you buy into the whole "more is better" philosophy, then
>sure go for a quad-core QX6850 and at least you won't have to worry
>about something being faster.

An overclocked dual core will easily beat a QX6850 in many
benchmarks and actual use. Certainly not ALL benchmarks and
actual uses, which brings up back to square one that each
use may benefit more or less from the clockspeed increase or
# of cores increase. Then there's the cost factor.
Overclocking is not for everyone but should be seen for what
it is - testing the processor in the specific environment it
will be running in, which is something Intel cannot do when
they spec these things, and getting as much as can
reasonably had from a particular specimen of processor,
which is something Intel can't do because it would create
too many different speed levels, and because it would
destroy their pricing model.

[Ext User(Howard Goldstein)]

On Sat, 05 Jan 2008 23:56:24 +0000, Matt <m.p.boulton@durham.ac.uk> wrote:
: Fair point. Would an overclock to a 1333MHz FSB (taking it up to 3GHz)
: for a Q6600 be feasible without a monstrous cooling system? I'm not at
: all keen about shortening the life of the CPU.

With the right motherboard, yes, absolutely you can. You can even
cool it with a cheap Arctic Cooling Freezer 7 air cooler as I used to
do here in Florida when the ambient temp in my office was around 30C.
As for going over 3ghz, it seems a lot of folks can, I'm not stable
anywhere over 3ghz :(