q9450 @ 4ghz air

[afreak]

The other q9450 thread seems to have gotten kind of cluttered, so here's my try at 4ghz.

I haven't kept up with all of the latest overclocking bios settings so it took me a few hours to achieve this speed. I'm not going to run at this speed, but I just wanted to see if I could get there.

http://www.flickr.com/photos/2...izes/o/in/photostream/
CPUID FSB 501

http://www.flickr.com/photos/2...210841/in/photostream/
voltages.

NB and FSB are pretty high...probably not a good idea to run at this for long. I didn't really mess with CPU voltage since it didn't seem to play much of a factor, so you can probably reduce these. The main things to getting the system to boot XP were NB voltage, FSB voltage, and NB GTL voltage reference.

I started by just trying to boot at 500mhz FSB with a lower multiplier and once that booted I incrementally increased the multiplier.
http://valid.x86-secret.com/show_oc.php?id=341089 (6x500)
http://valid.x86-secret.com/show_oc.php?id=341092 (7x499)
http://valid.x86-secret.com/show_oc.php?id=341109 (8x499)
http://valid.x86-secret.com/show_oc.php?id=341112 (8x501)
At each step I had to increase NB or FSB voltages. starting at 7.5x (not shown) I needed to start modifying the NB reference voltage...which I had no idea what it did at the time.


Setup
-------
Q9450
Asus Rampage Formula
G.Skill F2-8000CL5D (2x2gb)
Tuniq Tower 120
Corsair hx650w
and for testing: a honeywell HT800 blowing into open case (http://www.amazon.com/Honeywel...ormance/dp/B0000BYDUC )



I'll probably end up just running 3.2 ghz(8x400mhz) under normal circumstances, but it was fun just seeing if this was possible! I didn't spend much/any time trying to make my system stable at 4ghz, but I did use it for 40 min before I decided to down clock it. I only attempted prime95 once at 4g and it froze immediately. Someone with more time could probably easily make this stable.

 

[harpoon84]

Yeah, pretty much confirms my suspicions regarding the Q9450, for max OC it will be FSB limited rather than silicon limited. Still, booting into windows @ 500FSB is impressive stuff, though I think you'll find it hard to achieve stability at this speed without added NB cooling, considering the voltages being used.

 

[aigomorla]

booting isnt the issue.

These chips will take 4.0ghz @ an insane low voltage.

The problem is the rest of the hardware. 1.8V on your NB is insane.

Im holding 1.35V, wouldnt dare push it higher then 1.7v on Water.


Its a nice Overclock, would love to see some stability tests on it tho.


Your cpu voltage is a bit too high also,. I wouldnt dare push higher then 1.45 on bios on these guys. And thats with watercooling.

1.5 and your asking for it.

 

[afreak]

Here are a few more reference voltages numbers

Everything set to auto (2.66ghz, 8x333, mem@400mhz ):
Vcore = 1.18
FSBT = 1.20
NB = 1.36

everything auto except FSB @ 400 (mem @400mhz)
Vcore = 1.22
FSBT = 1.33
NB = 1.63

mem @ 500mhz
Vcore = 1.22
FSBT = 1.34
NB = 1.63


So my NB voltages run a bit high even under "normal" conditions.

 

[runawayprisoner]

I think you're overdoing it with FSBT? vCore and vNB sounds about right to me. But up to 490MHz, I only needed 1.42v FSBT. Try lowering that one?

 

[afreak]

you're right that my FSB voltage is a bit high, but I think I needed to up that in order to get my NB reference voltage multiplier to work out.

I believe Vfsb * NBref = perceived Vfsb voltage

or something like that.

so to get a perceived nominal Vfsb of 1.2 with a multiplier of 0.67x, I need 1.2/0.67 = 1.79v for Vfsb.

now that I think about it, the naming scheme doesn't sound right. it should be called FSB terminination voltage reference instead of NB reference voltage.

 

[aigomorla]

Originally posted by: afreak
you're right that my FSB voltage is a bit high, but I think I needed to up that in order to get my NB reference voltage multiplier to work out.

I believe Vfsb * NBref = perceived Vfsb voltage

or something like that.

so to get a perceived nominal Vfsb of 1.2 with a multiplier of 0.67x, I need 1.2/0.67 = 1.79v for Vfsb.

now that I think about it, the naming scheme doesn't sound right. it should be called FSB terminination voltage reference instead of NB reference voltage.

I always known this as the fsb termination voltage.

Also fsb termination voltage shouldnt be higher then vcore voltage.


And if you need that much voltage for your fsb termination, your going to burn something out on your board really quick.

 

[sofarfrome]

it is amazing with these chips in that you need very little vcore to get high OC but the Nb and VTT voltages will kill a board trying to run 24/7 at a high OC.

 

[runawayprisoner]

Unless you go extreme cooling on those components, that is.

 

[aigomorla]

Originally posted by: runawayprisoner
Unless you go extreme cooling on those components, that is.

ummm no even then, the hardware cant handle that much stress.

And trust me all my gear minus my laptop and my mom's rig is all watercooled.

 

[runawayprisoner]

Indeed. But mine ran fine at slightly lower settings 24/7 on air. But then again, maybe it was because voltages weren't too high then.

 

[Rubycon]

Originally posted by: aigomorla

And trust me all my gear minus my laptop and my mom's rig is all watercooled.

At sea it's easy as falling overboard to get a lappie watercooled. Saltwater kills a motherboard faster than a hammer too. :laugh:

Someone needs to cue up Peter Gabriel's Sledgehammer while "fine tuning" a motherboard that desperately needs it. (Abit comes to mind)

 

[aigomorla]

Originally posted by: Rubycon

Originally posted by: aigomorla

And trust me all my gear minus my laptop and my mom's rig is all watercooled.

At sea it's easy as falling overboard to get a lappie watercooled. Saltwater kills a motherboard faster than a hammer too. :laugh:

Someone needs to cue up Peter Gabriel's Sledgehammer while "fine tuning" a motherboard that desperately needs it. (Abit comes to mind)

ROFL....

yeah so does the condensation every morning!

 

[Rubycon]

Originally posted by: aigomorla

ROFL....

yeah so does the condensation every morning!

No condensation here UNLESS I take something outside into the warm air. I have an oven that keeps camera equipment at 94°F for this purpose. :Q

Condensation itself is not the problem but outside the salt covers everything.

 

[Dadofamunky]

Originally posted by: Rubycon

Originally posted by: aigomorla

And trust me all my gear minus my laptop and my mom's rig is all watercooled.

At sea it's easy as falling overboard to get a lappie watercooled. Saltwater kills a motherboard faster than a hammer too. :laugh:

Someone needs to cue up Peter Gabriel's Sledgehammer while "fine tuning" a motherboard that desperately needs it. (Abit comes to mind)

LOL...

Originally posted by: aigomorla

And if you need that much voltage for your fsb termination, your going to burn something out on your board really quick.

I think we're loooonnnnngggg overdue for Intel to have their IMC, eh? I think their external memory controller has pretty much come to the end of the line. Frustrates the crap out of us OCers, anyway.

Originally posted by: Rubycon
No condensation here UNLESS I take something outside into the warm air. I have an oven that keeps camera equipment at 94°F for this purpose. :Q

Condensation itself is not the problem but outside the salt covers everything.

Hey, you could always move to a more civilized place. It IS named the Bermuda Triangle for a reason, after all. (I'm sure Bermuda's great, I've never been there) I have a mother- and father-in-law who live on a stinkin' island in the middle of the Bahamas and I frankly have no idea how she keeps her PowerBook alive... I think they're certifiable...:Q

 

[BigMoosey74]

There are probably others but you are the first person I have seen validate the 9450 at 4GHz. Your X48 chipset looks like it is doing quite nicely. Are you stable at 4GHz? I saw someone else that validated the first X3350 at 4GHz on a P35 but they couldn't run demanding applications.

 

[Link]

Originally posted by: afreak
Here are a few more reference voltages numbers

Everything set to auto (2.66ghz, 8x333, mem@400mhz ):
Vcore = 1.18
FSBT = 1.20
NB = 1.36

everything auto except FSB @ 400 (mem @400mhz)
Vcore = 1.22
FSBT = 1.33
NB = 1.63

mem @ 500mhz
Vcore = 1.22
FSBT = 1.34
NB = 1.63


So my NB voltages run a bit high even under "normal" conditions.

afreak, you just proved my suspicion that there's something wrong with the 0219 Rampage bios. It seems if you set the FSB to 400mhz or higher, the bios sets the NB voltage to 1.61~1.63v if you leave it at auto.
I had P5E which only needed 1.30v on NB to keep the system stable, but when I swtiched to Rampage, 1.30v wasn't enough. I thought x48 should require less voltage on NB than x38. There's something fishy going on here.

 

[BigMoosey74]

^ Very true. I would stay away from auto settings. Those voltages should be a far overshot. I am curious to see if you back down the MCH voltages, how low the X48 will go.

 

[Acanthus]

isnt X38/X48 good to 1.7v with a fan?

P35 can take it and its on the same process.

 

[runawayprisoner]

Originally posted by: BigMoosey74
There are probably others but you are the first person I have seen validate the 9450 at 4GHz. Your X48 chipset looks like it is doing quite nicely. Are you stable at 4GHz? I saw someone else that validated the first X3350 at 4GHz on a P35 but they couldn't run demanding applications.

G35, to be more precise. And 4GHz wasn't stable to do anything other than a quick SuperPI 1M run and a validate link. I needed more voltage through my NB. Seems like 1.73v is the bare minimum for 490, and 1.81v is the bare minimum for 500 stable. It's the chipset that is important in this case, it seems.

 

[afreak]

Originally posted by: BigMoosey74
There are probably others but you are the first person I have seen validate the 9450 at 4GHz. Your X48 chipset looks like it is doing quite nicely. Are you stable at 4GHz? I saw someone else that validated the first X3350 at 4GHz on a P35 but they couldn't run demanding applications.

I did not make an attempt to make the system stable at 4Ghz, but I browsed the web for 40 min before I manually rebooted and down clocked.

However, at one point when I first got it working at 4Ghz I tried prime 95 and my system froze instantly.

Originally posted by: Link
afreak, you just proved my suspicion that there's something wrong with the 0219 Rampage bios. It seems if you set the FSB to 400mhz or higher, the bios sets the NB voltage to 1.61~1.63v if you leave it at auto.
I had P5E which only needed 1.30v on NB to keep the system stable, but when I swtiched to Rampage, 1.30v wasn't enough. I thought x48 should require less voltage on NB than x38. There's something fishy going on here.

You're right that auto overvolts. I've since brought my nb voltage down to ~1.4 with fsb@400mhz and memory@500mhz. I may be able to lower it even more, but I haven't tried.