• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Pushing Q9550 past 4 GHz?

Strawberry

Strawberry

Strawberry

Associate
Joined
29 Dec 2009
Posts
649
Location
Germany
Sorry if this is the wrong section, I didn't know whether to post here or in Overclocking & Cooling.

Hey! (First post here :D)

I've been trying to hit 4.2 GHz on my Q9550. It has a locked multiplier of 8.5, so my target FSB is 495 (4207 MHz). I've seen some posts floating around the internet of people getting 495 Mhz / 4.2 GHz to be stable, however no matter what I try I get a flood of lockups or rounding errors.

I think the issue might be my RAM. I have 6 GB of Kingston DDR2-8500 (1060 MHz), however my mobo (P45 Neo-F) apparently has some notoriety with 1060 MHz. (Currently it's stable at 940 MHz) A BIOS update is supposed to fix this issue, but I have not for the love of christ been able to create a bootable USB flash drive, and I don't have a CD/DVD or floppy drive in my system. I'm not sure whether my BIOS is already updated either way.

My current clock:

FSB 471 MHz
CPU 4004 MHz
DRAM/FSB 1:1 (942 MHz DDR2)
Vcore 1.43 (everything else set to auto; auto sets 1.45 for vcore but it's stable at 1.43)

I've tried raising MCH to 1.45 and VTT to 1.4 (I don't want to go any higher than that), as well as Vcore to 1.45 but it's not stable at 4.2 GHz.
Do I need to raise any of these voltages any higher? Is there anything else I should be raising instead? I don't feel too comfortable going above those values as I intend to keep this system running for several years and don't want to reduce it's lifespan.

(I'm on liquid cooling by the way, after a few hours of prime95 I get 55° max in a rather heated room, if I open the window regularly it drops to around 40°)

Disclaimer: This is the first PC I built completely myself, also my first overclock, so I'm new to a lot of this. :)

Thanks in advance!
 
I guess not. I still want to try though, nothing better to do.

Right now it's more for bragging rights than anything else, as I don't really notice a notable everyday usage / gaming increase over stock (2.83 GHz).

I do, however, use my machine for lots of huge mathematical calculations - the performance boost can definitely be seen on these. (Eg. fractal generators or simulations of neural networks)

When a single 2048x1365 image at little over 36893488147419103232x zoom and several thousand iterations takes as long as a minute to calculate, a 5% boost in speed makes itself noticable in the long run, especially considering I can zoom in to 340282366920938463463374607431770000000x zoom.
 
Last edited:
Nahhhh 4ghz is pretty easy these days, the new target is 4.2ghz. :P

This seems to be what most i7 920s will do on air!
 
i had a 9550 and i struggled to get 4GHZ for ages, in the end settled for 3.8 thats after i put the NB under water so i could put 1.4 through it.
might be worth looking into the clock skew settings to get the most from the CPU, is your CPU EO?
 
Yes it is E0. I've never heard of Clock Skew. What's its effect on overclocking and what should I be looking for in my BIOS?
 
Last edited:
Well Let me be the first to say that if you require huge mathematical calculations you are better off running it at stock vice overclocked. After all one digit off means life and death in that field, no? I certain ly wouldn't rely on a "stable" overclock for those purposes.
 
Well, technically no as I'm using it for private purposes. For my current science project I am actually running the test 10,000 times and using the average of those results as my final result. (Since the initialization is randomized).

I definitely do not run anything error-critical, but I'm still gonna make sure I survive at least 12 hours of prime95.
 
Update: Using in-place small FFT tests in prime95, my system seems to run stable so far (around half an hour), a LOT better than the large FFT tests, which were failing after as little as 2 minutes. (Assuming the entire system didn't lock up 10 seconds after starting the test)

I'm starting to suspect my RAM is key to the rounding errors and not my CPU. Will try getting RAM stable instead, I'll check back here.

Current Speed: 4.1 GHz

Edit: Just a random question: Does the correct application of thermal paste make a large difference? If so, how large?

When I built in my H50, I just pulled off the stock intel cooler (left a small smudge of thermal paste on it), and pushed on the new cpu cooler block (had to rotate it slightly in the process as well so it's slightly smeared). How bad is this?

I'm going to re-apply it but I don't have any thermal paste nor any alcohol to clean it off, so I'm hesitating before I go out and buy.

Edit 2: Been running for nearly an hour and a half and in-place small FFTs are still stable. Will continue running for longer, but RAM/NB is definitely my failure point.
 
Last edited:
I decided to have a shot at pushing my q9550

First relax and I mean REALLY relax your memory timings. Bump the dimm voltage to a safe 2.1

Currently I'm on 3.91ghz (460mhz) @ 1.36 Vcore running full torture - 32C idle 50C load

Not sure if I want to bump the vcore and go for higher clocks as my previous stable is 3.6@ 1.3Vcore

We'll see
 
I lowered my Vcore by a lot and the CPU stress test is still running stable (not a single hour been running it practically the whole time), what do you mean by relaxing my memory timings? Give me some examples.

I'm not too keen on all of the memory stuff, but I know my sticks are 5-5-5-15. I currently removed 2 of my 3 sticks (running 2 GB only right now) to see if I can get it stable that way.

Can the CPU clock / Vcore have any effect on memtest's results? I'm getting confirmed (random) errors in memtest now, it's not any specific address though and every re-run is different so I'm ruling out hardware failure. My Vcore is actually stable at 1.36 (running 4086 MHz atm), but can this have any effect on my RAM failing?
 
Blend is unstable, small FFTs is the one that's been stable for so long. Memtest reports errors. I have managed to get the rate down a lot by decreasing the memory timings though. (Running at 6-6-6-20 right now to see if it's stable. Is this the right thing to do?)

In a moment I will go back down to 3.6 or so to see if memory is stable at that clock.
 
definitely your RAM dude. Make sure it's off auto in the BIOS and shift it to a fixed speed.

I'm running pc2-6400 @ 920mhz.

Yes please get back to blend. If it won't blend it's no good.

When I started messing around an hour back - it wouldn't blend for more than a second. Getting there now ;)
 
The best 4.2 clock I have gotten so far blends for around 10 minutes btw.
Currently memtesting at 3.6 GHz just to make sure.

Also, my SPD says 5-5-5-18 but the site I bought it off says 5-5-5-15 in the product description. I'm using 5-5-5-18 right now just to be safe.

One last question: Does anybody know what the "Enhance Setting" in AMI Bios means? It says it improves stability/performance when overclocking especially for dual channel, but some people have gotten huge problems with similar "enhance" settings in other BIOSes if I recall correct, so I'm not sure if I should Enable or Disable it. It's Disabled right now.
 
Last edited:
Northbridge voltage is MHC right? If so, I've tried up to 1.46. Right now it's 1.4. Thanks for the fast replies!

Update: Okay 3.6 is definitely stable (100% memory coverage 0 errors). I was beginning to think my RAM stick may be busted. Back to BIOS *yay*!
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom