Pushing Q9550 past 4 GHz?

Strawberry · 31 Dec 2009 at 03:12

Kingston 2GB DDR2 1060 MHz 5-5-5-15 2.3V
http://www.arlt.com/oxid.php/cl/details/anid/1010469/listtype/search/searchparam/kingston+2048mb

Technically my RAM isn't even overclocked as it's running below 1060 MHz. I believe P45 Neo-F boards have had some issues with 1060 MHz ram though but BIOS update 1.7 fixed it (my BIOS screen reports 1.9 so I'm guessing that can't be the issue).

Update: Not sure how I did it, but I got memtest to report 100% coverage 0 errors again, running at exactly 4 GHz. Gonna blend it for a while and report back but if this works, I'll plug in my other two sticks, test it again, after that trying to push it higher with these settings.

Subliminal Aura · 31 Dec 2009 at 03:30

Oh well I'm losing.

1 core keeps dying on blend. Going to try bumping PLL voltage before sleeping

Strawberry · 31 Dec 2009 at 03:52

Anything above 4000 MHz is still catastrophically unstable for me (Blending errors within 1 minute).

Oh well, back to finding a stable 4 GHz clock and I'll leave it at that.

Subliminal Aura · 31 Dec 2009 at 04:00

yeah I'm not having much dice with my 3.91 either.

vdroop on this board is amazing. Just had to bump some more vcore (last blend failed after 20 minutes, best so far)

I think it's something like 1.38 in the bios ... cpuz tells me it's 1.26 under load.

Thinking it's wiser to go back to my cooler 3.6

cotton_flex · 31 Dec 2009 at 04:01

I managed to take my Q9550 e0 to 3.92GHz stable. This is on the IP35 Pro with 8Gb pc8500 ram.

I gave up on the magic 4GHz barrier. I can boot but its not stable at all. I'm putting this down to the fsb limitations of the P35 chipset and amount of ram.

With your P45 you should be able to get 4GHz with a bit of tinkering.

Strawberry · 31 Dec 2009 at 04:02

What temps are you getting at 3.9?

And vdroop on this is amazing too. 1.43 in bios and 1.36 on load (cpuz).

@cotton_flex: Could you list your 3.92 voltages? I want to mimic them and go from there if possible. Would be appreciated.

Subliminal Aura · 31 Dec 2009 at 04:07

3.9 ghz

Coretemp

35C - 56C

Sensor
50C Full load

I also have my GPUs on full load too and they're at 70C each

Strawberry · 31 Dec 2009 at 04:12

Yeah, that's pretty much what I'm getting too.

I'm just wondering why I can't seem to get a stable 4 GHz clock anymore. ]:
It's been stable for 8 hours at 4000 MHz before. Don't remember the exact voltage though, I believe Vcore was a bit higher however.

I'll keep experimenting until I get 4 GHz stable again then go to bed and leave it running. Thanks for all the support.

Subliminal Aura · 31 Dec 2009 at 04:17

Take care mate. G'night and thanks for keeping me up too grrrr ! LOL

Think I've finally got..... ARRRRRRrrrrrrrrgggghhh !

Spoke too soon died after 25 minutes. Going to save this profile and try again another day

Night

Strawberry · 31 Dec 2009 at 04:31

wh3eee, 30 mins and counting. That's a 1400% increase in stability so far.

Subliminal Aura · 31 Dec 2009 at 04:35

Strawberry said:
wh3eee, 30 mins and counting.

Don't do this ! I'm all ready for bed now.

Besides, I still think you're putting way too much vcore in that chip !

There must be a crazy difference between idle/full load ?

Strawberry · 31 Dec 2009 at 05:31

No sleep for me!

I finally got a decently stable 3933 MHz (still no errors).
But my Vcore is down to 1.33 BIOS and 1.30 load.

I'll still try to hit a stable 4000 MHz.

Edit: Full list of settings I changed:

FSB : 462 MHz
Vcore : 1.335 V
MCH : 1.376 V
VTT: 1.35 V
(everything else left [Auto])

What should I increase for FSB 471? I tried the same voltages but it error'd after 2 minutes. Am currently blending 464 at same voltages (3.95 GHz)
What about FSB 495?

Edit 2: Been running for 8 minutes so far which makes it 600% more stable than most othe... damn rounding error. Brb pulling FSB down to 462 again and blending it further until I find some different setups I could try.

Edit 3: Raised to 4012 and bumped vcore up to 3.6 blending it now. Wish me luck!
Edit 4: Failed, increased MHC a notch and blending again. Let's see what happens.

Strawberry · 31 Dec 2009 at 06:32

Alright, calling it quits for now.

Here's my max stable clock:

Edit: One last question. Should I have SpeedStep on or off? (lowers cpu freq when idle) I have it on, could this be making my system unstable in any way?

Strawberry · 31 Dec 2009 at 18:34

Yay, another day of blending!

Currently testing 3984 MHz, seems stable so far (10 mins), here's hoping it'll continue that way.

Edit: Damn error'd at 12.

geowal19 · 31 Dec 2009 at 18:47

40degrees sounds too good to be true tbh wid a H50 at 1.4v 4Ghz

Strawberry · 31 Dec 2009 at 18:56

Open case and a Noctua S12B pulling cold air into the radiator.
I did relax my Vcore a lot though.

Update: I just noticed that I can *lower* my CPU multiplier if Intel SpeedStep is disabled. So what I'm doing right now is trying to get an FSB of 500 stable (3920 MHz still), then pushing my cpu mult back up to 8.5.

Subliminal Aura · 31 Dec 2009 at 19:20

1.344 Volts ? Is that under full load ?

setter · 31 Dec 2009 at 19:23

1.344 isnt to bad for 4ghz, i had a q9550 that ran 3.8ghz on 1.216 (stock vid of 1.200) albeit that was on a p45 asus p5q deluxe mobo.

TheTard · 31 Dec 2009 at 19:42

That's asking a lot, i think you will be able to push it but would it really be worth it?

Strawberry · 31 Dec 2009 at 21:58

Full load droops to 1.30v. I haven't tried going lower though, but it's stable, not too hot, and I'm happy with it. Up to 1.36 is considered safe by intel for this chip.

I really can't get my RAM stable at these high FSBs though, even though it's advertised to support 1060 MHz. Even when running the CPU at 3928 MHz (FSB of 491 MHz, mult lowered to 8) I cannot get my RAM stable at all, even though CPU-only tests are stable (small FFT).