Hitting an FSB wall.

Associate
Joined
31 Aug 2009
Posts
1,015
Location
Surrey, UK
I'm trying to OC my Q8300 on an Asus P5Q Deluxe, and appear to be hitting a wall with the FSB - at 450.

I've tried tweaking everything. Voltages, straps, even taken down the multi on the CPU. Prime95 fails usually before 20 mins. MemTest is fine, can run that for 30 mins no problem.

I've had my vcore up around 1.36, nb @ 1.42!, fsb termination voltage at 1.4 volts. Basically taken all the voltages as high as I can get them as a form of troubleshooting and I still get a problem.

Yes I've tried loosening my timings. 2x2GB Kingston HyperX 1066 5-5-5-15 (default)

It literally seems that when the FSB is over 450, Prime fails regardless of ANYTHING else.

What else can I change to support the FSB above 450?

I've exhausted Google. I just can't find specific enough help from simply searching.
 
Is the vcore tied in ONLY to the final rating of the CPU after the multi has been applied?

So 450 fsb at 6.0 multi will require the same vcore if the fsb was at 450 and the mutli was 7.5?
 
Try the northbridge with lowered volts instead of increased. I know it doesn't make sense, but I've seen it work with quads on the Asus rampage formula. Might help to know the BIOS revision you are using :-)
 
I've been reading for about the last 6 hours about GTL reference voltages. I've gone from almost instant Prime fail, to holding for around half an hour with the FSB at 450.

I'm pretty sure the problem lies in the GTLs. Just a case of learning how they work; they're not as simple as other voltage settings as they're actually percentages rather than direct levels.

Currently got 0 and 1 GTLs down at .630, with a FSB termination voltage of 1.36 and a 1.35 vcore (just for testing). I'll try taking the NBv down a bit tomorrow.

Right now I feel like my brain is going to melt.
 
Pete.Dakota said:
  • Intel® Core™2 Quad Q8300
  • Asus P5Q Deluxe
  • 4GB Kingston HyperX 1066

Hey Pete,

You seem to be taking the road-of-pain route towards your overclocking! :D

If you want to make life simple I would slam all the voltages back on [Auto] and start again . . . heres some simple steps for you

  • Intel® Core™2 Quad Q8300 (7.5x400) 3.0GHz DDR2-1066 [8:5]
  • Ai Overclock [MANUAL]
  • FSB Frequency [400]
  • DRAM Frequency [DDR2-1066]
  • Load-Line Calibration [Enabled]
If you can confirm the clock above works fine and is prime stable try this next step

  • Intel® Core™2 Quad Q8300 (7.5x450) 3.37GHz DDR2-1080 [6:5]
  • Ai Overclock [MANUAL]
  • FSB Frequency [450]
  • DRAM Frequency [DDR2-1080]
  • vCore [Auto] *
  • Load-Line Calibration [Enabled]
*vCore may need manually adjusting if the auto setting doesn't supply enough juice

Leave the CPU Multiplier on [Auto] or set it at *maximum" (7.5) . . . by manually lowering it to [x6] you are increasing the pressure on the northbridge clock (NBcc) which is the "real" FSB . . . when you lowered it to [x6] and try to run a 450MHz FSB you were actually running the FSB (NBcc) at 525MHz which it may not have liked? . . . or at least not like with your attempted manual override!

You don't really wanna be farting around with vNB, vTT, or any GTL refs at this stage if at all possible . . . put them all back on [Auto] if you like the simple life! ;)
 
I disagree emphatically with the conclusion of Wayne's post. Auto is no place for northbridge voltage or fsb, and pushing the limits on P45 with quad cores does mean gtl refs. 450 is low for the core2duo range, but then the Q8300 doesn't tend to overclock as well as the highest end 775 quad cores. Some northbridge straps will work better than others, and this may vary with bios version. It's related to the internal timing of the P45 chip, where slackening off the timings works much like changing to cas6 on the ram does.

But then, I'm strongly in favour of the road-of-pain approach, so "take the easier option" never sits that well. As a warning though I spent a good three months trying to get 4ghz out of a q9550 (8.5 multi) on a P5Q board and it just wasn't having any of it. 3.85 was all it had to give under air cooling. This was probably not time well spent.

One thing of note is that the optimum gtl refs will drift a bit as you change other voltages. I'd suggest decreasing to a stable speed, then playing with the gtl's and vcore/nb/fsb to see what values allow you to take the main voltages the lowest. Then overclock with these values, but expect to go back and tweak them slightly when you can't quite get the next notch stable.

[auto] on the p5q is unlikely to undervolt anything, keep a close eye on temperatures if you go down that route as it may cheerfully push 1.6V without telling you about it. Careful with the ram too, some of the boards had a nasty trick of overvolting by 0.08V relative to the bios setting iirc. Oh, it may also treat the lowest manual settings identically to auto, so if the lowest northbridge voltage is 1.2 you probably want to set it to 1.22.
 
I can't get my Q8300 above 3.4ghz without touching the GTLs. Simple as that. It's a poor OCer; only sports a 7.5 multi.

If I set the FSB above 450, regardless of vcore, fsb term, nb voltage, strap, dram clock - anything - it fails in prime.

I'm beginning to learn more about GTLs now, and they seem imperative for a stable OC on 45nm. Especially if you have a low multi. I don't expect to get my Q8300 above 3.5ghz, but I'd like it stable at 3.4.

Luckily my board uses GTL references instead of more direct mv control, so I don't have to do much math :)
 
Back
Top Bottom