Learning as I go... the going has been pretty easy so far, which is good, but as they say, a little bit of knowledge is a dangerous thing :evil: I'd like to know if I'm completely off the mark, or pointed generally in the right direction.
Equipment: ASUS P5Q Pro, latest BIOS (1603), Q6600 G0, L814B433, 1.300 VID, 4 * 2 GB Corsair XMS, DDR2-800, 5-5-5-18, spec'd for 1.9 volts.
Like I said... so far, so good. I was concerned about possible issues from having all the RAM slots populated, so I began the OC process by OCing just the FSB first. Set CPU multi to 6, FSB to 400, RAM to DDR2-800, then keyed in all the voltages for NB, VTT, etc. to their default values.
Maybe I worried too much? Everything appears to be fine. RAM checks out, large FFT Prime checks out, so it doesn't look like the NB is experiencing any particular stress by having all those address lines to deal with... time to move on to CPU.
First stop: 400 * 8, 3.2 GHz. Previous farting around had suggested that the CPU's VID of 1.300 probably wouldn't carry this level of OC, and that remains true.
Around this point, I deviated somewhat from normal OC testing procedure. This is where the "little bit of knowledge" becomes dangerous, because I had just finished reading a bunch of information on GTL so I could better understand its function.
Well, with that "little bit of knowledge" in hand, I decided to use Large FFT Prime in order to test for stability in my CPU OC, because I thought it would be a good idea to look for any potential for raising VTT/GTL in order to stabilize the CPU OC rather than relying just on brute force Vcc.
Well, I may have found some... Large FFT Prime testing at 1.30625, 1.3125 and 1.31875 generated outright BSODs, so perhaps that's indicative of errors occuring that some GTL tweaking might correct... 1.3250 was large FFT stable for a full 12 hours, which is good. I didn't Prime test this for a full 24 because my real OC target is 3.6, though I've been using the PC for the last couple days @ this setting for inconsequential tasks, and it's held up just fine.
As for 3.6... only initial testing has been done so far; the temp log that was generated while stress testing 3.2 was somewhat alarming, so I popped the waterblock off of the Quad to look at TIM spread/contact... it was abysmal, so I pulled the CPU and gave it a lapping. By doing so I basically shredded my warranty, soaked it in kerosene and set it on fire, but temp-wise, it was effective... instead of the CPU cracking 60º C load @ 3.2 and 1.325v (under WATER), it loads up into the low-mid 50s now. Still not fabulous for water perhaps, but it's a "quiet" loop, not a max performance loop. I can deal with parts running hot, as long as they're not unreasonably hot, if it maintains my blissful silence.
Back to achieving 3.6 -- before I broke off to address cooling issues, I noticed the same sort of pattern emerging while upping voltages that I saw while getting to 3.2. I only got as far as testing out 1.4125 Vcc, which got me into Windows and let me do some SuperPi iterations, but when attempting large FFT Prime @ this Vcc it's back to the familiar BSOD and core dump/restart. At least that lets me know that this chip will quite probably do 3.6, and if I can achieve it by way of brute forcing Vcc, I'll guesstimate that it'll take somewhere between 1.425 and 1.45 in order to do so.
And that's where I'm at. I'm not really stuck, I sense that I'm getting somewhere with this process, though I do want to check in and make sure the methodology is... sensible.
...And if so, when I move on to the "refinement" process of the OC that I have in mind (increasing VTT/GTL and reducing Vcc), is there any particular way I should go about doing so... should I back off Vcc until I lose stability, then give VTT/GTL a boost? Something else? Do I even know what the hell I'm talking about?
Well, that's why I'm posting... this wouldn't be the first time I fancied myself smarter than I really am, so... if that's the case, slap me into shape.
(gently)
TIA.
Equipment: ASUS P5Q Pro, latest BIOS (1603), Q6600 G0, L814B433, 1.300 VID, 4 * 2 GB Corsair XMS, DDR2-800, 5-5-5-18, spec'd for 1.9 volts.
Like I said... so far, so good. I was concerned about possible issues from having all the RAM slots populated, so I began the OC process by OCing just the FSB first. Set CPU multi to 6, FSB to 400, RAM to DDR2-800, then keyed in all the voltages for NB, VTT, etc. to their default values.
Maybe I worried too much? Everything appears to be fine. RAM checks out, large FFT Prime checks out, so it doesn't look like the NB is experiencing any particular stress by having all those address lines to deal with... time to move on to CPU.
First stop: 400 * 8, 3.2 GHz. Previous farting around had suggested that the CPU's VID of 1.300 probably wouldn't carry this level of OC, and that remains true.
Around this point, I deviated somewhat from normal OC testing procedure. This is where the "little bit of knowledge" becomes dangerous, because I had just finished reading a bunch of information on GTL so I could better understand its function.
Well, with that "little bit of knowledge" in hand, I decided to use Large FFT Prime in order to test for stability in my CPU OC, because I thought it would be a good idea to look for any potential for raising VTT/GTL in order to stabilize the CPU OC rather than relying just on brute force Vcc.
Well, I may have found some... Large FFT Prime testing at 1.30625, 1.3125 and 1.31875 generated outright BSODs, so perhaps that's indicative of errors occuring that some GTL tweaking might correct... 1.3250 was large FFT stable for a full 12 hours, which is good. I didn't Prime test this for a full 24 because my real OC target is 3.6, though I've been using the PC for the last couple days @ this setting for inconsequential tasks, and it's held up just fine.
As for 3.6... only initial testing has been done so far; the temp log that was generated while stress testing 3.2 was somewhat alarming, so I popped the waterblock off of the Quad to look at TIM spread/contact... it was abysmal, so I pulled the CPU and gave it a lapping. By doing so I basically shredded my warranty, soaked it in kerosene and set it on fire, but temp-wise, it was effective... instead of the CPU cracking 60º C load @ 3.2 and 1.325v (under WATER), it loads up into the low-mid 50s now. Still not fabulous for water perhaps, but it's a "quiet" loop, not a max performance loop. I can deal with parts running hot, as long as they're not unreasonably hot, if it maintains my blissful silence.
Back to achieving 3.6 -- before I broke off to address cooling issues, I noticed the same sort of pattern emerging while upping voltages that I saw while getting to 3.2. I only got as far as testing out 1.4125 Vcc, which got me into Windows and let me do some SuperPi iterations, but when attempting large FFT Prime @ this Vcc it's back to the familiar BSOD and core dump/restart. At least that lets me know that this chip will quite probably do 3.6, and if I can achieve it by way of brute forcing Vcc, I'll guesstimate that it'll take somewhere between 1.425 and 1.45 in order to do so.
And that's where I'm at. I'm not really stuck, I sense that I'm getting somewhere with this process, though I do want to check in and make sure the methodology is... sensible.
...And if so, when I move on to the "refinement" process of the OC that I have in mind (increasing VTT/GTL and reducing Vcc), is there any particular way I should go about doing so... should I back off Vcc until I lose stability, then give VTT/GTL a boost? Something else? Do I even know what the hell I'm talking about?
Well, that's why I'm posting... this wouldn't be the first time I fancied myself smarter than I really am, so... if that's the case, slap me into shape.
(gently)
TIA.
Facebook
Twitter