E6400 - an impossible OC?

[AWhackWhiteBoy]

So I built this system over 2 weeks ago and I'm getting fed up with it. I can OC ram no problem but as soon as I push the CPU the system becomes unstable. The best I can do is 2.6ghz, but when I stress test it the system just reboots after 10-20 minutes with no errors or warnings.

The CPU never gets hotter than 45C under load, the north bridge runs at about 50-53C. I've tried pushing the Vcore as high as 1.55V.

Here is an example of some meager settings that are UNSTABLE:
Screencap

What exactly am I missing? I used to be OC savvy back in the barton days but a lot of this stuff is new to me. I feel i'm pretty up to speed on this new hardware, but maybe there is something I'm blatantly missing? The only thing I can think of is the PSU, however for a 60 or so watt CPU one would think 550 would be plenty.

Any input would be greatly appreciated.

DFI Infinity P965-S - stock bios
Core2 E6400 -Tuniq Tower HS
2gigs of G.Skill(Micron D9) PC6400 - 4-4-4-12
550W Coolmax PSU
eVGA 8600GTS
Vista x64 Ultimate

 

[PCTC2]

for one, you're probably pushing your RAM too hard on the divider. You will want to run memtest at least a few hours b/c you have pushed your RAM at really tight timings and it may be unstable. you are 100 MHz OC on your RAM at tighter timings. I think it might be your RAM. plus, you might need to boost vDIMM and vMCH
plus, why is your multiplier at 7x? why don't you set it to 8x? also, did you disable C1E and EIST? did you lock your PCI/PCIe clocks?

I have my E6400 @ 3.33 GHz. but you just seem to be pushing your memory way too tight. Try CL5 or keep your RAM at stock and try pushing your CPU first so you know it isn't the RAM. good luck.

 

[Zenoth]

Try a Memory divider to avoid over-clocking it or keeping it within its default speeds, to put it out of the equation. For example if your Memory, by default, runs at 600Mhz, then try a divider (which should basically decrease your Memory speed from its default) that will keep it at or around 600Mhz despite the CPU being well over-clocked above it's default speed. I don't know how much that can be done with C2D's and DDR2 Memory, since I'm still on old technology myself (939 CPU, DDR1), but in my case, I can push my CPU a little more, and make it more stable if I try to keep my Memory at or around 200Mhz (it's default speed), to perhaps something like +/- 195Mhz to 215Mhz.

If however I keep my Memory running at a 1:1 ratio, then it will constantly get higher in speeds, following the CPU over-clock itself, and that's where it gets complicated because you need to consider a lot more possible variants that might be affecting your OC stability. Well I'm almost telling you how to OC basically, that's not my initial intentions, you said you were good at it before, but you might have just forgotten to isolate components from your equation, that's a well known trick and first step for OC'ing. Just keep your Memory at default (or around) speeds, and OC your CPU as much as you can within those Memory boundaries, and then once you've found that comfortable spot (not too much vCore and not too much heat) then do the opposite, try to isolate your CPU and OC your Memory as much as possible while trying to maintain the CPU as less OC'ed as possible. That way you'll find a balanced setting, eventually, and hopefully.

 

[AWhackWhiteBoy]

Originally posted by: PCTC2
for one, you're probably pushing your RAM too hard on the divider. You will want to run memtest at least a few hours b/c you have pushed your RAM at really tight timings and it may be unstable. you are 100 MHz OC on your RAM at tighter timings. I think it might be your RAM. plus, you might need to boost vDIMM and vMCH
plus, why is your multiplier at 7x? why don't you set it to 8x? also, did you disable C1E and EIST? did you lock your PCI/PCIe clocks?

I have my E6400 @ 3.33 GHz. but you just seem to be pushing your memory way too tight. Try CL5 or keep your RAM at stock and try pushing your CPU first so you know it isn't the RAM. good luck.

I've stressed the ram at those speeds and timings for quite some time while the CPU was underclocked and it checked out. The vDIMM is at 2.3 and the vMCH at 1.35V in that particular cap. Apparently its fairly common for the D9 chips. Even when the ram is at stock speeds and the CPU has similar settings its unstable. I have the multiplier lowered because if I put it any higher it would have crashed far sooner.

C1E and the NX bit is disabled. I do however have the "Thermal Management Control" enabled. Doesn't that just step down the system clock and voltage when the CPU is idle? Should I turn this off?

PCIe isn't locked, but I manually set it to 100mhz

 

[PCTC2]

yes disable TMC, because it can cause instabilities.

 

[evetstech]

DFI 965P is very picky on things. Raise your VTT up, and your NB also, and you'll be able to achieve higher clocks. I'm using 1.37v for VTT and NB. Hope this helps.

 

[AWhackWhiteBoy]

I just shut off the TMC and gave the NB 1.55V and I've gotten farther than ever before, 2.56Ghz. What a typical NB voltage, esp when its passively cooled. 1.55V seems out of control, or has this been the missing link all along?

You sure I need to raise my VTT? I read up on VTT and GTL, they go hand and hand and this board doesn't offer much in the way of adjusting the GTL resistor network so its kinda moot. Did you notice any real different when it was lowered?

 

[CTho9305]

I think StopSign might be willing to trade you for another E6400. He can OC all of them (even E6300s) to 2.6GHz stably. If he isn't willing to, check with apoppin.

Originally posted by: StopSign
Every E6300/6400 will overclock to at least 2.6 GHz, which in my opinion is a "meaningful" increase in performance. Each and every one of these chips will reach that level at stock voltage and will be perfectly stable.

CTho9305, unless you can disprove the above, your argument is wrong. By disprove, I mean find me ONE E6300 or E6400 that failed to overclock to 2.6 GHz at stock voltage.

Originally posted by: StopSign

Originally posted by: CTho9305
Buy 10,000 E6300s. OC all of them to 2.6GHz (stock voltage / air cooling) for a year. If, after that time, any one fails to properly run any program I can find / write, I get all the CPUs. If after two years, the failure rate is more than double Intel's normal failure rate, I get all the CPUs. If they all work, I buy you 10,000 new equivalent-price CPUs. Would you take a bet like that? I sure as hell wouldn't.

Sure I would. You gotta be crazy to turn that deal down.

 

[evetstech]

Text Trust me =)

 

[AWhackWhiteBoy]

Originally posted by: evetstech
Text Trust me =)

Seriously, Thank you so much for the input.

After a few hours of tweaking I think I settled on my best options. For whatever reason the ram becomes unstable with CPU OCing so I thorttled it back and jacked up the CPU.

Best result so far is 3.2Ghz passively cooled.

These are the settings I settled on.


If I go any further the system becomes unstable and the voltage to frequency yield ratio is getting a little out of hand. Is 1.45V too much for a Vcore, esp when its passively cooled? The warmest it gets is about 60C, same with the north bridge.

Thanks again!

edit: BTW I'm going to do a better job stressing it tonight

 

[MarcVenice]

Passively cooled, with a huge casefan doing basicaly the same as the fan that's supposed to go with the HS. Not bad though

As for the voltage, I think you need 0.05 or 0.1 vcore extra for every 100mhz extra when the cpu won't go higher on stock voltage. Let's wait for someone who knows for sure, but it's been posted around here.

I wonder tho, WHAT did you change that allowed you to do 3.2-3.6ghz from your initial overclock that didn't let you even do 2.6ghz ? One more question, what's wrong with using the 8x multiplier? You do realize that the extra memory bandwith you're getting since it's running 1:1 will not yield you any extra performance, exept for stupid benchmark programs like sandra ? It's been tested before. The only reason that evetstech is using a 7 multi is because he has a e6320, and not a e6400

What I'm trying to say is, get the memory to run slower, but with better timings. That would be better.

 

[AWhackWhiteBoy]

Those are actually the memories stock timings. They can go up to 500mhz with those however when OCing the CPU I seem to be stuck at 470ish max. So I'm actually not loosing anything, and I'm more comfortable with the 1:1 ratio.

As for 2.6 to 3.2, it was the VTT voltage like evetstech suggested. I read an article that seemed to write off the option off as unncessary. I jacked it way up to almost max and it did wonderful things for my OC.

As for the HS, if I put the fan in it I only get a 2-3C difference under load. The 140mm PSU fan and 40mm deep 120 do wonderful things. This heatsink needs very little airflow anyway, it hardly gets hot. The copper base never gets warmer than 95F.

 

[MarcVenice]

I dunno what VTT is, so I looked it up lol, if interested, check it out, search for vtt, and stare at the picture http://www.intersil.com/data/wp/WP0538.pdf

As for your HS, hmz, interesting, I guess I should see what happens if I stop the fan on my ac 64 pro.

 

[evetstech]

I'm at 1.5V, of course I got it on water now, and my temps hover around 40-45C full load. A safe temp is around 60C. VTT is very important for clocking your memory and cpu higher, at least for the P965. The VTT, or FSB Termination Voltage essentially is used for connecting your cpu, memory, and other chipsets together, and have them sync together. When you try to clock your memory and your cpu at high rates, you'll need more voltage in order for the sync to keep up. That's my theory, I don't know if it's accurate or not, but It seems to be plausable.

 

[lopri]

nevermind. You did it already.