E6600 Temp

[sjandrewbsme]

I'm getting the following temps (according to Core Temp):

Idle: 32 Celcius

Orthos small FFT (CPU Test): 64 Celcius Max (after about 1/2 hour)

The room temp is about 70 degrees F. Does this sound about right? The reason I ask is that I've become very hesitant when it comes to my Tuniq tower and the torque at which I seat the HS. My second P5N32-E SLI just died, and I think it might have been from a shorted CPU HS bracket (this is just a guess - one day it's suddenly shorted on the PW-ON header and a tiny trace behind the bracket looks like it might be scratched). Since then (the first P5N32-E SLI died from a blown MOSFET which has nothing to do with heat - but it's made me scared none the less) I've been very timid when it comes to overtightening the bracket.

It makes sense to me that more torque would result in better temps - but I'm hesitant to tempt fate. I think these numbers are pretty decent - any input is appreciated.

 

[BonzaiDuck]

I'm currently testing the limits of my E6600 on the Striker motherboard with Crucial Ballistix DDR2-1000's.

I'd been using Passmark CPU Burn-in Test for the preliminary testing, and the load temperature would peak at around 36C -- (up from 25C idle).

When I ran Orthos at the same settings, the monitoring software showed a peak of 50C.

Since I'm using one of the best coolers available, I was stunned. I know the Ultra 120 beats your Tuniq on thermal resistance, and you'd expect a load temperature of a few more degrees. But the comparison would depend on things like our respective VCORE and over-clock settings.

The advantage of the Ultra's design -- it uses spring-loaded screws, and you simply tighten them down all the way. The design of the back-plate makes it impossible for the screws to tension the motherboard in any harmful way.

 

[BonzaiDuck]

. . . . even so . . . . when I installed the cooler, I was also a bit shy about the bracket. There are no solder-pins, traces or solder-joints on the back of the Striker where the bracket fits, but I put a layer of Black silicone RTV gasket-maker on the back-plate, and after it formed a "skin," I shaped it level with the rubber grommet of the backplate that surrounds the caps opposite the processor.

 

[sjandrewbsme]

My only overclock settings are:

1500 FSB for CPU frequency of 3370. I'm using DDR2-1000 @ 5-5-5-15-2T (Ballistix) @ 2.2V. Other than that, everything is "stock" - no other FSB or voltage changes.

I've read mixed reviews on the accuracy of Core Temp - I haven't tried much else in terms of monitoring. According to the BIOS, the CPU Temp is 28C when I boot.

What I did for the backing plate is instead of re-using the double-sided tape (again), I mounted a whole bunch of rubber pads (the little bumper feet for the bottoms of components like CD Towers). I've had these laying around for years - I metered them to make sure they had infinite resistance. I was thinking about doing something along the line of RTV but this way worked out quite well - it distributes the mechanical load quite well.

Orthos blend gives me 50C - only the CPU test gets my proc > 50. Prime95 gets to about 52. I also have my CPU fan at about 30% - where I can't hear it. I've tried running it faster (where it's quite a bit noisier) but I haven't seen a big difference with cooling. I'm also using AS5 thermal grease. I expect it to improve over the next couple months as the AS5 "sets".

 

[BonzaiDuck]

Just to share some reliable gossip here.

I read a review and some forum posts that had the quality of "white-papers." The forum post was an extensive test of limits with an E6600 on my Striker Extreme board. The poster -- try "Eva2000" @ i4memory.com -- checked BIOS settings against BIOS monitor (and windows monitoring which gets the numbers from the BIOS monitor) -- and checked all of those readings against actual multimeter tests for available voltages.

Her work confirmed that my Seasonic PSU is not really running 3% under the requirement on the three rails. All those voltages are understated by the Striker board, and her readings confirmed for me that my rails are all just about right on the money. But in addition to the three PSU rails, someone else confirmed that for certain earlier BIOS versions, things like VDIMM, while monitored values are less than "set" values -- were actually a tad higher than set values.

These observations were also true for the P5N32-E, AS I REMEMBER(!) It was an ASUS board using the 650i chipset.

In my later BIOS revisions from those tested by "Eva," ASUS apparently corrected various bugs in the discrepancies of VCORE, VDIMM and other voltages, but the monitored values are still a tad under the set values. And apparently, the actual values come back in line with the set values.

You're using the same memory that I have, although I refuse to twist the voltage up to the warranty limit, and am adhering to a self-imposed limit at just short of 2.1V.

I'm also limiting my VCORE to 1.475V. [Eva2000 cranked it up to 1.525V in order to reach 3.6 Ghz FSB QDR=1600, DDR=800, 1:1 @ 400 external frequency.] I need to go back and look at the temperatures she was getting, but I'm sure her tests were done under air-cooling.

I can only get to around 375 Mhz - FSB = 1500 1:1 with DDR=750 at the processor's stock multiplier -- under my self-imposed voltage limits. But I can drop the timings of the Crucials down to 4,4,4,8 at either 1T or 2T -- and they may still have some slack in them.

But I'm still mystified -- given the thermal resistance of my cooler -- how my load temperatures with these over-clocks at VCORE 1.475V are pegging at 50C, but the CPU idles at 25C (given today's room ambient.)

Certainly, there's no doubt -- Orthos really puts the screws to your system -- other tests I've used don't push the CPU "load" value beyond 38C.

 

[sjandrewbsme]

Orthos (Small FTT or whatever it's called - the CPU test) is the only one that can get my CPU > 53. Orthos blend pegs @ 52 ish as does Prime95.

Right now I think I'm satisfied leaving it where it is. I just updated BIOS, I might try it again and see if Core Temp displays a different temp reading (I went from version 0602 to 1002 - four versions newer iirc).

I've never messed with VCORE in the hopes of additional overclocking. I figure that when I find an application that needs more "oompf" I might. For the time being, 3370 is plenty.

FWIW my PSU is a 500W Enermax Liberty. Also, my HS is lapped (not the heat spreader on the CPU, just the HS.

 

[thilanliyan]

I get higher temps with Intel TAT (about 5C higher) than I get with Orthos. However, any errors always pop up in Orthos, not TAT.

 

[sjandrewbsme]

TAT doesn't run on mine - I've tried multiple links to what I presume to be the same source file, but they all don't work.

 

[BonzaiDuck]

I want to stray "off-topic" here.

Per the test programs, I've noted the following anomalies which "bother" me.

PassMark CPU Burn-in Test, Orthos, and S&M all provide the wrong info about my processor when it's been over-clocked with a different muliplier than the Intel default (9). So if I set the multiplier to 8, and the nVidia monitoring software shows a CPU speed of 3.0 Ghz, Orthos et al show 3.6. It either takes the existing multiplier and makes assumptions about the external frequency to falsely report the CPU speed, or it gets the true external frequency and applies only the default multiplier.

PassMark at least recognizes the processor as "Core 2 Duo," even though version 5.1 was released before C2D became available.

But there seems to be some disconnect between software that otherwise works properly and presents accurate information, and the nVidia chipsets for Intel processors.

On the matter of VCORE settings -- and this might be the same with the P5N32-E SLI board. The Intel voltage range specification for the E6600 is 0.85V to 1.35V. I set up my Striker board with the default "Auto" settings initially. BIOS and Windows monitoring reports that the "set" value for "Auto" vCore (and other stock settings) as 1.44V.

Per the article at i4memory.com by "Eva2000" -- which I mentioned above -- you can get to 3.6Ghz with the 9 multiplier if you're willing to set your VCORE at about 1.5V. But that just seems too high for me, and the temperatures would certainly be higher than those we discussed here without more attention to cooling.

In the range of VCORE settings in my self-imposed limit, there is a very slight rise in idle temperature between a "default" Vcore value and the upper limit in my range. When you go beyond that -- it gets . . . . . more noticeably hotter.