djstx2003: I had a feeling your water temp was a little warmer than you thought it was!
It's definitely wrong to adjust TjMax. I've continued to work with Xello and he's also seeing a 4C difference at full load between the two sets of cores within his Quad. That's very normal and during testing seems to continue right up to TjMax and beyond.
I was getting a consistent 5C difference but the CPU fan was likely on a very low setting to get things this warm or maybe even off some of the time. The point is that difference is normal at full load. At idle this delta should decrease to virtually zero. Surprisingly, I took the heatsink off once and let it get up to a similar temperature but the difference between sets of cores decreased to next to nothing which sort of proved the sensors are OK. I think it has to do with how the IHS contacts the cores but I haven't been motivated enough yet to gut a good processor in the name of science.
The odd part about your CPU is that the difference is fairly uniform from idle to full load. That doesn't usually happen.
My official test as sort of outlined in the documentation suggests using 333x6.0 ~ 2000 MHz for 45nm and about 1.10 volts. Similar voltage is more critical than MHz. Looking at your reported idle temps and your water temp, I wouldn't bother using a calibration factor on core0 or core1. They are both reporting about 41C to 42C at idle which considering your water temperature, is probably a very accurate core temperature. A water cooled Quad will usually idle about 5C above your water temp during my official test. There might be a slight variation depending on the quality of your set up and 101 other factors but this is a good baseline to shoot for.
I'd just use negative calibration factors on core2/core3 (approx. -2.0) to bring their temps at idle down so they equal core0/core1. I think this will provide you with the most accurate approximation of your core temperatures from idle to TjMax. Leave TjMax set to 95C for all of them.
That's my best guess. Your sensors are far better than most Quads. Maybe after an RMA they reach into a bag of processors with good sensors and give you one of them to keep you happy.
Making any comparisons to what Everest or PcProbe or SpeedFan report for the CPU temperature isn't usually a good idea because you don't know how accurate that sensor they are reading is. The Intel DTS sensors are excellent as long as you know where they are accurate and where they are not.





Reply With Quote

Bookmarks