Xtreme Low Q6600 G0 Temp on Air !!! (Lots of Pics)

[Spanki]

...Tjunction - DTS = coretemp and I even had a screen per the demand asking to show the Tjunction remaining from the coretemp just to show that the core temp is actually picking up the 100c Tjunction properly.

I think you must have missed my earlier response to this, but it's worth repeating...

Note that what I highlighted above may be misleading. The only thing your math proved was that CoreTemp was in fact using 100C as it's Tjunction reference value - not that 100C is the value it is supposed to be using for this stepping.

But we already knew that it was using 100C, because it was listed on the window . The question at hand is whether or not 100C is the proper value. Note that Tjunction is not a register on the chip that CoreTemp can simply read to find out. It's basically looking to see which cpu is being used and using a hard-coded value, depending on whether it's a mobile, conroe-duo, allendale, quad, etc.

So, until Intel verifies what value should be used, we just can't be certain that 100C is correct.

What makes me think it's changed for the G0 stepping? Here's my reasoning...

1. Temperatures being reported for the new E6x50s are way off - obviously 85C is no longer the proper Tjunction value to be used for those.

2. If you look at the Intel alert/notice about the stepping change for this particular cpu (Q6600), you'll see the following:

Tcase for the Intel® Core™2 Quad processor Q6600 and Intel® Xeon® processors X3220 and X3210 on G-0 stepping has been increased by 11 oC. Tcontrol offset will remain the same relative to increase in Tcase which will help reduce acoustics

...also note that 'Tjunction' is also (and perhaps more properly) known as 'maximum Tcase value'. So, what Intel did with this stepping (in addition to some optimizations that allow lower overall voltages) was to "let the cpu get 11C hotter, before cranking up the cpu fan (and/or throttling itself)". This leads me to believe that the Tjunction value was also increased - maybe by 11C, but possibly by some other number.

I'd also note that (at least speaking for myself), there is some confusion/ambiguity about how all of these values relate to each other... Tcase, for example, is (as far as I know) the temperature probe in the center of all cores and is the one that the BIOS and most apps report as 'cpu' temperature, but the way the above is worded, they seem to be specifically talking about 'Tjunction' (or TcaseMax)... they've obviously increased the top-end value of the allowable temperature range.

Given all of the above, I think it's a fairly safe assumption that the 'proper' Tjunction value to use for a G0 stepping Q6600 is 111C - but we really just won't know until Intel shares this information one way or the other (the only thing they've said for certain so far is that the mobile cpus use 100C - at least at the time they said that).

So basically, I don't think we can say for sure what the core temperatures truely are, but for practical/discussion purposes, using CoreTemp's numbers is as good as anything else until/unless some clarification comes from someone who really knows. As mentioned earlier, the only truley meaningfull measurement is whether or not your system is stable at some given speed. If it is (and the voltage you're feeding it is not dangerously high), then it doesn't really matter what the temperature is.

That being the case, the more meaningful number that CoreTemp can give you is the actual DTS value (enable the "Show Delta to Tjunction" option). This tells you how much headroom you have left - regardless of what all the various temperatures are reading. XT showed this @idle earlier, but idle temps never have interested me ... I'd like to see some DTS values for some of these overclocks while under load.

[XtremeTiramisu]

Please look at my post above
I just edited it.
I hope this will clear out things

[fornowagain]

...also note that 'Tjunction' is also (and perhaps more properly) known as 'maximum Tcase value'.

snip

I'd say that's close, but Tjunction and Tcasemax are not the same thing. Tcasemax is the maximum temperature for the IHS. Its on page 74, here, 62.2C for the B3 Q6600. The Tjunction is the on die temp, Tjmax is the maximum die temp. Tcontrol is the delta to the the thermal trip point TCC and has a negative direction -ve, as the the core heat increases the Tcontrol tends to zero. TCC is the throttling temperature. There is some debate, but Tjmax is not TCC. Tjunction is a generic value for a cpu type and the TCC value is the factory set value and can't be read externally, probably within a few degrees. So programmes like coretemp are accessing a register or assuming a value, if thats assigned correctly or reflects the actual Tjmax is a different matter.

Tcase for the Intel® Core™2 Quad processor Q6600 and Intel® Xeon® processors X3220 and X3210 on G-0 stepping has been increased by 11oC. Tcontrol offset will remain the same relative to increase in Tcase which will help reduce acoustics

That's saying the Tcasemax is now 73.2C (71C), the throttling point has the same offset as before so TCC has also increased by 11C. The short of it is that accurate absolute values are problematic, different programs use different values. Coretemp could well be reading the wrong register values for Tjmax and its not the 'actual' throttling point. Without an overheating throttling chip and a thermocouple attached, its still going to be guess at an absolute value.

Have a read on these links it might clarify a bit on the finer details. The first one is where I've tried to explain it as best I can; but you're right the deltaT via DTS is all that really matters. In techno speak its, "PROCHOT# is activated when DTS=0 for the maximum safe operating temperature."

http://forums.overclockers.co.uk/sho...8&postcount=14
http://www.xtremesystems.org/forums/...d.php?t=131008
http://www.xtremesystems.org/forums/...d.php?t=136804

One point though, if you know the temperature of the air blowing over the HSF and you know the approximate thermal resistance of said HSF. Then for stock voltage and speeds then the chips wattage will be close to the curves specified by Intel specs. If you know that lot, then you "should" be able to calculate an approximate value for Tc. At least then you'll know how accurate the PECI bios temps are! There's a worked example in the link above if you can be bothered.

The only thing I'd add, which may or may not be obvious is...

- A G0 stepping Q6600 at xxxMHz @1.5v very likely runs the same temps as a B3 stepping Q6600 at the same xxxMHz @ the same 1.5v.

If the G0 has a lower vcore at default that the B3 and ends at the same vcore at a higher clock it could end up using MORE power even with a lower TDP. The additional power used comes from change in voltage squared. Obviously its not very likely as a lower default vcore generally means a lower vcore for the same clock.

1. Q6600 B3 @ 3.6GHz 1.275v stock, 1.50V oced.

Pd = 105*(3600/2400)*(1.5/1.275)² = 217W

Suppose for a G0 a lower stock vcore and TDP, frequencies the same, so its the change in voltage that counts. Its using more!!

2. Q6600 G0 @ 3.6GHz 1.20v stock, 1.5V oced.

Pd = 95*(3600/2400)*(1.5/1.20)² = 222W

Now if it uses less voltage, but the same overvoltage delta. Which is fair if its the same process.

3. Q6600 G0 @ 3.6GHz 1.20v stock, 1.425V oced.

Pd = 95*(3600/2400)*(1.425/1.20)² = 200W

[Decami]

Been reading this forum for months now. I always do that before i join to get the jist of the atmosphere. These forums are awesome I must say. Great info from well rounded people. This thread had alot to do with registering lol

anyway, to the point.

Amazing OCs Tiramisu. looks like the G0 is actually all its cracked up to be. Being that ive been staring at close to the same rig you have for the past month now, it makes me excited. 10 days till order form and counting. I'm getting excited.

May I point out. I have been staring at the Antec 900 for quite some while now and reading up on it. That may have alot to do with your temps also. Since this case while not the greatest is amazing at air flow, being where the heatsink sits sorrounded by fans with amazing intake from the front.

Nice to join you guys. and GL on 3.8-3.9 Tiramisu. lol

[Spanki]

I'd say that's close, but Tjunction and Tcasemax are not the same thing. Tcasemax is the maximum temperature for the IHS...

---- lots-o-good-info-snipped ---

Thanks for the comments and additional links - interesting stuff.

[unclewebb]

1. Temperatures being reported for the new E6x50s are way off - obviously 85C is no longer the proper Tjunction value to be used for those.

I compared the reported load temps of my B2 E6400 to a new G0 E6550 and at the same MHz, core voltage and room temperatures, our load temps were almost identical. For both processors, TjMax is assumed to be 85C.

The idle temperatures for his E6550 were being reported below ambient which is impossible but my B2 E6400 can also report idle temperatures below ambient. That doesn't prove that the maximum Tjunction isn't 85C. It might just be that the data coming from the on chip digital temperature sensors isn't accurate at low temperatures. These sensors are designed and calibrated for managing the thermal throttling and thermal shut down of the processor. No where does Intel document their use for measuring idle temperatures.

My opinion based on readings from an IR thermometer is that they are very accurate from 50C to 100C. Below 50C they lose accuracy and when used to report idle core temperatures, will usually report readings that are too low.

I believe the TjMax for the new Q6600 is likely 100C as CoreTemp 0.95 is assuming but without any documentation from Intel, I'm only guessing and so is anyone writing temperature software.

None of the above matters if you set CoreTemp 0.95 to "Show Delta to Tjunction temp." CoreTemp is accurately reading that Intel documented register and displays it in real time.

If you're doing some serious overclocking and you're running stable you really don't have to worry too much about temperatures. When fully overclocked, you'll lose stability before you reach the throttling point and you won't get anywhere near the THERMTRIP# shut down point.

[XtremeTiramisu]

I compared the reported load temps of my B2 E6400 to a new G0 E6550 and at the same MHz, core voltage and room temperatures, our load temps were almost identical. For both processors, TjMax is assumed to be 85C.

The idle temperatures for his E6550 were being reported below ambient which is impossible but my B2 E6400 can also report idle temperatures below ambient. That doesn't prove that the maximum Tjunction isn't 85C. It might just be that the data coming from the on chip digital temperature sensors isn't accurate at low temperatures. These sensors are designed and calibrated for managing the thermal throttling and thermal shut down of the processor. No where does Intel document their use for measuring idle temperatures.

My opinion based on readings from an IR thermometer is that they are very accurate from 50C to 100C. Below 50C they lose accuracy and when used to report idle core temperatures, will usually report readings that are too low.

I believe the TjMax for the new Q6600 is likely 100C as CoreTemp 0.95 is assuming but without any documentation from Intel, I'm only guessing and so is anyone writing temperature software.

None of the above matters if you set CoreTemp 0.95 to "Show Delta to Tjunction temp." CoreTemp is accurately reading that Intel documented register and displays it in real time.

If you're doing some serious overclocking and you're running stable you really don't have to worry too much about temperatures. When fully overclocked, you'll lose stability before you reach the throttling point and you won't get anywhere near the THERMTRIP# shut down point.

Another thorough explaination unclewebb, thanks for posting it here. This is my first time I see you on XS and you rock over at the [H] forum