Penryn QX9650 maxed temp 70C. tCASE, or tCORE?

[Noel 4,486]

I notice Core Temp displays a temp that is roughly 10 deg higher than either ASUS Probe II or Realtemp display. Realtemp's display can show 4 separate "core" temps as does Core Temp, but these four will display as the same temp, suggesting to me Realtemp is actually measuring tCASE.What do you know?Noel

[Guest D17S]

If you have 2 separate monitors showing 4 individual cores at approximately the same temps, we can be relatively assured these temps are "Tj" (Temp junction). These are the temps that are taken directly inside each - individual - core. Often you will see these temps cluster in pairs with up to 10C splits. This is normal. -- These are the critical temperatures to watch -- . Intel also placed a temp sensor in a position that is in the center of their "case" (the CPU die). This is called Tc (Temperature, case). Intel placed this sensor equidistant from the individual cores under the IHS. This was provided as a quick reference for users wanting a single temperature reading to monitor their CPU temp.A server farm operator has no time to monitor each individual core. This is the temperature Intel provides for normal industrial operators. The single Tc sensor is positioned at a physical distance from the 4 Tj sensors. The temp at the Tc location will be Cooler than the temps at the 4 Tj locations because of this distance. For a Q6600, Tc will be approximately 15C cooler than the Average of the 4 Tjs. For other CPUs, this difference is 10C. I expect this is the 10C difference you are seeing. However in any case, Tc is irrelevant. It's just a quick reference "surrogate" for the individual Tjs.The only way to know what's going on is to know what's going on. I found Asus Probe II to be entirely inaccurate. It appeared to me they intentionally skewed their Tc number "scary high" so a novice builder would run that temp "safety-low." They did this with Vcore too. Their bios overstates Vcore by .5v. That's a lot BTW. And their Probe thinggy showed Vcore another .5v above that. Funny game they're playing. Try speedfan.http://www.majorgeeks.com/download337.htmlThat'll give you another Tc vote on what's what.

[Mace 1,902]

Thanks for the heads up on Asus Probe II. I use SpeedFan but hadn't made comparisons.Looks like I still have lots of headroom in the temp department...RhettFS box: E8500 (@ 3.80 ghz), AC Freezer 7 Pro, ASUS P5E3 Premium, BFG 8800GTX 756 (nVidia 169 WHQL), 4gb DDR3 1600 Patriot Cas7 7-7-7-20 (2T), PC Power 750, WD 150gb 10000rpm Raptor, Seagate 500gb, Silverstone TJ09 case, Vista Ultimate 64ASX Client: AMD 3700+ (@ 2.6 ghz), 7800GT

[Guest Nick_N]

Speedfan may need to be calibrated and may not reflect the correct core tempits best to use a verification software which is certified by the software maker to read the diodes correctly and place the offset into speedfan if needed. Since Intel still seems a bit skidish on releasing data in this area you do need to confirm the software, like Core-Temp or others have been updated to read the processor you use correctly.Asus Probe is well known for being untrustworty and innacurate around the CPU and I have never used it for that purpose

[Guest nd55b]

The most reliable measure of Tj is Real Temp:http://www.techpowerup.com/realtemp/Core-Temp, Speedfan, and a few others overstate their temperatures by about 10C on the 9650's (and relatives).

[Noel 4,486]

>The most reliable measure of Tj is Real Temp:>>http://www.techpowerup.com/realtemp/>>Core-Temp, Speedfan, and a few others overstate their>temperatures by about 10C on the 9650's (and relatives).But, as I say, when Core Temp was showing different temps that made sense for different loads on each core, RealTemp showed 4 cores at exactly the same temp. That is suspicious, and moreover, it is the same as with ASUS PROBE II. Thank you all for your input.Noel

[Guest D17S]

Truly suspicious. I've always seen all 4 bounce around independently. Two will be higher and the other two will be lower.Try Everest's system stability test screen. It's quick, easy and is close enough.

[Noel 4,486]

>Truly suspicious. I've always seen all 4 bounce around>independently. Two will be higher and the other two will be>lower.>>Try Everest's system stability test screen. It's quick, easy>and is close enough.We are in the Mother of All Wild Fire seasons, so I've not been able to run my AC cooler except during non-smokey days. So, I've found with a fan blowing directly over critical parts with the case cover off, a regular 12" house fan that is, my Tj temps from Core Temp are coming in as high as 74C. So, not running overclocks to the same degree I can with AC on. With my AC on I am running 4.2Ghz with Tj's staying under 60C which is awesome.

[Guest D17S]

So what's Tjmax on the those 45nm Qs? The big Qs (65nms) are fine well into the 80s. Tjmax is 95-100C. Point being, maintaining 60C is entirely unnecessary. Any temperature below Tjmax is safe and sane.Intel integrated these thermal throttling limits (Tjmax) to - Prevent - damage. Any temp below Tjmax Will Not damage the core. After all, they Did have a choice. Would Intel have put these limits beyond the damage zone? What kind of idiocy would that have been? However on the other hand, Intel is actually an acronym for INTegrated ELectronics. Clearly (a response might present), this makes No reference to INTELligence! However consider: If throttling was occurring After damaging a CPU, we'd easily see this in quarterly revenue reductions for warranty expense . . . and it would be Big. These thermal auto-throttling limits were put in place to accommodate their major sources of revenue, industrial users. They buy Billions of $$ of these things . . . then install them in undercooled server farms. Intel had to protect itself from this user group. If a core approaches a temperature where damage will occur (i.e., a warranty event become at least arguable), the core will declock or entirely shut down - Before - this damage occurs. Again, the core declocks or entirely shuts down - Before - this damage occurs. Worrying about thermal damage is what Intel did, so We don't have to! If this thermal protection mechanism did not work, we'd see it in INTC's stock price . . . in a hurry.That said, consider that a jet engine has this same overtemperature safeguard functionality. However an operator would be Nuts to simply bang a thrust lever to the wall and hope for the best. During an initial TO roll, the PNF (pilot not flying) has a beaded stare on EGT as thrust is advanced to TO. However the crew does Not set power to 500C if the engine's power is rated at 725C (Yea, let's leave 200K lbs of freight on the ramp!) . . . nor do they simply sit and stare at EGT for the rest of the flight. If power sets and is stable, they move along to other aspects of flight . . . like flying the airplane (or using ones computer). A responsible operator will use the engine's designed-in and safely-available power. Likewise, a knowledgeable computer operator can responsibly use their CPU's entire thermal envelope.Set your clock to somewhere below Tjmax. I really expect the Frigidaire will end up being entirely unnecessary. If it's stable at comfortable increment below Tjmax, move on to "flying the airplane." Coddling a core at 60C simply misunderstands the functionality that is built into these CPUs. Go ahead and use your machines (CPU's) capability. It was designed for that purpose. BTW, my sound problem magically went away (right around the time I installed the latest Nvidia driver). I'm back to 3.6Ghz at Tj(avg) hi-70sC. All's well. Have you tried Tampa's Kai Tak X? Getting 15FPS in the 744X . . . but it's a good (stutterless) 15.

[Guest ziporama]

My Gigabyte X38 board reports a 10C difference in temperature between cores 0/1 and cores 2/3 on the QX9650 so what you are observing matches my own experiences - and makes no sense to me.I typically don't trust the on-board sensors because they are not typically accurate, and have next to no documentation on how they operate from the manufacturer. A calibrated thermocouple or infrared imaging device would be much better, but then again, two devices that aren't around the typical household, combined with the fact the cooler sitting on top of the CPU doesn't make readings particularly easy :)I'm running my 9650 at 4GHz, 1.45V stable with 4Gb of memory. If I remove 2 GBs, I was able to go up to 5.1GHz (10x 510MHz FSB) on water cooling and 1.6V, relaxing timings on my Corsair memory. Thinking the 4x1Gb was a problem, I tried a pair of 2x2Gb OCZ with same specs, and never was able to run a stable torture test beyond 405MHz FSB at 10x multiplier. There is something about going to 4Gbs that isn't related to the slots on the board in my case.Still, at 4GHz, FSX rocks...Cheers,Etienne

[Noel 4,486]

>So what's Tjmax on the those 45nm Qs? The big Qs (65nms) are>fine well into the 80s. Tjmax is 95-100C. Point being,>maintaining 60C is entirely unnecessary. Any temperature below>Tjmax is safe and sane.As a side benefit though, in my opinion, are higher overclocks and cooler memory, NB and GPU. Higher temps, more electrical resistance, higher volts required, I thought that's how it worked. Plus, it's kinda warm in this room right now, since it's 100 degrees outside :() I like cool myself, but your point is well taken. Yes I am aware of the protection schemes. I've heard 100C is Tjmax, but Core Temp shows 105C.>Intel integrated these thermal throttling limits (Tjmax) to ->Prevent - damage. Any temp below Tjmax Will Not damage the>core. After all, they Did have a choice. Would Intel have put>these limits beyond the damage zone? What kind of idiocy would>that have been? AMD were idiots then, since they actually produced processors with no built in protection. I guess intel did to in the past.

[Guest D17S]

Oh I'll bet if they coulda, they woulda. Reducing warranty expense is the real driver of this autothrottle technology. We just get CPUs that can't hurt themselves. True 'nuf that increasing temp will increase resistance, that will require increased volts to maintain a flow rate (amps). For instance, measure a light bulb's resistance with an ohm meter. WaY less that 1 ohm, maybe? The flow rate (amps) implied by a < 1 ohm resistance when pressured by 120 volts should blow the main-breaker of the City of Chicago. You make a very accurate observation. It's the immediate heat occuring in the filiment that increases the resistance so your house doesn't explode into a fireball when you turn on a table lamp. However, consider your CPU premis: Is the maintenance of a flow rate (amps) the determiner of the ability to generate a clock rate. Kinda doubt it. Transistor operations are more based on pressure (voltage) levels. But this is sooo beyond me, it's simply pure entertainment (for me) to even think about it. Thanks for the opportunity!