100celsius for Penryn Tjmax, so how close can you get to it and still be safe?

[GundamF91]

It looks like I am getting rock solid stability on the Yorkie Q9450, running 3.8Ghz at 72c, water cooled, and I think it can hit 4.0Ghz if I up the voltage more. The problem is the temp, I've always heard the redline is 70c or so, and the 45nm may be even more fragile to high temp. Even though Tjmax is 100c, would it be a bad thing to run the chip at 70+ temps?

 

[Markfw]

what is your current vcore ? You can easily kill yorkies over 1.4, and over 70c is not good....

 

[gwai lo]

You'll want to switch from reading the temperature to the delta to tjunction max. That's read the registers, whereas the "core temperature" is calculated by subtracting the delta by from the tjunction. If you read Intel's presentation, you'll find that it's not exactly accurate because each chip is calibrated individually at the factory.
http://intel.wingateweb.com/US.../SF08_TMTS001_100r.pdf

Of course, one of the slides also says that a 1°C change in real temperatures doesn't always translate to a 1°C change in the delta, nonetheless from what I understand (someone correct me if I have this all completely wrong) the "delta to tjunction max" will be the most accurate figure you should be paying attention to. With that said, theoretically you can run 1°C under Tjunction Max and be fine, however I think most people are maintaining a 10-15°C buffer to be safe.

 

[GundamF91]

I am using CoreTemp.99.3, so it has Tjmax fixed at 100c, thus Delta to TjMax is basically 100-72=28c. THere's still some room to go, but I am not sure it's healthy.

vCore wise, I have been able to get stable under 1.35v, so it's definitely doable, but Intel's own processorfinder says Yorkfield's upper llimit is 1.3625v

 

[VirtualLarry]

As I've indicated several times on here, I've been running a pair of E2140s at 9C to Tjmax for some time without any ill effects. Then again, if I do kill them, they only cost me $60 ea, so no big loss there.

 

[aigomorla]

Originally posted by: VirtualLarry
As I've indicated several times on here, I've been running a pair of E2140s at 9C to Tjmax for some time without any ill effects. Then again, if I do kill them, they only cost me $60 ea, so no big loss there.

65nm vs 45nm larry.

:T

You cant compare like that. I still have yet to kill my E2180.
I actually gave up trying to kill it.

Can we say energizer bunny? 😛

 

[Tempered81]

72.4C is Tcase for Dual core 45nm E7xxx & E8xxx, which is like "~26C-27C Distance to Tjmax". Intel would love for you to keep it under ~73C according to Tjmax of 100C. I think the Q9xxx 45nm is the same Tcase, I know the QX series are a little lower.

Originally posted by: gwai lo
you can run 1°C under Tjunction Max and be fine, however I think most people are maintaining a 10-15°C buffer to be safe.

I usually keep a 20C buffer minimum. This would be ~79C max on 100C TJmax. Larry's using a 10C buffer & as Aigo said, hes on 65nm Conroe.

If you want to follow Intel's rules then you'll need a Vcore of = 1.3625v & core temperatures of = 72.4C. Say by chance you do run to 1C under (aka 99C) then your processor will shut itself off, or throttle using TM2 & PROCHOTT before it lets you overheat it. I suppose that if you dive into the ranges of 73C - 99C you risk the chance of electromigration?

 

[GundamF91]

I think that it's an "either or" thing....either you run too much vCore into Yorkie, ie. 1.4v+ even if you have phase cooling to keep it cool, or that you get too close to Tjmax even if you don't pump up the vCore. So in either situation, you can prematurely cuase increased electromigration and degrade your processor so it would take more and more vCore to keep up the speed.

I decided to settle with lowly 3.6Ghz, 1.275v and cooler 65c temp (Linpack). I figure that should last me long enough until Sandybridge. I can't afford to buy another Quad should this one burn up.

 

[VirtualLarry]

Originally posted by: jaredpace
72.4C is Tcase for Dual core 45nm E7xxx & E8xxx
(...)
If you want to follow Intel's rules then you'll need a Vcore of = 1.3625v & core temperatures of = 72.4C.

No, you don't need core temps less than that, that Tcase, not Tjunction. Too many people seem to believe that. There a delta between Tcase and Tjunction, I believe it to be around 15C for 65nm dual-cores, it may be less for 45nm, I don't know. But that would allow me to get to a Tjunction temp of 73 + 15C, or 88C, before I start to bump into TcaseMax limits. So for me to load at 9C to Tjunction (100C in CoreTemp 0.96), that gives me 91C, so I guess I peaked a little bit over the limit. But I'm close. Suffice to say that I'm not worried about temps.

(Now that I got an AC, my temps are around 10C lower anyways.)

 

[Tempered81]

copied this from RT discussion:

Processor--power(W)--Tcase max--Casing IR(idle)--Tjmax----absolute temps at tjmax (+/- few C according to intel)
E7x00-------65---------74------------------------------100
E8x00-------65---------72.4-----------95--------------100--------97? or 98?+/- (on ones measured)
Q9650-------95---------71.4---------------------------100
Q9000-------95---------71-----------------------------100
QX9650----130---------64.5----------------------------95
QX9770----136---------55.5----------------------------85

Q6600 GO--95----------71-------------95-------------100?--------97? or 98?+/-
Q6600 B3 -105---------62.2-----------85--------------90?
E6850------65----------72-------------95-------------100?--------97?or 98?+/-
E6700------65----------60.1-----------85?-------------90?
E6600------65----------60.1-----------85--------------90?--------87?or 88?+/-

QX6800---130----------64.5-----------90?-------------95?
(compare to QX9650 with known tjmax)

E4400 L2--65-----------61.4----------85---------------90?

 

[wwswimming]

the rule of thumb we used on some radios for the government was -
if we use commercial parts, we got to keep the junction (the actual silicon)
at 85 C, or less.

so for a device cranking out 5 watts with a junction case thermal resistance
of 3 degrees C/ watt, the goal would be to keep the case at 70 C, or cooler.

wonder what ?jc (<== that's one FAT theta !!) is for the Q9450 ?

the 9450 is such a classic chip. & the P5Q Deluxe - good MB.

guess before i look up the spec - 9450 outputs 100 watts, ?jc is .1 (common
for bigger parts), temp rise case to junction is 10 degrees C.

i suggest to keep a thermocouple on the case & keep the case temp
below 70 C. except you'd have to modify the heat sink to attach the
thermocouple to the CPU case. i guess they have software for this, that
reads out CPU temp ?