Safe temp for Q9300

[Avrithor]

Since the consensus seems to be that the Q9300's TjMax is 95c, that means that it should be safe to run it at up to 70-75c since that's still 20c+ distance to TjMax, correct? That still seems super hot to me but if TjMax is supposed to be the temperature where the chip shuts down to avoid damage then 20 degrees is a healthy margin, yes?

 

[unclewebb]

TjMax is not anywhere near where these chips shut down. At TjMax or 95C these chips will start to throttle themselves and slow down and maybe even automatically reduce the core voltage depending on the motherboard and how you have it set up. Thermal shut down doesn't happen until approximately 125C.

One user sent me a screen shot where he made it up to 123C and was still able to get a screen shot of his Q6600. All of the Intel mobile core based chips have a documented shut down temperature of 125C. The desktop chips are likely the same but that number is not documented by Intel. Most software craps out long before the processor does.

You don't have to be concerned about temperatures when running an Intel Core based processor. Your computer will lose stability long before it ever has a melt down. These chips are built extremely well with multiple safety features. As long as you have a heat sink attached and a fan more or less running, you'll never have a problem.

 

[Avrithor]

Wow, that is an amazing screenshot, unclewebb! Very interesting info. Does the higher temperature still have an impact on the expected lifespan of the chip, though? Or is it pretty much the case that as long as you're at a safe voltage and the system is stable, you're all set?

edit: As an aside, I notice you have an E6400 running at 1.5v+...I was under the impression that over 1.45 isn't advisable as the max specified if you look in Processor Finder is 1.5. Mine is currently at 450x7, 1.45v...the board can pull a higher FSB easily but I'm nervous about giving the CPU more volts. What's the best way to determine what's truly safe for 24x7 operation? (I mean, obviously yours hasn't melted into goo yet.)

 

[BillParrish]

I will attempt to avoid one of my 'book sized" posts.

The short answer to your question is "it depends on if replacing the CPU would be a issue".

As Unclewebb has shown, it is almost impossible to destroy a c2d with temp, it will protect itself. Voltage is a different issue and "youse pays your money and takes youse chances with voltages over 1.5 ish volts (depending on CPU family and the Intel Data sheet Absolute max Vcc rating you should look it up for your chip. Intel.com ).

On thing to keep in mind is that per Intel spec (and that assumes whomever built your motherboard bothered to meet the spec) voltage "overshoot" by the CPU voltage regulator circuit when going from light to full load can be 50mV or .05 volts. Making it easy for the CPU to see 1.55 volts if you set (and it happens to be accurate) a Vcore of 1.5 volts. These are very short transients voltage spikes that can over time stress the transistor junctions leading to bad things happening, or not.

Its pretty much un-arguable that the cooler electronics runs the longer it lasts. But these things are so cheap now and testing is so expensive (and by the time the testing is done the part is obsolete anyway) all you can do is "accelerated" testing with temps very high, look at the fallout and try to extrapolate backwards for less extreme conditions. I haven't seen anything recent publicly available. Per the Intel warranty you can run a c2d at a delta of 1C to Tjmax and they will warranty it for 3 years. So back to the "can you afford to replace it. My advice, just use common sense and burning or blowing up a chip for an extra 100MHz when spending the time/money tweaking the video card would yield so much more system improvement (assuming gaming is the primary use) seems silly to me. However I admit to doing some really really silly things from time to time.

What you really need (like me) is a cpu . with a higher multiplier so we can run fast without the insane FSB speeds that stress the entire system.

 

[Avrithor]

I will attempt to avoid one of my 'book sized" posts.

I don't mind book sized! I ask in order to learn.

The short answer to your question is "it depends on if replacing the CPU would be a issue".

As Unclewebb has shown, it is almost impossible to destroy a c2d with temp, it will protect itself. Voltage is a different issue and "youse pays your money and takes youse chances with voltages over 1.5 ish volts (depending on CPU family and the Intel Data sheet Absolute max Vcc rating you should look it up for your chip. Intel.com ).

On thing to keep in mind is that per Intel spec (and that assumes whomever built your motherboard bothered to meet the spec) voltage "overshoot" by the CPU voltage regulator circuit when going from light to full load can be 50mV or .05 volts. Making it easy for the CPU to see 1.55 volts if you set (and it happens to be accurate) a Vcore of 1.5 volts. These are very short transients voltage spikes that can over time stress the transistor junctions leading to bad things happening, or not.

Its pretty much un-arguable that the cooler electronics runs the longer it lasts. But these things are so cheap now and testing is so expensive (and by the time the testing is done the part is obsolete anyway) all you can do is "accelerated" testing with temps very high, look at the fallout and try to extrapolate backwards for less extreme conditions. I haven't seen anything recent publicly available. Per the Intel warranty you can run a c2d at a delta of 1C to Tjmax and they will warranty it for 3 years. So back to the "can you afford to replace it. My advice, just use common sense and burning or blowing up a chip for an extra 100MHz when spending the time/money tweaking the video card would yield so much more system improvement (assuming gaming is the primary use) seems silly to me. However I admit to doing some really really silly things from time to time.

What you really need (like me) is a cpu . with a higher multiplier so we can run fast without the insane FSB speeds that stress the entire system.

Aha. So, per Processor Finder then the max VID voltage I would want to set for a Q9300 is going to be 1.31v to account for potential 0.05v spikes, maybe a little more (though the mobo is an ASUS, it's likely within the specification, right?). Keep the VTT that low too probably as I've heard these 45nm chips are sensitive to it, and not lose too much sleep if the delta to TjMax is a bit low. I mean if Intel still warranties them up to 1c below TjMax that tells me they're confident in the chip's ability to operate at that temperature.

 

[Down8]

I got my Turion TL-50 up to 102*C. Accidentally left it on a blanket on the couch for an hour or so. Oops.

I've seen 80*C+ on my E2180 recently. Come to find out the heatsink was bolted down on one side - oops again. Not that it's solid, it won't go above 50*C on a warm day.

 

[unclewebb]

For 24/7 use I wouldn't go over 1.40 volts for the newer 45nm CPUs or over 1.50 volts for the previous 65nm chips. I like to keep my chips for a while and my original P3 - 450 MHz CPU is still in service for internet use.

Temperatures are rarely an issue as long as you have a heatsink attached. It's more important not to get carried away with too much core voltage. Even one high voltage run might be enough to permanently damage a CPU and the 45nm chips seem to be a little more delicate than the previous generation. If you can't afford to buy a new one, use some restraint.

 

[chrisf6969]

assuming that 70-75C is under 100% load (on all 4 cores), then you're fine.

As your processor will probably average a single digit % load over the course of its life!

I mean you rarely ever hit 100% load, UNLESS you're folding@home with the multi-core version of it or sometimes while encoding. But even gaming often only loads a processor like 50% +- a lot of the time.

Use your processor as normal, and leave a temp monitor app in the background, your max temp will probably be more like 50C+- since you're never going to have the quad that bogged down.

Oh yeah, and what uncleweb said : < 1.4v for 45nm (your chip), & < 1.5v for 65nm

 

[Avrithor]

Okay, excellent. The Q9300 machine I'm working on is a friend's, to be used for gaming at LAN parties. I set the core voltage to 1.31v and have so far left the FSB termination voltage to auto, and I've already been able to overclock it quite a bit with load temperatures still in the mid 60s running two instances of Orthos to stress all 4 cores. I definitely won't go any higher on the vcore and just see how far it goes on 1.31 and leave it at that to be safe. Thanks for the info! I have a Q9450/P5Q-E on the way for myself so it'll come in handy for that too.

 

[BillParrish]

Actually the "overclockers voltage" is the Absolute Max Vcc from the datasheet, you are looking at the range of "stock" voltages Intel ships the chips at. For a Q9300 Vcc max is:

http://download.intel.com/design/processor/datashts/318726.pdf

page 19, table 2-2 1.45 volts.

Me personally, I would stop around 1.375V as a personal limit, IF MY OC NEEDED THE VOLTAGE.

If your current results make you happy there is no reason to go higher. I just wanted to clairify for anyone else that wandered into this thread what the datasheet maximum was. Your approach is very conservative and has yeilded good results so no need to push it unless you just want to. At your current voltage setting there is absolutely nothing to be concerned about.

 

[Avrithor]

Okay, now the issue we've run into is that the system passes 24+ hours of Orthos (2 instances w/affinity set so all 4 cores get utilized, Large FFTs) just fine, but it crashes in games. So that's going to involve bumping the northbridge and/or southbridge voltages by very small increments until it stabilizes, right? If it's stable for Orthos that shows the processor and the memory at least are good to go, it's when you bring the video card into it and fire up games that it has issues (the video card itself is on stock settings, no OC). The motherboard is an ASUS P5E-VM HDMI, G35 chipset. Is there a datasheet somewhere I can look up to find out what the max voltage for the chipset is?

EDIT: Aha, I found it! Under "Absolute Minimum and Maximum Ratings," "MCH Core," "VCC," it gives 1.375v. Now I just need to look up the ICH9R sheet for the southbridge.

EDIT: Oof, the ICH9 datasheet is harder to read. Any help here?