Prescott Throttle Temperatures?

I'm currently using a 2.8E in an asus P4P800SE motherboard. I can run the cpu at 3500mHz with stock cooling just fine stable for hours of ut2004, Farcry, Doom 3, or Sandra Burn-ins... no, I'm not prime95 stable, but I suspect it's my memory holding me back from attaining that.

Anyways, my problem is that Asus PC probe and MBM both never report my temps higher than 58 C. That is the hottest I can get my processor to register, and the times it does hit 58, it almost always goes back down to 57 or 56 immediately. Turning off the "throttle CPU" option in bios has no effect on this maximum temp.

I know that Asus boards tend to report temps 5-10 C lower than actual, so I've basically just been asuming that my actual temperature is somewhere around 68-70C.

My question is this, and I hope someone can help me: At what temperature does a 2.8E Prescott start throttling itelf? I want to know this so that I can figure out exactly how inaccurate my temperature readings are at load.

Where did you read that your board reads temps 5-10 C lower than usual? My Dad has that board and he gets temps up to 66 C on full load. So would a more acturate estimate of his temps be around 76 C on load? A reason or a point in the right direction would be good. Also if you know what the max recomended temp is for the prescott line that would be great.

Yes, Asus boards do report temps 5 - 10 degrees lower than real temperature is. Most P4s slow down at 75 - 80 degrees but Intel itself says some Prescotts may have this pushed higher, depending on their thermal characteristics. But so far all P4s I had, including some ES and Prescotts, slowed down when reaching this temperature.

If you want to know whether your P4 / PM slows down, download CPUMSR and see page Thermal Monitor. You can also turn duty-cycles off or set it to custom level.

EDIT: With Vcore increased, Thermal Monitor may not work properly, not slowing down even in situation where overclocked CPU crashes system.

Petr great program. I have just run it on my machine (AMD 2500) and under the features tab it says that 'System Call Extension' and 'System Data ECC Checking' are disabled. Is there any point in enabling them for me or will it just cause problems?

Petr said:

EDIT: With Vcore increased, Thermal Monitor may not work properly, not slowing down even in situation where overclocked CPU crashes system.

Click to expand...

This is very true. No thermal "automatic" or "built-in" protection is a substitute for proper protection (ergo, proper cooling) and monitoring. Also, when overclocking, higher voltage means less temperature tolerance for both stability and CPU lifetime.

Excellent article on the innacuracies of thermal monitors:
http://www.bleedinedge.com/reviews/abit_asus_temps/abit_asus_temps_01.html

I downloaded CPUMSR and have read through the documentation. Very useful program, thanks for the link. I've yet to try this, but it seems as though if I disable Thermal monitor 1, my CPU will no longer throttle itself at high temperatures. So if I select this option then run a load test, and my temp goes higher than 58C, I can be relatively certain that that is the temperature that my cpu is throttling itslef. But the original problem still remains... is there a way to determine the ACTUAL temperature that the processor thermal protection control kicks in?

So, as I understand what's been said, there isn't a set temperature where the 2.8 prescotts start throttling themselves, and that the throttling temperature is dependant on the thermal characteristics of the individual processor? I would have throught that this would be a value that would be the same for all processors of the same model/stepping. Am I going to need an extenal thermometer to get this value?

Once again, my goal is to discover the difference between what my Asus Probe thinks the temp is, and what the actual temp of my processor.

This utility should save some people a lot of time if they're trying to do the same thing I was. It grahs CPU load, as well as throttling in realtime. Very Handy when testing OC's on pentium 4 procs:

http://www.panopsys.com/throttlewatch.htm

Oh, and just a note, using CPUMSR to disable TM1 is very very very not reccomended.

I was just coming to post that link here . I'll still put it in its own thread .

I just used that throttlewatch proggy and if i oush my pressie over 60c it throttled.

You don't need a proggy to see TM1 thermal throttling. Just open your task manager's performance tab, and pull down 'view' and tick 'show kernal times'. This will show your OS kernal usage in the color red, displaying from the bottom of the the 'cpu usage history', while the normal cpu use will still remain in green.

When TM1 thermal throttling occurs, the OS' red graphing will rise, with a corresponding fall in the cpu's green graphing, and it will be in a continous fashion similar looking to 'noise' on an O-scope. As soon as the TM1 throttling stops, the red and green graphs will seperate, and become stable. It's a lot easier to spot than I explained, and once you've seen it once or twice, it's easy to identify.

FWIW - My Pressy C0 3.2 @ 3738 (267 x 14 - using the 'cpu unlock' feature in the BIOS) starts to TM1 throttle at ~ 60*C on my P4P800SE, as displayed using AsusProbe. This is with the 'cpu internal thermal throttle control' set to disabled in the BIOS.

With 'cpu thermal throttle control' set to enabled, the throttling begins around 58*C, as displayed in AsusProbe. This second form of throttling does not appear to be the usual Intel TM1 thermal throttling I've seen over the last 1-1/2 years. I believe it may be the relatively new externally controlled 'On Demand' throttling. I'm not sure if it is implimented to benafit the mobo manufacturers or Intel.

Disabling it, will not intefere w/the TM1 throttling inherent to your P4, and will gain you several degrees centigrade, and some x-tra performance and stability at the very top end, if you're running on the edge (like me - running 2 Distributed Computing clients concurrently, 24/7).

BTW - Anyone wanting to mess w/a Pressy on air, would be best off going for the D0 steppings, or waitng for E0, rather than using a C0, IMHO. My 3.2 C0 is an easy 5 - 6*C hotter at both idle & load, than my 3.2 D0, plus the D0 OC's slightly better due to its cooler running, when using identical cooling set-ups. The nice advantage w/the 3.2 CO, though, is the ability to 14X multi (if your mobo's BIOS supports it), but keeping it cool on air, w/a nice nice OC, is an absolute B****!!!

HTH

Strat

Ooops, sorry -

Didn't address the OP's original question:

P4C/E's generally begin TM1 thermal throtlling bertween ~70 - 75*C, as measured directly in the center of the IHS. For your chip this appears to be 69.1*C.

ref: Intel Spec finder filter for P4E/800MHz/90nM/IMB/S478.

(click on your s-Spec# for the 'thermal spec')

The reason your Asus mobo reads low is 'cuz it measures temps at the socket, not the IHS or die. You are are definitely thrermal throttling at 58*C, slipping back a few degrees, resuming normal operation, and then returning to thermal throttling. This is TM1 throttling, and is internal to the cpu, so cpu-z or any other s/w utility that measures your cpu's clockspeed will not show a decrease in cpu speed. But cycles will be dropped internally w/i your cpu, and you DEFINITELY will be taking a performance hit. Since the temp is measured by the Asus onboard diode, the diode is your reference point, and you'll get the same 'incorrect' temp measurement, irrespective of your s/w temp monitoring utility.

I'd highly recommend using a good aftermarket HSF combo (or other superior cooling methods), so you can enjoy the full performance of your cpu. If you'd like to see how badly your performance is due to the TM1 throttling, try running some long benchmarks or a distributed computing client. Something that will run for 20 - 30 mins. Then replace your stock HSF w/a good aftermarket one, and rerun your benchies. I strongly suspect you'll see a significant performance improvement.

BTW - Complete thermal shut down occurs at 135*C die temperature. This prevents catastrophic failure. Not sure what temp that equates to at the IHS, but you should never run into it, unless you've got a misaligned or missing HSF.

Been there.

GL

Strat

Thanks Stratcat for that Link to the Intel Spec Finder... I googled for that information for months and came up with nothing.

My Processor throttling at 69.1C sounds just about right. Confirms what I've been thinking all along. My estimation of 68-70C was errily accurate.

I have a 3.2 prescott mobile, right now its divX encoding at 73C and hasnt throttled a bit according to that program.

Ive yet to see it break 73C, as when it hits 74 it drops to 67 and back to 73...but no throttle occurs, that i know of. Being a mobile, maybe it clocks down instead, and the programs dont register that. Stupid prescott.

I assume the temp sensor hits its max at this point....

Noob question. What is thermal throttling? I have a setting on my Abit board's BIOS that allows me to set it at 25%, 50%, 75%, etc. What does that do?

Trottling makes CPU automatically downclock itself when overheating.

That is however not one hundred percent true . Until the temperature gets extreme on Pentium 4 chips , throttling will not affect the clock speed . So there will be no clockspeed drop . What happens is that the CPU will begin to introduce waitstates into its cycles and thus decrease the work done per cycle and thus reduce the energy both used and dissipated ( the CPU will basically have cycles where it does nothing or at least much less than it would do at 'full speed' ). This of course would also reduce the performance of the chip or its effective speed , long before the actual MHz starts to be affected . This is why many people don't realise that their P4 is throttling , because the MHz or GHz speed looks the same .

Hamm3r said:

Trottling makes CPU automatically downclock itself when overheating.

Click to expand...

So if I am looking for a 250 fsb and have only been able to hit 248 before, should I disable it? My temps are always under 61 deg C. so I expect my CPU has never been throttled.

Edit: Thanks CowboyX, we overlapped on posting. I guess it would have no impact on my fsb OC.