[Mike360000]

Well I am out of ides, I just don't know what else to try....
After doing what I posted in the below links, I am not much better off with my cpu temps than before I started. Read the 2 links again too catch up on my problems and what I have done:

http://forum.oc-forums.com/vb/showth...554#post757640 (Last post!)

http://forum.oc-forums.com/vb/showth...threadid=94280

Basically it boils down to this, my cpu with an Alpha PAL 8942 on it with a YS Tech 48 cfm fan is still only bringing my cpu temps down to no less than 48c (stock & at idle) with my outside the box 8" fan picking up fresh ac air from the nearby vent, piping it to the back of my case in 4" flex vent pipe. When the ac isn't on my cpu temps stays around 53c. System temp with ac on and my fan picking up that air, runs about 31c and about 36c with the ac air off.

The outside fan air does make for breazy/turbulent airflow inside my case, which also now has a slight positive pressure from this forced air. I have a bay cooler containing 2x120mm fans, and my Enermax 465P ps, which has 2 fans. And 1-90mm fan case fan in the front bottom of my computer, which pulls in air.

I can't figure out why my cpu temps WILL NOT go down any further? I've put the heatsink on correctly and I had tried reseating my previous Sunflower h/F in hopes that would reduce my cpu temps, but NOTHING has worked! My vcore voltage is running about 1.69 to 1.72v. The only thing I know that does affect my overall temps, both cpu and system is that my room ambient temp usually stays between 74 and 78 degrees f. (Only when the room ambient temp is at 72 degrees f will my cpu go down to 49c!

I used AS3 and know I have applied that correctly also. I even tried my cpu at stock speed and it still ran average temps of 48c at idle! I thought about adding some inside baffles to direct the air more towards the motherboard and away from the bottom of the power supply's bottom pickup, fan and because of the way the Alpha blows out air instead of blowing it down onto the heatsink. My inside wiring is fairly pinned up and out of the way, although it could be better, and I have the round HD cables. But I just don't know anything more to try! So what is my problem? Would placing baffles help?

This is driving me nutz! I'd post some pics of my cooling if I had someplace, but I think I gave a good description of it in another of my linked posts. BTW I am running an AOpen H600A case. (Link to case http://www.aopen.com/products/housing/h600a.htm)

The 4" duct pipe runs into the back of the AOpen case, where the two add-on fans (1 shown in pic at AOpen site) supposed to be. I made a air-handling box that attaches to the back of my case and holds the duct pipe.

Cheers,
Mike Lamb

[Zendel]

I feel you pain. On the following system

P4 1.6a - cpu
Abit TH7ii - Raid - motherboard
2x256 Samsung RDRam

with 2 exhaust and 2 intake fans in addition to the power supply fan I am still getting idle temps of 45c and loaded of 60c+. On boot up the bios reports 50c. Is this a bad batch of chips or something?

[Zerstorer]

I wonder if its due to the temperatures being taken from within the CPU core?

Anyway I have a Chieftec full tower casing with 3 intakes and 3 exhaust fans. P4 1.6A@2.368Ghz at 1.9Vcore(measured)

Idle:
Case temp=34C CPU=50C
Full load
Case temp=38C CPU=59C

Using a Volcano 7+ at high fanspeed.(6800rpm).

I've assembled a few more setups for friends using BD-7 mobos and the temperatures are similar.

[Mike360000]

Well, I hate that saying, misery loves company, but at least I now know I am not alone!

But I do know this, which I stated in one of the posts in the links I posted. Once your P4 reaches a temp of about 66c, you're history as far as cpu performance goes. Because your P4 will start throttling back at that point and running Sandra's cpu arithmetic benchmark test will prove this. Before loading my cpu, I run the Sandra cpu bench and can got app. 4500 on the math part. Then I can run the cpu burn in (only) for about 12-15 loops and then run the cpu math bench again, and will get about 3500, IF my cpu temps reached about 66c during the burn-in test. Otherwise I will stay at about 4500 points.

So I am left with little choice but to down my oc and vcore volts until I can maintain cpu temps of less than 65c under a load, for say 1/2 hour, or for however long I set the Sandra burn in test for. If it wasn't for my cpu heating problem, I could run at 2.4 ghz +. I can bootup and run there now, but the heat is like 61c at idle!

Right now, sitting here with the ac running, and my outside fan picking up and piping the cold air to my case, I get the following temp redings in Hardware Dr.:

Idle
Case 36c/ cpu 52c
Geez, this is ridiculas!

Cheers,
Mike Lamb

[Mike360000]

Here's a shot of my fan. -- Mike

[Mike360000]

Here's a shot of my air handling box attached to the back of my computer. -- Mike

[macklin01]

Mike,

Thanks for your note alerting me to your new thread. That's quite a ducting setup! I didn't quite catch what your room and case temps were ...

I think at this point, where you obviously have a good amount of air coming and going from your case, we should try to investigate the flow pattern and get a feel for temps at various locations.

Here's an idea for making an "airflow probe": Get a chopstick (an excuse to order out!) and tape a piece of light thread to it. Insert it into various parts of your case and make a map of the airflow in your case. You might come up with various thicknesses of your arrows to represent the relative strength of the flow at one point or another (or vector length). It should give at least a 2D impression of the general flow.

Next, I'd recommend getting a cheapo $10 digital thermometer at home depot. The kind I hope for is the type that has a wire lead to a probe. (intended for indoor/outdoor temps on the same unit) Place the probe in different places in your case and see if heat is getting trapped in one place or another. Maybe all that flow has produced an "eddy" somewhere that traps heated air. It could be an interesting experiment, if nothing else ...

I'm sorry that I can't come up with more at this point ... I'll let you know if I have any other ideas. -- Paul

[Needitcooler]

I have come to the conclusion that either this board reads high, or we all got a bad batch of chips. I have done seemingly endless mods to my case, and the best idle temps I have ever gotten were 45 C. I even took the side of the case off and blew a huge living room fan right at the HSF, and that dropped my temps all the way to 44 C. I don't even turn MBM on anymore. I just deal with the fact that this whole temperature thing really sucks. If anyone ever figures out what the problem is, they would be my ultimate hero.

[davefred99]

Guys relax,
I had the same problem with my abit St6 mobo. it always reads about 7c-10c above My other Vh6t mobo or my Msi mobo for that fact. I know you guys are working with p4 mobo/combos and mine is a tuly but the fact is that many of the newer motherboards now use the on die internal temperature diode to measure cpu temps this is not the same as the on board socket theristor that most older boards use. There is a post somewhere in the forum on this fact and a link that will tell you how to calibrate your temp readings for better results. Basicaly it requires that you run your chip at its max stable oc and vcore and measure the ambient intake air temp under full load and also your cpu temp reading also under full load. Then set your bus speed as low as you can without changing your vcore and take the same measurements. Then divide your highest speed by the lowest speed, This will give you a ratio like 2to1 depending on your actual low speed setting and high speed setting. You then take your highest oc cpu temp and subtract the intake temp. Do the same for your lowest cpu speed settings. divide the higher number by the lower number and see if it is equal to the oc ratio. if not and it usualy isnt then your tmps are off. you them have to start subtracting from the two numbers until the ratio is the same. The final result will tell you how much to subtract from your internal readings for a calibrated tempurature.

[macklin01]

Originally posted by davefred99
Guys relax,
I had the same problem with my abit St6 mobo. it always reads about 7c-10c above My other Vh6t mobo or my Msi mobo for that fact. I know you guys are working with p4 mobo/combos and mine is a tuly but the fact is that many of the newer motherboards now use the on die internal temperature diode to measure cpu temps this is not the same as the on board socket theristor that most older boards use. There is a post somewhere in the forum on this fact and a link that will tell you how to calibrate your temp readings for better results. Basicaly it requires that you run your chip at its max stable oc and vcore and measure the ambient intake air temp under full load and also your cpu temp reading also under full load. Then set your bus speed as low as you can without changing your vcore and take the same measurements. Then divide your highest speed by the lowest speed, This will give you a ratio like 2to1 depending on your actual low speed setting and high speed setting. You then take your highest oc cpu temp and subtract the intake temp. Do the same for your lowest cpu speed settings. divide the higher number by the lower number and see if it is equal to the oc ratio. if not and it usualy isnt then your tmps are off. you them have to start subtracting from the two numbers until the ratio is the same. The final result will tell you how much to subtract from your internal readings for a calibrated tempurature.

That's a really interesting setup. A very reasonable experiment.

If I read your setup correctly, it is:

speed_hi / speed_low

= (temp_hi + delta - temp_amb) / (temp_low + delta - temp_amb),

where

speed_hi = overclocked speed
speed_low = stock speed
temp_hi, temp_low are corresponding temps
temp_amb = intake temp
delta = correction offset.

If so, then solving for delta gives

delta = temp_amb + (temp_hi - R temp_low) / (R - 1),

where R = temp_hi / temp_low.

This should speed up the process. -- Paul

[Zerstorer]

Originally posted by Mike360000
But I do know this, which I stated in one of the posts in the links I posted. Once your P4 reaches a temp of about 66c, you're history as far as cpu performance goes.

I concur with this. I encountered the same issues when I didn't seat my heatsink properly. Thermal throttling kicked at around 65 degrees and it kept the processor at a steady state of 75C for 8 hours even though that was barely 20% contact with heatsink.
Thank god for thermal throttling!

Idle
Case 36c/ cpu 52c
Geez, this is ridiculas!

Geez Mike,

You case temp is way too high for a setup with such good ventilation. Have you tried increasing the exhaust flow? I was thinking that maybe you have too much positive air pressure?

Coz my case temps are usually only 3degrees above ambient temp. I live in the tropics with 30deg C room temperatures, case temp is usually 33-34C.

[Zerstorer]

Originally posted by davefred99
Guys relax,
I had the same problem with my abit St6 mobo. it always reads about 7c-10c above My other Vh6t mobo or my Msi mobo for that fact. I know you guys are working with p4 mobo/combos and mine is a tuly but the fact is that many of the newer motherboards now use the on die internal temperature diode to measure cpu temps this is not the same as the on board socket theristor that most older boards use. There is a post somewhere in the forum on this fact and a link that will tell you how to calibrate your temp readings for better results. Basicaly it requires that you run your chip at its max stable oc and vcore and measure the ambient intake air temp under full load and also your cpu temp reading also under full load. Then set your bus speed as low as you can without changing your vcore and take the same measurements. Then divide your highest speed by the lowest speed, This will give you a ratio like 2to1 depending on your actual low speed setting and high speed setting. You then take your highest oc cpu temp and subtract the intake temp. Do the same for your lowest cpu speed settings. divide the higher number by the lower number and see if it is equal to the oc ratio. if not and it usualy isnt then your tmps are off. you them have to start subtracting from the two numbers until the ratio is the same. The final result will tell you how much to subtract from your internal readings for a calibrated tempurature.

I think this description can be found on the Artic Silver webpage. But rather tedious to apply and I'm lazy.

I'm of the opinion that our temps are actually correct as they are measuring the internal temps of the CPU(which should logically be higher) and not the surface temperature. In fact I'd rather work on keeping this highish measured temps low for safety than to use a probe on the surface(which some people did) and gain a false sense of security.

Nothing to worry about, as long as ur CPU is not throttling.

[macklin01]

Originally posted by Zerstorer
Thank god for thermal throttling!
That speaks loads for Intel. I've always viewed reliability as one of their chief selling points.
You case temp is way too high for a setup with such good ventilation. Have you tried increasing the exhaust flow? I was thinking that maybe you have too much positive air pressure?
That's why I recommended trying to map the airflow in the case. I'll bet there's not much dust, though.
Coz my case temps are usually only 3degrees above ambient
Not bad!! -- Paul[/B]

[davefred99]

Originally posted by macklin01

That's a really interesting setup. A very reasonable experiment.

If I read your setup correctly, it is:

speed_hi / speed_low

= (temp_hi + delta - temp_amb) / (temp_low + delta - temp_amb),

where

speed_hi = overclocked speed
speed_low = stock speed
temp_hi, temp_low are corresponding temps
temp_amb = intake temp
delta = correction offset.

If so, then solving for delta gives

delta = temp_amb + (temp_hi - R temp_low) / (R - 1),

where R = temp_hi / temp_low.

This should speed up the process. -- Paul

Heres a link
http://www.articsilver.cam/diode_calibration.htm

[macklin01]

Originally posted by Zerstorer
I think this description can be found on the Artic Silver webpage. But rather tedious to apply and I'm lazy.
A pretty tedious process. All I could do is simplify the math ...

I'm of the opinion that our temps are actually correct as they are measuring the internal temps of the CPU(which should logically be higher) and not the surface temperature. snip and gain a false sense of security.
I'd definitely agree with you there. I have a question -- does the PIII measure the internal die temp? I was under the impression it did, but now I'm second-guessing myself ...

Thanks! -- Paul

[macklin01]

EDIT: Oops -- sorry for the double-post! -- Paul

[macklin01]

Originally posted by davefred99

Heres a link
http://www.articsilver.cam/diode_calibration.htm

Thanks for the link. Hmm ...

Oops. Should be
http://www.arcticsilver.com/diode_calibration.htm -- Paul

Oh, and judging by that link, that expression I gave for delta was correct. Just be sure to mind the negative sign. (I.e., if delta comes out negative, you must retain the negative sign.)

[Zerstorer]

Paul,

I'm not sure if the P3 temps were derived from the chip itself. In fact, I'm slightly doubtful, coz I used to get 34C-43C temps on my P3-600E@788Mhz. Add to the fact the HSF did radiate a certain degree of warmth(though lesser) when felt compared to the P4(not very scientific) I was thinking that the 2 temperatures ought to be pretty close if they were taken from the same point.

Another case in point would be AthlonXP temps...my friend's computer has a reading of 40-50C yet the heat buildup in his case is obviously much higher than mine.(Plz don't tell me his HSF is much better..:P)

[macklin01]

Originally posted by Zerstorer
Paul,

I'm not sure if the P3 temps were derived from the chip itself. In fact, I'm slightly doubtful, coz I used to get 34C-43C temps on my P3-600E@788Mhz. Add to the fact the HSF did radiate a certain degree of warmth(though lesser) when felt compared to the P4(not very scientific) I was thinking that the 2 temperatures ought to be pretty close if they were taken from the same point.

Another case in point would be AthlonXP temps...my friend's computer has a reading of 40-50C yet the heat buildup in his case is obviously much higher than mine.(Plz don't tell me his HSF is much better..:P)

Destroyer,

Thanks for your response. I'm not sure. The temp changes have always been fairly instantaneous with my CPU readings in MBM5, which makes me lean towards internal diode. It may well depend on teh version of PIII, too. (I have FCPGA-2, not a slot version.) Also, I can see no trace of a thermistor on the motherboard.

Out of curiosity, I emailed the motherboard maker to ask. I'll let you know if I hear back from them. Always interesting stuff to find out ....

Oh, and I don't think he has a better HSF; His temps are on the order of 10C higher, which is consistent with the higher heat buildup you sensed on his AMD system as opposed to yours. Maybe I read this all wrong?

Thanks again for the response. -- Paul

[Mike360000]

Macklin01,
Phew, you gave me a lot of info.
First my temps again:
My room/ambient temps usually run about 25-26c.
My case temps ran 40c or higher (idle) before my mods and now since the mods, they usually run 32-34c. (I have seen lower, but only when I set the room ac lower than 25c and when it is currently blowing so my case fan picks up the cold air.)
My cpu temps have varied least of all, from 55-56c to 50-51c after mods. (all idle).

Now consider loaded temps: (This is taken with an oc of 128 fsb, DDR 266 ='ing 2300 mhz - cpu. Vcore set to 1.75v. Reads 1.69v - 1.71v in HWM.)
Before mods:
Case/System 44-51c.
CPU 66+ (System would kick out of Sandra burn-in. Games would do the same, especially UT which is heavily cpu dependent.)

After mods I could run my entire system at 129 fsb/2320 mhz, and have slightly better cooling.
After mods idle:
Case/System 31-36c.
CPU 52-54c

CPU/System under load after mod:
Case/System 46c is the highest temp I've seen after a 25 loop Sandra burn in test.
CPU 64c (Same test)

However tonite I did one more easy mod, just to test a theory.

If you can pic in your mind how the P4 sits on a BD7 mobo, and how the Alpha blows air up and out into the case instead of blowing down on the heatsink. And then picture my vented/piped air coming in at a complete/direct 90 degree right angle, right into the path of the blowing fan on the heatsink, and it seemed to me the air from the duct was never fully reaching the mobo. This seemed to make me think the heatsink would not be getting a good path of direct air from the vented in air. So I put a baffle, diagonally, from the left rear corner of the case, sloping down to the left (nearest) side of the fan, making sure the vented air would not be caught up in the air being expelled from the heatsink fan. At the same time this directed air not only to the heatsink itself, where it would be sucked through the heatsink, but it also put air directly down onto the mobo, including the chipset. (I did a trial attachment with the baffle using duck tape, of course, until I am sure of the results.)

Well so far tonite the results have been encouraging. While my idle temps have not changed at all, both my loaded temps for my system and cpu have went down by 4c each! Plus I am running 1 mhz higher, at 129 fsb, and .25v higher! And this was at 26c room/amb temp! Still the idle temps has me stymied.... Why doesn't my idle temps go down? (I ran a 24 loop Sandra cpu burn in to check the system.)

I think DaveFred99 had it right about not worrying about, along with his suggestions, but I do want to run a decent oc and keep temps to where I know my programs will not crash or my cpu not throttle back and lose performance.

I have done a little sheetmetal work in the past, as 1 of my degrees is a 2 year degree in HVAC. But I'm afraid all those figures on averaging and testing the heat exchanges is far more time consuming and troublesome than what I could gain from just experimenting. Like I said I am beginning to firmly believe ANY temp is fine for the system and cpu as long as it does not start to crash programs or throttle back the cpu. My main thing now is to see how well I can tune my airflow to get the most from my cooling to my cpu and system, REGARDLESS to idle temps. And like I said I've already seen factory off the shelf P4s that run almost as hot as mine. (My stock speed tests concerning temps did not vary on idle more than 2c from my oc'ed idle readings. Although with my system loaded the stock settings would be from 2 to 4c cooler. Still this is high considering what many other people claims.)

Finally concerning my fan/blower, I can get another model at Wally World's that looks and fits almost identical to the one I already have, for 5 bux more, but it *supposed* to be a stronger fan, moving more air. The only reason I didn't get it to begin with was that Honeywell wasn't specific on the airflows of each fan. Although the ads say 1 has more power as well as being a 3 speed. I did make 1 other change to my fan also. I added a light dimmer switch to control the fan speed ans on/off settings. It works fine, keeping me from going under my computer desk.

Cheers,
Mike