WTF it can't be that hot!

[peach]

Anyway – here’s the deal. This is background info, for impatient types, skip to the 3rd paragraph. A few months ago I decided to upgrade my computer a little - mainly to get rid of the noise. It was a 1.2 axia o'clocked to 1.4 on KT7ARAID in a custom painted & modded to hell & gone case that was noisier than a nail factory (~13 fans - with two of them 120x38 sunons). At the same time I acquired another 1.2 athlon. What to do? Ah hell build me another one with then anticipation of the thunderbirds, stuff the guts of the old one in a new quieter Al case and set up the old case with an as cheap as I can find it mobo and an old duron I had laying around. So I got a couple nice matching aluminum cases and set'em side by side with a kvm. I built the new computer around a gigabyte 333ddr raid motherboard. Now in the wind tunnel (my old case as I affectionately referred to it as) my case temp was room temp ~72ºF and proc usually ran around 78ºF (84ºF under ful load). Of course that’s the abits motherboard version. But in the new case and no longer oc'd (same mobo and proc and hsf) it jumped to 130ºF - I can't remember exactly I just remember it was very high. It had a Sh!tty hs and loud fan so I replaced it with the same one I was matching in my new computer, a full cu evercool with a aluminum 60mm 4600 rpm fan on top. Having both running at the same time, their temps still were running in the mid 120's. Hmm. I am using artic silver and I am pretty positive I am doing everything right (this is like the 8th year I have built computers - no expert but certainly not a beginner). So I ducted the intake to the proc to the front of the case (wasn't using those bays anyway) and added an intake fan in front, an exhaust fan in back behind the proc and a slot cooler below the proc in each comp. Ok now temp in old faithful (with new skin) is down two around 105ºF and in the newbie is around 118ºF. For whatever reason I had put off getting the newbie up to speed for a few months. I had gotten xp and some standard apps going, but hadn't bothered moving the 60 gigs worth of personal info and tweaking everything. Anyway, my wife’s laptop somewhat died and I figured ok ok i'll demote the old faithful and start moving everything over.

So I am running my new computer with virtually the same apps, the same case, same heatsink, same processor, but a newer mobo with better ram: cl 2.0 ddr333 vs cl2.5 sdr133, and a sweet quadro dcc video card (yeah baby). Not much really different as far as stability and temps go right? And of course it is running a little hotter @118ºF. Now I do punish my computers with a lot of multitasking and intense cad work, but i'll be damned this thing is crashing all the time. ~5-20 min on the files (thank god for autosave, I have it set to 1 min). And ~5 times a day with a full hang (no mouse movement) and another 2 or three blue screens of death (haven't seen those since win98) granted xp pro isn't as solid as nt4 but this is really bad. So I install mbm. Wow they've improved it. Hmm I see my temps are even higher than I thought there around 124º and 125º. Sh!t. I guess I nee to do something. So I pull the duct and case sides off. 129ºF. Crapola. Its late. Wtf why not. I pull out a trusty 120mm sunon and fashion a pretty good temp reducer to the front end of the duct. I cool this mofo and find out if its crashing b/c of temp or not. For all I know it might be the ram.

So here’s the final situation: an all cu evercool 60mm square thin finned hs, with one an aluminum 60mm 4600 rpm fan but also ducted with a 100+cfm 120mm sunon at the mouth of fresh 72.5ºF air intake. The hsf fan is spinning at nearly 5000 rpm. heh heh. Anyway, the damned thing is reporting 118ºF. I say no way. How likely is the temp probe on the gigabyte mobo wrong?

Is it possible that my hs is screwed up? As a mechanical engineer I understand that pretty much it all comes down to hs area, hs material, and the all important airflow. This hs is fairly large, is all cu, and i hva got about all the air one can imagine stuffed down its throat. wtf?

BTW I have been typing away this post after having done this mod and have had no crashes and the temp under full load and under no load has stayed almost a rock solid 118. I really belive this probe is wrong. Of course this mod can only be temporary as it is very loud.

Thoughts anyone? Crash, you still around? I am not sure its not the video card drivers either (same as geforce3 detonator driver). And of course, how do I tell if it is bad ram? The ram is GeIL 2.0 supposedly handpicked by OC system with a cu spreader. Case temp was ~98ºF when crashing, now about 93º with case open and all.

On a separate note, how effective are the Tt smart II fans. I am thinking of making the switch to that and keeping the duct. Also, you guys think a switch to 2600 would be noticeable enough for the purchase?

<font color=blue><i>On Company time...

 

[lhgpoobaa]

I would say there is a pretty high likelyhood that the temp sensor is wrong, or its reading a different one... If its reading from the internal sensor then its bound to read higher.
If its the external one then +/- 10C is common... Hell one guy posting here reckons his temp goes DOWN when he overclocks... hows that for screwed up.

Ive tested out a thermaltake smartfan 2... its a nice enough fan. Its a bit weak moving air and buzzy at its slowest speed. So buzzy that i ad to return the first. The replacment was better but still not perfect.
The advantages of course is that it has a VERY wide range of working rpm... And to be honest it sounds a bit sick at 1500rpm... just too low for it really. It sounds best at medium 2000-3000rpm.
At full bore its allmost as loud as my 4900rpm foccused flow 80mm delta screamer! But moves slightly less air!

So if you want something thermally controlled or manually adjustable then this is the fan for you. (the temp probe is long and thin... great for squeezing in tight spots)
IF you JUST want a slow quiet fan or just a high powered fan then other alternatives are better.

One final comment. If you think your ram is playing up, run MEMTEST86. A very useful and free tool. Great for detecting bad ram and ram thats been overclocked to much as well.

<b>My Idea of a DOS attack is to stop posting helpful information at Toms Hardware.</b>

 

[peach]

Thanks for the tips. I read that thread on the temp goign down with oc. it made me laugh, but not as much as the page I was just reading - the <A HREF="http://home.planetinternet.be/~bbaeyen2/thb/index.html" target="_new">Barton 3000+</A> spoof on thg. Damn that was funny.

Thanks for the tip on memtest86. I d'loaded it but it requires a floppy and mine has never worked. I never really need it except on rare occasion like this. i wonder if i can trick it to writting to a small hd partition and using flipping that to a bootable cd with nero. hmmm maybe.

If its reading from the internal sensor then its bound to read higher.

uh, i know exactly what you are asking, but I thought there was no internal diode in an Athlon 1.2 The mbm notation is a little cryptic but I belive this is from the sensor touching the cpu from below. I can understand it being low from lack of good conduction and variation of temperature from the die and the bottom of the cpu but reading high? 10º high? I would guess that not to be true but am now hoping that's the case.


<font color=blue><i>On Company time...

 

[lhgpoobaa]

Opps. your right. Those Tbirds had no internal sensor.

As for it reading high, the temp reported really depends on two things, the airflow under the socket, and the distance of the little diode to the bottom of CPU core. In certain instances the diode is high enough to actually touch, giving massively higher temps than one a few mm away.
Also ive seen the use of waterblocks give higher undersocket temps than standard fans... as the waterblock has no fan to move air around the socket area.

Even different manufacturers have done things differently. ASUS boards are well known to report higher temps than other brands.
When i had to replace my ASUS KT133A with an Iwill one, my "apparent" cpu temp of my tbird 1200C dropped 6-8C. Same case, same cooler, same small amount of arctic silver. Go figure.

So my good man, the temperature being reported is really meaningless unless you have some refrence point. The best refrence point is the temperature at which it locks up or crashes.

I knew on the Iwill board my 1200C would stop working at ANYTHING above 52C period. So so long as i kept it below 52C i was happy



<b>My Idea of a DOS attack is to stop posting helpful information at Toms Hardware.</b>

 

[peach]

<font color=blue>

In certain instances the diode is high enough to actually touch, giving massively higher temps than one a few mm away.

<font color=black>I think I actually bent the sensor up to ensure it would touch. I figured if it was touching - it would be that more accurate, right? or did they put in a buffer assuming it wouldn't touch? Gee I didn't think of that....
<font color=blue>

the waterblock has no fan to move air around the socket area.

<font color=black>I wouldn't think air could move in there at all, even in the wind tunnel ;-)
<font color=blue>

So my good man, the temperature being reported is really meaningless unless you have some refrence point. The best refrence point is the temperature at which it locks up or crashes.<font color=blue>

<font color=black>well said. and apparently, with my temp set upas it is, it's rock solid at 118, crashes a little at 122, more at 125 and unbearably at 127. okidoki.



<font color=blue><i>On Company time...

 

[lhgpoobaa]

Well if i had an under socket sensor i think i would want it touching the cpu surface... get the most accurate temps you can.

118 is a little close to the temp taht you start seeing crashes, so maybe consider slightly better cooling... unless of course it never gets above 118.

<b>My Idea of a DOS attack is to stop posting helpful information at Toms Hardware.</b>