high CPU temps under load

bonza32

Honorable
Apr 2, 2013
24
0
10,520
Greetings!
I have a old Compaq Presario SR1610NX that I was trying to tweak the last bit of performance out of (unfortunately, I'm quite poor on a limited income), so I saved my pennies and found a Athlon 64 4000+ San Diego CPU (AMD Athlon 64 4000+ Processor ADA4000DAA5BN) that I popped in to replace the original Sempron 3200+. Noting that the new CPU had a higher TDP, I purchased a Spire SP708S3-1 AMD Socket 754, 939, 940, AM2 Sleeve Bearing Cooling Fan w/3-Pin Connector (up to 4000+), which appeared to state it was sufficient.
When I tried to use Furmark's testing program to determine the performance gain by this upgrade, CPUID Hardware Monitor indicated a rise in CPU temp to in excess of 90C within 60 seconds...using Speedfan, I have confirmed that at near idle conditions, such as having a web page open to write this, the CPU fan is running at 100% speed of 4237 RPM and the CPU temp refuses to drop below 50C with a ambient temp of 39C inside the case...if I place a load on the CPU, such as playing C&C3, the CPU temp skyrockets to in excess of 80C.
Is this just a case of a insufficient CPU cooler? I would like to play around with this and try to OC it using ClockGen, but with these CPU temps while running at just 2.4Ghz, that seems to be quite dangerous.
Any suggestions (aside from buying a new computer, that is...too poor atm...)?
 
I used a HTK-002 CoolerMaster thermal compound kit after removing the original stock paste from the Spire. It gave me a little template and a spreader...I applied the paste to the square in the middle of the template using the spreader, then peeled the template off and attached the heatsink/fan to the motherboard.

I used Isopropyl Alcohol and a soft cloth to remove the stock paste, if that makes any difference...
 
According to the information from Speedfan, the high number (53C atm) is coming from the AMD K8 chip from the core sensor located on the PCI bus...there is a Ambient temp coming from the SMSC H/M Block chip on the Atilgp SMBus (38C atm), and a third called Temp coming from a LM75 chip located on the Atilgp SMBus (49C atm).

When I turn on the Furmark test, the first number immediately rises to 84C+ (within seconds), the second number doesn't change, and the third rises to 67C. Approx. 15 seconds after I exit the Furmark test, the numbers return to the previous readings.
 
It sounds like you did a good job of installing the cooler / thermal paste. It does matter what you use to clean and it's good that you cleaned the stock paste off. Isopropyl Alcohol is the best stuff to use.. Next time thou make sure to use a "lint free" cloth. A coffee filter is a good example. The lint can cause issues.

Double check your installation. Make sure the heat-sink / fan is securely fastened and all the way on. No gaps etc. You shouldn't be getting temps that high. Double check your temps with a second piece of software too.. Hwmonitor is good and often recommended.

If all else fails at that point I would re clean / re apply the thermal paste / heatsink. There's a chance that you got a bubble, or it didn't spread evenly and is causing it to insulate instead of transfer the heat.

I prefer the pea size method when applying. (you can see various youtube vids for it) Others have other methods.
 
Using CPUID Hardware Monitor, the max temp achieved is 106C for Core #0 while using the 60 second Furmark test...turning the system off now to check my install...will update after checking.
 
I took it all apart, re-cleaned and re-applied the thermal paste and then re-ran the Furmark 60 second test.

Results were slightly better...according to CPUID Hardware Monitor, during the test the CPU temp went to a max of 87C...quite a bit cooler then the previous 106C...but isn't that still too high for this CPU? It does appear to be cooler at near idle...about 47C...with CPUID HWM, and CPU-Z minimized, and this web page open...

Suggestions?
 
Re-ran the Furmark 60 second test several times...getting between 87 and 92C peak temp off the CPU...

Could it be a problem with the CPU? I did buy it used from Amazon...I tried to play C&C3 but it keeps crashing and the CPU temp soars to 100C+...
 
If the CPU was that hot (100C+), it would boil water, so to manually test if it's a sensor or not, while you're looking at the screen saying 87C put your hand NEAR the CPU (NOT on it)...and see if you can feel a LARGE difference in ambient temperature around the CPU...87C is 188-189F...so you would know the difference between that and the room. If there is not a significant difference, then it sounds like your sensor is bad.
 
You know, the fact that under load, with these high temps being reported, I can reach in there and touch the side of the heat sink...and it doesn't feel all that hot...has occurred to me. But then again, doesn't this CPU have a "under the cap" heat sensor...and if there was a failure of the heat sink to draw off heat build up, the two readings begin to make sense...the temp difference between the LM75 sensor and the CPU core sensor would reveal the inefficiency of the HSF/thermal paste combination and explain the rapid rise in CPU core temps under load.

Now that logically makes sense...what's causing the inefficiency is another matter entirely. As previously pointed out, the thermal paste application is critical to ensuring the proper efficiency.

Methinks my noobness at this has allowed me to apply way too much compound via the recommended "spread method...every time I take this apart, I apply a little less and use the "blob" or "dot" method and the "at idle" CPU core temps improved to near parity with the ambient temps...but the "under load" temps are still 86C via the "under cap" core sensor...while the LM75 indicates a considerably lower 66C with a ambient of 40C...within two minutes of applying the Furmark test...and those numbers continue to slowly rise the longer I apply the load.

 
You very well could need to reapply your thermal compound, yet again :)

Though, if you employ the "dot" method, and the temps don't fall inline, and your board/cpu/heatsink don't feel warm at all, then there could be an off chance that your thermal sensor is just out of whack...it happens sometimes on boards that a certain sensor is just way out of bounds or goes bad.

Though, I would recommend trying that method again, retesting, and seeing if the temp drops any. Also, what kind of case cooling do you have? The case may not be moving enough air internally as well, and that could cause internal temp readings to be slightly off...(though not a great deal...just a few degrees).
 
Well something isn't right...I tried a CPU load test (http://www.fossiltoys.com/cpuload.html) and the numbers started climbing...according to Speedfan's log, the CPU's core temp rose all the way to 110C before the amdk8.sys (AmdK8+0x3332) crashed with a BSOD code: 0x100000D1 (0x220CC, 0x5, 0x0, 0xFFFFFFFFF5D4168B) according to WhoCrashed...bunch of gibberish to me, but it might mean something to those more Enlightened in the "Ways of MicroSoft"...

This is just a standard Presario case...as I've been doing so much work on it, I've left the side off, so it's "open air" on the left side...

Looked up the thermal paste I've been using...CoolerMaster HTK-002-U1...and in comparison to Arctic Silver, it seems to be a much inferior product by a straight comparison of the numbers...if that means anything or not, I don't have the experience to say...

Ok, going to take it apart again and re-do it...I'll post back with the results.

BTW, THANK YOU ALL for your kind assistance!
 
Bad...as in insufficient? I've tried to find out more about this "Spire SP708S3-1 AMD Socket 754, 939, 940, AM2 Sleeve Bearing Cooling Fan w/3-Pin Connector (up to 4000+)", but its apparently discontinued...I have discovered it was called Speeze KestrelKing I and that it is a extremely budget priced product...read "cheap"...

That "up to 4000+" quantifier at the end makes me wonder if I purchased something that just can't do the job...
 
Ok, the "taking it apart and re-doing it" over and over isn't changing the numbers...every time I place a load on it, the CPU slowly gets hotter and hotter until the machine BSOD's...

So the problem is either the CPU is defective (even though it appears to test just fine and runs perfectly on near idle conditions such as now) under load...or the HSF/thermal paste is failing to remove the heat.

Once again, I'm not well versed in this, but what are the chances that a "defective" CPU runs great...except under load, when it gets so hot it eventually faults and the machine BSOD's?

Not being real knowledgeable in this area, I have no idea...so lets assume (I hate that word...) that the problem isn't the CPU...after all, it only faults when it gets really, really hot...

I've gone ahead and ordered a Cooler Master Hyper TX3 - CPU Cooler with 3 Direct Contact Heat Pipes (RR-910-HTX3-G1) with an additional fan (Cooler Master BladeMaster 92mm PWM High Air Flow Silent Case Fan R4-BM9S-28PK-R0), a "Y" connector to run them both off the motherboard CPU fan connection, Arctic Silver 5 Polysynthetic Silver Thermal Compound Paste, as well as two additional one GB sticks of PC3200 RAM....just in case I'm wrong about the CPU, I've also ordered at Opteron 180 dual core CPU, so unless I'm totally installing this setup wrong (after watching several videos of thermal paste application on YouTube...I doubt it...) I should be covered either way.

Once again I wish to thank you all for your help and when all these parts come in, I'll post how it all worked out. :)


 
A quick update here...while I'm waiting for my parts to come in, I saw something about a new type of thermal paste that supposed to be far and above everything else called Gelid GC-Extreme.

Now this stuff was claiming some pretty big performance, so I picked some up and tried it out.

For the last 15 minutes I've been running the CPU test from http://www.fossiltoys.com/cpuload.html on 100% load...and my CPU core temp is staying right around 60C...when it used to BSOD me in less then 5 minutes with CPU core temps in excess of 100C.

I'm impressed!

I can hardly wait to get my upgraded cooler and try OC'ing this Athlon using this stuff... :)
 
hi bonza32, forgove me if i have missed it reading through this thread but this could be a power supply issue. From what i have learned over two previous builds is that if the wattage of the power supply being used is not sufficient then this can cause severe overheating problems. This is because the chipset, cpu and gpu are all fighting against eachother for power which isnt there, this causes them all to get hot. what is the wattage of your power supply?
 
Brand new (less then two weeks old) Diablotek DA600 series 600watts...I got to say this new paste is QUITE impressive...using ClockGen, I pushed this Athlon 64 4000+ from 2.388Ghz to 2.448GHz (very mild, I know) but the temps stayed the same and it doesn't BSOD...I have to learn more about OC'ing because when I pushed it to 2.502GHz, I got a ntoskrnl.exe (nt+0x3A9A3)
Bugcheck code: 0x1000000A (0x2D8CB0, 0x2, 0x0, 0xFFFFFFFFE0BDC9A3) BSOD error code after load testing it for about 5 minutes...but the temps were steady at 60-61C for the CPU core. :)
 
Received all parts except for Opteron 180...won't be here for a few more days.
Installed two new 1G sticks of RAM.
Noticed voltage on RAM dropped from 3.3v to 2.5v on all four sticks of 1G...and timings, as indicated with CPUID SPD page, are all the same now...they were different with two sticks of 512 in Slot #3 and #4.

Using ClockGen, pushed CPU to 2.507Ghz (a ~9% overclock) and tested CPU with http://www.fossiltoys.com/cpuload.html for ~30 minutes at 100% load.

CPU core temp (via SpeedFan) stayed between 55 and 58C.

All appears to be well. I would have to say the GC-Extreme thermal paste made quite a bit of difference, but I do wonder at that voltage drop...less voltage=less heat, right?