Are BIOS temps idle or load?

[Len]

Just put my new system together (see my sig below) last night. Haven't installed Windows yet but was tweaking my BIOS settings. I'm using a generic cooler "rated" for up to 1.2 Ghz CPUs that I bought with my 1.2 Ghz CPU at TCWO so I could get the CPU warranty extended from 15 days to 1 year knowing I would probably need a better cooler to overclock. The BIOS reports my default CPU core voltage at 1.83v. I'm running all defaults at this point. No tweaking yet other than the BIOS.

Anyway with a room temp of 26C, my BIOS reported "system temp" is 32C and CPU temp is 57C. I'm assuming "system temp" is the temp inside my case somewhere? If so then my CPU temps seem to be about 25C above that which seems quite high to me. I guess its ok until I can get a Glaciator IF the BIOS temps are full load temps and not idle temps. If they are idle temps then I might be in trouble because under load in Windows they could easily go into the 60's or higher if my cooler is getting overwhelmed which I think is getting too high.

Does anyone know if the BIOS temps are idle, load or somewhere in between?

 

[AMDGuy]

BIOS temps are at idle. Since the only work the CPU is doing is displaying BIOS settings. That temp is way to high. I'm suprised they sell a cooler that allows it to go that high. I suggested to someone else to try the WPCREDIT BIOS editor trick where you change register 52 to EB. This is supposed to tell the chip to idle. It's possible yours is set to EC. Setting register 52 to EB tells it to idle when not in use making the temps much lower. If you do a search on the board you should find all the info you need to check this.

Just a suggestion.

 

[e_storm]

I've heard that bios temps are not idle or full load, but somewhere inbetween. There was a long thread about this a long time ago (maybe 6 months???) in this forum. I originally thought what Amdguy said, but popular opinion in this thread said the cpu was under some load, even when in bios.

 

[Len]

I did notice that on my old K62 system, booting to a DOS prompt (real DOS - not a DOS box under Windows) would make the CPU temp rise as if it were under load.

BTW, I forgot to mention in my original post that the CPU cooler came with a thermal pad which I removed. I then applied a thin even layer of Radio Shack heatsink compound to the core.

 

[e_storm]

This may sound like a dumb question, but how did you apply the thermal grease? Did you apply it with your fingers or a razor blade? Just trying to rule that out as a possiblity for high temps. I know applying with the fingers is a no-no... thats what I did before I knew.

Possibly trying cooling programs like Rain might help temporarily til you get a new cooler... if that turns out to be the problem.

 

[Len]

I first put some on with my finger and then spread it around. Then I took a very thin piece of plastic and used it to thin out the compound. Honestly, I'm not sure if I used to much or not. I can see the core through some of the compound but on some parts all I see is heatsink compound. Does that sound like its on too thick?

I really need to get windows installed, run Prime95 and check the temps with MBM5 to see if my temps are dangerous. This is a gaming box so it will be under full load often. I guess that will also help answer the question if the BIOS temps are idle or full load.

 

[Hoot]

There's more to this than meets the eye. The other night, I turned on my PC and had to run upstairs for some emergency with my 7 year old. My PC was sitting at the Windows logon screen, waiting for my password. I know from experience, that this is before the halt-on-idle is enabled. When I came back downstairs, 20 minutes later, I noticed my digital thermometer showed a fairly high temp, for my system. (36.1C). This sparked my interest in a long nagging theory. I finished booting into Windows and shortly after I got to the desktop, the temp plummeted to 22C. Typical for my system idling, with halt-on-idle running. I fired up Prime95 Torture and let it run for 20 minutes. The CPU temp was only 33.4C. CPU stressing programs are supposed to load the CPU very close to max, but my temps were lower than waiting at the Windows logon. I reproduced this result sitting in bios and also booted up in DOS mode. Both of these scenarios also coincide with halt-on-idle not being implemented yet. It appears that even when running a CPU stressing program, there is still a small amount of benefit from halt-on-idle. The exception to this is running BurnK6.exe, which is a burn-in routine that runs from DOS.
What throws me is how high the temp got with such little activity. Think about it. Waiting in bios, DOS, or the Windows logon screen, the PC is sitting there, refreshing video, polling interrupts and essentially waiting for the <Enter> key to be struck. Hard to believe these small housekeeping routines could generate more heat than an accepted CPU stressing program, with the only difference being the halt-on-idle routine. Any other theories?

Hoot

 

[e_storm]

Len (Jun 13, 2001 09:12 a.m.):
I first put some on with my finger and then spread it around. Then I took a very thin piece of plastic and used it to thin out the compound. Honestly, I'm not sure if I used to much or not. I can see the core through some of the compound but on some parts all I see is heatsink compound. Does that sound like its on too thick?

I really need to get windows installed, run Prime95 and check the temps with MBM5 to see if my temps are dangerous. This is a gaming box so it will be under full load often. I guess that will also help answer the question if the BIOS temps are idle or full load.

Thats gonna be an awesome system! By the way... almost the exact same specs I'm looking to put together right now. I have the DDR and case all set up, waiting to get the board and chip.

Anyway, I'm no expert, but that sounds like a perfect amount of compound. Putting it on with the finger might be something though... not sure how much that could really raise temps. I have a PIII 700@875, 1.75v that runs at 38-40C at full load (I put the grease on with my finger), which seems a bit hot for that chip... I'm thinking theres a possibility it might be the "finger thing".

I'd be interested in hearing what that thing Hoot was talking about turns out to be. The "halt-on-idle" or whatever it was.

 

[Len]

Hoot, your observations regarding high temps at seemingly idle times seem to concur with mine. With any luck 57C will be the max I ever see under load in Windows. I can live with that until I can get a better cooler in a couple weeks. Don't you wish you could just go in Wal-Mart and buy a Glaciator?

 

[AMDGuy]

Hoot (Jun 13, 2001 09:20 a.m.):
<<snip>>I fired up Prime95 Torture and let it run for 20 minutes. The CPU temp was only 33.4C. CPU stressing programs are supposed to load the CPU very close to max, but my temps were lower than waiting at the Windows logon. I reproduced this result sitting in bios and also booted up in DOS mode. <<snip>>

Any other theories?

Hoot

I've also noticed as my computer waits for my Network Logon that the temp will get quite high if I don't log on. My theory is that it's cycling waiting for input. Mine system was about the same temp (within 1C) when I let the Logon screen sit there vs. running Prime 95 Torture Test with Priority set to 10. I think "with Priority set to 10" is the key. You have to go under the advanced menu and change the priority. The higher the number the more CPU resources it uses. When I run Prime 95 with default priority I can easily navigate web pages, play MP3s or do other work. With Priority set to 10, the computer is hesitant at doing anything. I can tell the system is much more stressed with priority set to 10. My max CPU temp was 40C after over an hour of Prime 95 Torture Priority 10. I'll test it tonight with Priority at 1 and see if it makes a difference. If it does I'll bet this makes a difference in a lot of people max temps.

 

[Hoot]

I guess it does pay to RTFM. ;D I don't think I have ever played with the priority setting. Definitely on my to-do list, when I get home from work. Thanks for the heads-up.

Hoot

 

[e_storm]

I'm gonna look around when I get home for the priority setting.. but is it in windows, network settings, or in bios?

 

[AMDGuy]

e_storm (Jun 13, 2001 12:54 p.m.):
I'm gonna look around when I get home for the priority setting.. but is it in windows, network settings, or in bios?

The priority setting is in Prime95 under the advanced menu. You have to enter in the password (it's in the readme file) to access that menu.

 

[AMDGuy]

Hoot (Jun 13, 2001 11:10 a.m.):
I guess it does pay to RTFM. ;D I don't think I have ever played with the priority setting. Definitely on my to-do list, when I get home from work. Thanks for the heads-up.

Hoot

I notice in your sig that you're at 1.95 V. Did you do the resistor or trimmer mod to achieve this? If I could get that voltage my Duron 650 will hit 1000 I'm sure. I have the info on the voltage mod, but I'm just not gutsy enough to do it yet. If you know of another way I'd appreciate the info.

 

[Hoot]

I used two 10-Turn 50k ohm trim pots. One for the voltage offset and one to fool the Over Voltage Protection circuit. If you only intend to go up to 2.00V, you can omit the OVP pot. The OVP usually doesn't trip until somewhere between 2.05 and 2.1V. Installing the Phase 1 and 2 mods is tricky solder work. I used a 12W micro pen and went directly to the IC pins. I used 30ga blue wire wrap wire and routed it over to the two trim pots which were mounted in an uncluttered spot about 3 inches from the IC. Start out with the trimpot set for maximum resistance. As you reduce the resistance, you will see the resultant core voltage go up, above the bios called voltage. I set mine for exactly .2V higher than what I am caling for in bios. That way, i can offset the bios called voltage without having to remeasure the actual core voltage. IE 1.85=2.05, 1.75=1.95, 1.65=1.85, etc.

Hoot

 

[AMDGuy]

Hoot (Jun 13, 2001 01:42 p.m.):
I used two 10-Turn 50k ohm trim pots. One for the voltage offset and one to fool the Over Voltage Protection circuit. If you only intend to go up to 2.00V, you can omit the OVP pot. The OVP usually doesn't trip until somewhere between 2.05 and 2.1V. Installing the Phase 1 and 2 mods is tricky solder work. I used a 12W micro pen and went directly to the IC pins. I used 30ga blue wire wrap wire and routed it over to the two trim pots which were mounted in an uncluttered spot about 3 inches from the IC. Start out with the trimpot set for maximum resistance. As you reduce the resistance, you will see the resultant core voltage go up, above the bios called voltage. I set mine for exactly .2V higher than what I am caling for in bios. That way, i can offset the bios called voltage without having to remeasure the actual core voltage. IE 1.85=2.05, 1.75=1.95, 1.65=1.85, etc.

Hoot

Thanks, I got Colin's notes on how to do it. I just gotta get up the nerve.

By the way, is there a way to be notified when someone responds to my reply? I get an e-mail if I started the thread....but otherwise I don't.

Thanks again.

 

[klosters64a]

It's about impossible to apply thermal paste "by the book." Which states "An even layer, one one thousandth(1/1000") of an inch thick."

Just put it on top of the processor's slug in as thin a coat as you can, without leaving bare spots. AFAIK, the "no finger" admonition refers to AS and ASII only.

The thermal paste in aluminum oxide form is thick and cakey, and is impossible to spread evenly. At least for me.

The thermal paste that's a clear liquid(from RS in a "pen,") is the easiest to apply. However, it's difficult to apply an even layer of it.(Even though complete coverage is a snap).

Arctic Silver is easier to handle than aluminum oxide. Unfortunately, making a perfectly uniform coating of it is difficult. I said screw the playing card method, and use a small modeller's flat brush to spread it on with.

Hmm. I wonder if anyone is working to create something that applies thermal paste perfectly?

 

[Thelemac]

Hoot (Jun 13, 2001 09:20 a.m.):
Any other theories?

I think that really, without the halt command being issued to the processor, it's still doing something, even if that something is just "Nope, nothing to do"...it still runs that through, where otherwise the mobo just sees that there is a nop and sends the hlt instead.

The type of work doesn't really stress the cpu more than any other, it might be more apparent when doing complex calculations, since adding 0 to 0 is really a no brainer.

I can't say I know this for sure, or how halt-on-idle really works, but that's the way that makes sense to me.

 

[Hoot]

Set priority to 10 in Prime95 and Torture mode runs maybe 1C hotter. MathMark runs hotter by another 2-3C and BurnK6.exe another 4C higher.

Hoot

 

[Len]

I finally got WinME installed and MBM5. It seems like the 57C I was getting in the BIOS is closer to idle since that's what I get in Windows doing nothing. I ran this "CPU Stability Test" program and the temp rose to 62C. Too hot for me. I may underclock to 1.0 Ghz or something until my Glaciator arrives next week.