Motherboard CPU Temp warning

[ekdikeo]

Hi guys My new system is all up and running (though I wish the processor ran a little cooler, I need to make some serious case adjustments, when I run it open, I drop from 31 deg idle to 21 deg) yay!

However, I started with the BIOS CPU Temp warning set to trigger at 60 degrees. I start up Prime95, as soon as it hits 50 degrees according to CoreTemp, the temp warning starts blasting. I bump it up in the BIOS to 65, the temp warning starts blasting when CoreTemp reaches 50.

OK, so, either CoreTemp is reporting bad numbers, or my BIOS is reacting to bad numbers.

Can I trust that the BIOS option to automatically decrease processor speed and such when it hits thermal emergency kick in properly? I don't know what to trust, what not to trust, etc.

 

[trents]

CoreTemp is actually reporting CPU core temps, not CPU socket temps. Your motherboard sensors are reporting CPU socket temps. CPU socket temps are typically higher than CPU core temps. What you described tells me the alarm is probably functioning normally. If you get a 10 degree drop in temps from opening the case you obviously have a case air flow problem? What make and model case are you using. How many, how big and what is the placement of the case fans?

CPU core temps are the ones that are critical. You want to keep them from exceeding mid 50s C.

Download and install HWMonitor. It will display both CPU socket temps and core temps.

 

[ekdikeo]

mmm.. nice, I think that is just the software I am looking for.

Yeah, I'm aware that I have serious case air problems, the case is an old one that my gf had laying around. One of it's three fans was dead, and the other two appear to be intake rather than exhaust, I need to turn them around. Strangely, there's no place in the front to mount an intake, so I'm guessing that i'm going to have to go with the side vent on it to mount thei ntake, and then have the upper fans all exhaust.

OK, I'm getting the alarm when my "TMPIN1" sensor goes above the temperature threshold, so I'm guessing that that is the CPU Socket temp. I guess I can disable that warning, if it's mostly bogus?

I also have a TMPIN0 that doesn't seem to be moving (43/40/43), and a TMPIN2, that is ranging from 75-105, and I don't know what that is connected to.

Prime95 starts blowing up when I get above 223MHz on the HT, a bit disappointing, but I'm used to my E4400 that gets a 60% oc on everything stock. This one just barely gets past 10% before hitting a wall, apparently. There's a ton more variables on this one than there are on the Intel though.

 

[trents]

Yeah, TMP1N1 is most likely the socket temp sensor reading.

It is not uncommon for those hardware monitoring programs to misidentify one or more sensors and give false readings. Don't worry about that.

I'm sure your CPU is capable of more overclock than you are now getting (with better case ventilation) but your are correct, there are a number of variables that need to be manipulated to get there such as CPU voltage, CPU-NB speed and voltage, HT Link speed, and Memory frequency and voltage. The trouble with using the HT Reference (FSB) to overclock is that it causes lots of other things to speed up at the same time whose frequencies need to be adjusted downward and whose voltages may need an upward push to compensate.

 

[ekdikeo]

I don't seem to have a lot of other frequencies that I can change, everything else seems to be multipliers. I just exchanged the original 4 gigs ram for 8 gigs, so i'm going to see if I can get any more megahertz out of it. At fully stock frequencies, i'm still hitting around 55deg at full load on Prime95 (after several minutes). So, I'm not too hip on the idea of increasing the voltage.

I'm really curious what TMPIN2 is, it starts at 75degrees, when I crank the CPU out in P95, it goes to 102. It could be some sensor for case air temp, perhaps, but it doesn't seem to be agreeing too well with my actual case temp sensor, which sits at 81(F) and stays there, with it open.

I've spent the last week dealing with hardware issues, so I'm going to spend this week getting real work done, and worry about the fans when I have some downtime.

Wish I could turn off the BIOS cpu temp warning from within Windows, I'm sick of rebooting. haha

 

[trents]

The multipliers are the frequency adjustments. For instance, the HT Link frequency and the CPU-NB frequency are multiples of the baseline HT Reference of 200 mhz. For example, if the HT Reference (FSB) frequency is increased from 200 to 227 and the multiplier is 10x NB the HT Link will show up in CPU-z as 217 x 10 or 2270 mhz. Since you want your HT Link speed to remain at or lower than the stock 2000 mhz for stability sake you would lower it to 8x which would be 1736 mhz. Same principle with the CPU-NB. But since the CPU-NB actually likes to rise with the FSB for stability purposes (to a point, that is) it doesn't require quite the diligence the HT Link speed does.

Moving from 4 gigs of ram to 8 probably has a negative effect on stability since it puts extra strain on the Integrated Memory Controller (ICM). Besides, there aren't many applications that will benefit from that much ram. There's this assumption out there that "more ram must be better" which is not true. It's only better if your applications really need it and it has a negative impact on oveclocking. Moving from a ram frequency of 1333 to 1600 can also have a negative effect on overclocking for the same reason. Ram is cheap these days and we're seeing a lot of people on the forum moving from 4 gigs to 8 gigs without a good reason and without realizing it can be a damper to their overclocking effort.

Can you increase the bios temp alarm threshold from 60 to 70 to eliminate that nag?

 

[ekdikeo]

I would tend to disagree, trents, as far as amount of RAM goes -- I had 6 gigs for a week on my main machine before I tried to go with 4 gigs on the new machine. It was an untolerable difference. Now that I'm at 8 gigs, everything is running smooth as silk again I am a developer, so I can definitely put it all to good use, but even in non-development things, or things that shouldn't suck up that much memory, I still notice a huge difference. 2->6 was mindblowing, 4->8 was a very nice difference.

If nothing else, for the startup time on Visual Studio -- 2 minutes with 2 gig ram, about 30 seconds with 4 gig, and it starts the moment I release my finger from the mouse button with 6+gig. And I don't do anything on the art side. I have artist developer friends who have 12 and 16gig and feel constrained with any less.

I don't dispute, though, that adding more RAM can constrain your overclock, there's 2x the number of bits there that can get muddled.

As far as my overclocking, nothing changed there -- 223 is the max so far.

I did knock the temp alarm to 70, as it looks like the highest that sensor ever goes right now is about 66, with the cores reporting 54-55.

SO, I should knock the HT Link multiplier down, and then try bumping again? Sounds like a plan, I'm on it.

** edit: 8x @ 225 fails p95 instantly, 9x @ 224 doesn't, running now.

 

[trents]

Of course, AV/image editing software is the most memory hungry class of apps. Personally, in moving from 2 gigs to 4 gigs of ram I notice very little difference in performance in boot times, internet browsing, offices apps and things like that with Vista or Windows 7.

Edit: Or, if you have a lot of apps open at the same time lots of ram would help. But as long as you aren't having to resort to virtual memory hard drive usage, having more and more ram shouldn't make a difference.

 

[RJARRRPCGP]

** edit: 8x @ 225 fails p95 instantly, 9x @ 224 doesn't, running now.

Looks like it's NOT an unstable core.
Now, I would check the DRAM clock and see it that's too high.
The DRAM clock may be too high.

 

[trents]

Looks like it's NOT an unstable core.
Now, I would check the DRAM clock and see it that's too high.
The DRAM clock may be too high.

Good thought.

It wold be helpful if you would post pics of CPU-z tabs: "CPU", "Memory", and "SPD".

 

[ekdikeo]

I got thru 15 tests at 9x@224 (before stopping it because i had to use the pc), temp got up to 57, I don't think I really want to push it any higher? not without getting some better cooling going?

Is there like some master guide to a good way to go about doing this, on a non-black edition?

I've got AMD's guide open but I haven't really had time to try to absorb it all. I guess I need to do it, but it seems like there's just a whole raft of stuff that doesn't apply.

(voltage was low at the time, i have all the power saving stuff enabled when i'm not playing with the numbers)

 

[trents]

Everything looks in order but your temps are an issue, maybe the issue. You seem to have uploaded those pics so that they are obscuring some of your post text. There is a tool built into forum for doing this so that wouldn't happen. Click on "Go Advanced" then scroll down to "Manage Attachments" then "Choose File" then "Upload" then "Submit".

 

[ekdikeo]

hehe.. i took the image with three CPUZ instances running over the top of my browser, that's the post from the top of the forum. didn't even notice what was behind them all when i snipped them! hah didn't realize there was an attach option, most forums i go to don't let you attach files, so i just threw it on my web server.

I think I'll need to get back to this in a few days, though, I really need to catch up on all my lost time in the last week thanks so much for everyone's helps here.

 

[ekdikeo]

Hey guys. Sorry to bring up a year old zombie, but I finally got back to messing with my OC settings . . . yeah, i know i said a few days .. and then time slipped away.

I don't know if I missed this in the original OC instructions that I followed, but I learned something very important about overclocking the Phenom II:

HT speed should remain as close to stock (2000) as possible.

Once I figured that out, I'm now running at 3750MHz, temperatures rarely exceeding 50C. I was able to boot at close to 4200MHz, although the temperatures over 4000 were insane, and I couldn't seem to get it into Windows over 3750. After I move in about a month, I plan to tear it down and seriously examine my cooling situation, and see if I can get it over 4ghz.

SO, lesson: Keep your HT clock at 2000MHz or as close as possible, set your FSB as high as you can get it with a lower multiplier, then adjust the multiplier until it stops working. Back the multiplier off by one, and increase NorthBridge and/or CPU voltage until stable.

My temps are now just barely above the stock 3.2ghz settings, with a good improvement in speed.