Heat Problem: SYSTIN & CPUTIN

[jerreece]

This is kind of an extension of a couple previous threads. Anyhow, I had my system OC'd to 3.3Ghz and things ran fine for two days. Lastnight the system began freezing every 15 or so mins while playing BF2142. So I started playing with settings. Couldn't get anything stable again. Started reading some more and hearing about MCP heat issues with the 680i boards. Here's how I found my current problem.

All the following notes are done under STOCK settings (NO OC).

I installed Everest Ultimate Edition Trial. It will monitor "MCP" temps while stressing CPU. MCP temp (as it shows it) is not a problem at sits around 43c. After running Everest for 2+ minutes "CPU" suddenly went from in the 40's to 101c!

I opened up CPUID Hardware Monitor, and it is showed the following:

SYSTIN 81c
CPUTIN 94c

SYSFAN 5400 RPM
CPUFANIN0 3900 RPM
CPUFANIN1 2600 RPM

Core 0 31c
Core 1 27c
Core 2 26c
Core 3 30c

Not sure what the deal is. I booted up my PC this morning, and am sitting here at idle.

SYSTIN -1c
CPUTIN 30c

I don't understand why last night at idle these same things were 81c & 94c respectively after running Everest, and then closing Everest and sitting at idle for 20 minutes.

If I run Prime 95 right now I get:

SYSTIN 25c
CPUTIN 46c

Shutting down Prime 95 results in those temps immediately going back down. I just began running Everest's stability test again. And all temps appear okay, until 2 mins into test when CPU suddenly hits 103c & the fans in the PC audibly charge up.

Everest must be testing something Prime 95 isn't testing, that ultimately is stressing something on my motherboard. What I don't understand is why CPUID Hardware Monitor shows CPUTIN & Everest shows CPU @ 100+ while both (and Core Temp) show each core in the 30's to 50's when idle or full load.

After Everest is stopped CPU hangs out around 98c... where before using Everest it was 30c. The temperatures WILL NOT reduce below 98c until I actually reboot the PC. At that time I get this again:

SYSTIN -1c
CPUTIN 30c

I'm afraid that when I'm playing games I must be suffering this same problem. Whatever is glitching and causing the CPU & SYS temps to skyrocket must ultimately result in my system freezing. But why doesn't this problem show up when running Prime 95? And why don't the temps go back down after shutting down Everest?

Just to note: At idle all CPU cores show in the 30's, as do both my hard drives. So there isn't an overall heat problem in my case.

 

[Andrius]

CPUTIN(this would be the CPU case temperature I guess) looks about right for a stock Q6600 with a good cooler.

With the values SYSTIN is showing either the Winbond chip has gone to chip heaven or it's a BIOS bug (have the most recent BIOS)?
Check if the northbridge heatsink is hot enough (you think it would be scary hot at 90°C )?

With gaming you also stress your PCI-E subsystem and GPUs. What are your GPU temperatures? Perhaps it's a sign of lacking airflow and all the heat gets cought up around the northbridge and CPU. I still doubt it really gets that hot.

 

[jerreece]

My GPU temps were good a few days ago. Tried to OC my GPU a few days ago (without luck). Anyhow, the guide I followed seemed to suggest my temps were fine on the GPU itself. Think it idled about 44c, and wouldn't let itself go over 71c or something like that.

What bothers me is the fact that SYSTIN and CPUTIN don't gradually go up. They sit at cool temps (CPUTIN @ 46 or so) for 2 minutes. Then within like 2 seconds hit 103c. After I shut down everything, and let the system sit at idle, the temps only go down to 98c for CPUTIN. The cores go down into the 30's. Only way to get anything lower than 98c is to restart the computer, at which point CPUTIN goes right back to 30c.

Just doesn't make sense. If something were getting to hot (for real) it would cool off when the system sat at idle. I've got 2 80mm fans blowing in from the side of the case. One points to the top of my CPU fan. The other points directly onto the side of my GPU (thus cool air hitting front and back of the video card). I have 2 80mm fans pulling cool air in from the front of the case. One blows directly back at CPU fan. The other blows back below the GPU.

Then there's also the 120mm fan blowing hot air out the back of the case from the CPU. Not to mention the power supply has a large fan drawing air out through itself as well. If I hold my hand over the 120mm fan the air that comes out is cool feeling. So I highly doubt my "case temp" is ever as high as 98c, let alone 103c. That's pretty freaking hot.

Anyhow, I put in a service ticket with XFX about the motherboard. Basically copy-pasted my post above to them. Waiting to see what they think about the SYSTIN and CPUTIN issue. Makes no sense to me as to why the temp would spike so rapidly, and never go back down until reboot. Makes me think there's a bug.

Also, I am using the latest P07 BIOS for my board. That was story of its own... had to put in 2 service tickets on that because it wouldn't flash originally. They had to email me a special copy/flash tool to make their own bios flash to this board. LOL Their site even gives you the download links based on your Serial #, yet it claimed I had the wrong version BIOS. That was a pain.

 

[Andrius]

Sounds like this board was flashed before you bought it.

Yes, it looks like a bug. Smells of an "offset overwrite".
If you say the air comming out of the case at the back is cool I wouldn't worry. If you have a thermal sensor probe (like this http://www.highspeedpc.com/Merchant2/merchant.mv?Screen=PROD&Product_Code=Temp1&Category_Code=Overclocking ) just stick it to the side of the cooler and see how far it goes. I can tell you my onboard HSFs never go past 50°C. The GPU is passive cooled and only goes to 60°C when the card is at 75°C (only during 3D Mark loops).

 

[jerreece]

So this "offset overwrite" issue. Is this the kind of bug that XFX would replace my motherboard? Or possibly a BIOS related flaw?

I purchased the motherboard new from TigerDirect. The box was sealed, as was the CD and motherboard inside. Highly doubt it was ever opened. Apparently though, these boards were either factory flashed with an XFX or nVidia splash screen, and each one uses a slightly different version of the same BIOS. (lol how stupid huh?)

Interesting information on the GPU though. I used ATI Tool to heat up my video card while I was watching the temp. What was interesting was it got up to 77c but the fan never went over 45%. Really weird. I had ATI Tool adjust the fan so it would hit 100% and now the GPU only goes to 73c. Not a very affective fan apparently....

 

[Andrius]

If it just a bug in the BIOS code that writes to the wrong register it should be an easy fix. A new version of BIOS.
Every board has a different serial so that's probably as much as the BIOS images differ. Sounds like they went overboard with the "uniqueness".

If the 45% was way less noisy I'd use the 45% or maybe 60%.
The limit is likely the heatsink not being able to remove enough heat from the chips and so fanspeed doesn't matter all that much.

 

[jerreece]

Ok cool. We'll see what XFX has to say as well. Hopefully it's just a stupid BIOS issue and I can reflash to be safe. If that's the case, it probably doesn't have any real affect on my OC ability or instability issues. That's a bummer considering these 680i boards aren't so hot for OC'ing quads.

GPU:

Thanks for your input on the video card as well. I didn't think about the fact that the heatsink on the card might be so poor. Makes sense though since going from 45% to 100% fan speed only changes temp by 4c. Pretty sad on the part of Sapphire who makes the card.

As long as it is safe to run that card at "73c-77c" I guess I'm okay with it. lol

 

[Andrius]

My GF6600GT ran 3.5 years at 70-75°C. It still works and kicks 8500GT ass.

If it's at 77°C on "low speed" and 73­°C on "high speed" I'd say the cooler is pretty good (probably low noise optimised). The fan is rather small (60-70mm) and thin probably around 8-10mm thick. The amount of air it moves is probably too small. Sapphire makes your card? Is it ATI based? Scary selection.

 

[jerreece]

Kind of disappointed in XFX right now. So far from entering my Ticket with them I've had contact with 2 people. The first one simply told me to update to the current BIOS (which is P07).

Ok, so even though this is the SAME guy I talked to back in February (you'd think they'd look up your previous contacts since they use "Ticket #s" and all. Anyhow, I'm already using P07. So, to appease them I flashed the P07 BIOS again.

Still having the same temp glitch with the CPUTIN and SYSTIN reporting extremely high temps in error.

I report details back to them. The next person then asks me what CPU and Heatsink I'm using and tells me they may be installed incorrectly. LOL The original post in this threat, all the details I put up there, are what I gave them in the first place. If my CPU was really running at 103c or 106c as CPUTIN claims, I wouldn't have a CPU anymore. Yet I keep telling them that Everest, CPUID's Hardware Monitor, Core Temp 0.97, and Sandra Soft's program ALL show my core temps around 51c at 100% load.

Not to mention, after a fresh reboot SYSTIN says it is -1c at idle. Seriously, Negative One Celsius? How is that possible if my house is kept at 68F?

I've noticed that when playing with Everest some more (and using HW Monitor to watch CPUTIN and SYSTIN) that Everest shows "CPU" jumping to 106c, as does HW Monitor. However, randomly that 106c will flicker to 51c (which matches what HW Monitor and Core Temp show my actual 4 cores to be temp wise).

So it's kind of apparent that the sensor reading the CPUTIN and SYSTIN temps is either not working right, or there is indeed a coding bug as you pointed out. But XFX seems convinced I must have installed the CPU and Heatsink improperly. They're not listening when I tell them it's not really 106c.

Oh yeah, and the second person told me:

"The only way to cause a cpu to increase in heat is by upping the clock speed and/or vcore voltage."

 

[jerreece]

Sapphire makes your card? Is it ATI based? Scary selection.

LOL Well, the X1950 PRO 512MB PCI-E card I'm running is made by Sapphire, and obviously is ATI based. It was the best I could get for its price range at the time. Seemed like it got better FPS ratings compared to it's equally priced nVidia counterparts. Traditionally though, I don't much care for ATI's Catalyst Control Center and crap, so I never install that.

Always have had nVidia cards prior to this one. Then again, I've always run AMD's before this Q6600.

 

[Andrius]

AMD was the "better" choice from slot A (1999-2000) to core2 (2006) so it's not hard to understand. I've had 2 AMDs in this timeframe myself.

I've been trying to "dump" NVIDIA since my GF6600GT but Every time I'm buying they have the better card in my pricerange. I really wanted to get a HD 3870 last month. But the GF9600GT was 30 EUR less (for about the same performance) and has passive cooling. Runs cool too (45°C idle on XP desktop, 70°C in games, the only exception : 99°C in 3D Mark for some reason (too much airflow in my case for it I guess)).