Overheating Video Card

[wessunman]

Hi all,

I realized I was posting in the wrong section. I've been having problems with my video card... My old card was an nVidia 7900gt. I could use that video card for anything BUT gaming and it would be fine. Once I started gaming, the screen would start flashing and ultimately my computer would freeze.
I replaced that card with a new video card, the Radeon 4850. Same issues apply. After gaming for about an hour, the game froze up and I was forced to reset the computer. Normally, I have an external room fan blowing into my computer to keep it from overheating. But I figured with the new VC, I wouldnt have to do it being that its winter and its cold lol!
The weird thing is, I dont even have the case on the computer, its open on both sides, so i'm just curious how and why my computer keeps over heating!

 

[drewp29]

So, it doesn't seem to be related to the gfx card . . . but it could be, however the fact that it does the same thing with the 4850 as with the 7900GT would seem to point to something else. Do me a favor and post your system specs, CPU, memory (manufacturer/model as well as speed), aftermarket cooling on CPU (if applicable), power supply, case, and all hard drives in the system. Heat problems can be hard to troubleshoot, but there are things you can do to narrow down the list of possible suspects. I have had issues with graphics cards overheating and causing system instability on several builds I have done for friends, but usually replacing the video card fixed.

Post your specs, and it will give me an idea of where to look . . .

 

[wessunman]

Hey Drew,
I forgot to mention.... with the nVidia... the computer would freeze within 10 min.... with the new video card, it goes over the hour mark atleast before it froze on me.

 

[cjl]

What PSU do you have?

In fact, post your complete specs...

 

[wessunman]

Sure. I have the:
- Asus P5B Deluxe.
- INTEL Core 2 Duo E6400
- OCZ GameXStream OCZ700GXSSLI 700W ATX12V SLI Certified CrossFire Ready Active PFC Power Supply - Retail
- Patriot PDC22G6400LLK 2GB Kit DDR2-800 PC2-6400 Dual Channel Memory Kit
- COOLER MASTER RC-532-SKN1 Centurion 532 Mid Tower Case Retail (case sux)
- Western Digital Caviar SE WD2000JS 200GB 7200 RPM SATA 3.0Gb/s Hard Drive - OEM

I have another HD in there, that is a 150gig that i have only for my OS (Windows XP)

 

[drewp29]

Actually, your case is not that bad provided you do have the front fan and the side fan. Compared to what most large PC builder companies, i.e. Dell, eMachines, HP, Gateway, etc. use it has much better air flow. First I would run some tests that are easy on the gfx card, but fairly cpu intensive to rule out a cpu heat problem.

Get CoreTemp from here: http://www.alcpu.com/CoreTemp/

Get prime95 from here: http://majorgeeks.com/Prime95_d4363.html

Run prime95.exe, choose Options --> Torture Test --> in place large FFTs, click OK

Now watch the cpu temps with CoreTemp. If there is a large gap between the cores, you may have a cpu heat problem (by large gap I mean, > 8-10 deg C). Also, accepted maximum temps for the e6400 are ~70C, but that is a max . . . during the prime95 test you should be between the 65 to 68 range (or lower) after about 10 minutes. If you are seeing temps greater than 70C, stop the test (!), and pull your CPU cooler and clean it, and reapply thermal paste. I recommend Arctic Silver 5 or Arctic Cooling MX2, but any silver based thermal compound will be better than stock. (edit: MX2 is not silver based and is not conductive which is why I like it ) If you don't have any, your local Best-Buy should have something silver based, though not as good as either of the two mentioned, it will do in a pinch.

Try this, and let me know what you get for temps. If the CPu isn't the problem, we can look at MB temps to see if its maybe a NB/SB problem. Basically, I am trying to assume its not the video card and then if nothing pops up with testing, then assume it IS the video card.

 

[wessunman]

Thanks drew, i'll give it a shot today after work!

 

[wessunman]

Actually, your case is not that bad provided you do have the front fan and the side fan. Compared to what most large PC builder companies, i.e. Dell, eMachines, HP, Gateway, etc. use it has much better air flow. First I would run some tests that are easy on the gfx card, but fairly cpu intensive to rule out a cpu heat problem.

Get CoreTemp from here: http://www.alcpu.com/CoreTemp/

Get prime95 from here: http://majorgeeks.com/Prime95_d4363.html

Run prime95.exe, choose Options --> Torture Test --> in place large FFTs, click OK

Now watch the cpu temps with CoreTemp. If there is a large gap between the cores, you may have a cpu heat problem (by large gap I mean, > 8-10 deg C). Also, accepted maximum temps for the e6400 are ~70C, but that is a max . . . during the prime95 test you should be between the 65 to 68 range (or lower) after about 10 minutes. If you are seeing temps greater than 70C, stop the test (!), and pull your CPU cooler and clean it, and reapply thermal paste. I recommend Arctic Silver 5 or Arctic Cooling MX2, but any silver based thermal compound will be better than stock. (edit: MX2 is not silver based and is not conductive which is why I like it ) If you don't have any, your local Best-Buy should have something silver based, though not as good as either of the two mentioned, it will do in a pinch.

Try this, and let me know what you get for temps. If the CPu isn't the problem, we can look at MB temps to see if its maybe a NB/SB problem. Basically, I am trying to assume its not the video card and then if nothing pops up with testing, then assume it IS the video card.

Hey Drew, with core temp the readings never fluctuated more than 3-4 degree celcius. Both were anywhere from 51-56 during the test

 

[wessunman]

Nvm.. during the prime95 test, the temps quickly reached around 72-75 c within 5 minutes! i stopped the test immediately

 

[drewp29]

If you haven't started to remove the heatsink, first boot to your bios --> during the boot process while it is checking memory, hit the delete key repeatedly . . . this should get you to the bios. If you are booting with the front screen on, i.e. during boot there is a front screen that usually has the motherboard manufacturer's name and some graphics instead of showing the memory test and what speed your CPU is running at, just hit delete repeatedly while this screen is up. When you are in your bios, there should be a selection called system monitoring or something like that . . . anyway, find the screen that has the values for your fan speeds and there should be an option for SMART fan speed. Look at the right side of the screen where it shows you the description of the options and whichever option says something about adjusting the CPU fan based on temperature you should disable. This will cause your CPU fan to run at 100% all the time instead of adjusting it according to temp. If this is already disabled, then you should think about redoing the thermal compound on your heatsink to get those temps lower during the prime test.

Don't worry about the temp it reached during the test. The e6400 should be able to reach 80 to 85 degrees Celsius without damage to the CPU. You don't want it to run for any length of time at these temps, but it shouldn't harm the CPU. Prime95 generates the maximum amount of heat your CPU will ever put out, and you most likely will not see these kinds of temps during normal use, even during gaming.

However, if forcing the CPU fan to run at 100% does not keep the temps at or below 70C during prime95 then you should definitely redo the thermal compound and retest it. If you have any questions about how to apply the thermal paste, look that up before you turn your computer off . . . a thin layer on the cpu should be all you need, some suggest just putting a drop in the middle of the cpu and letting the heatsink spread it during attachment, but I still like to spread with a small paddle so I make sure the coverage is complete. Just a suggestion - it takes less thermal paste than you think. And don't get the thermal paste on anything but the top of the CPU (silver based compounds are conductive and could short out the MB if you get it on any circuit paths).

Let me know what you find out.

Drew

 

[drewp29]

Hey cjl, anything else to add? You usually have some good insight . . .

Or anyone else for that matter

 

[drewp29]

If you are using the stock Intel CPU cooler I would highly suggest an aftermarket cooler even if running at stock speeds. I've been in the 'box fan cooling necessary' situation, and it sucks big time.

 

[wessunman]

Hey drew... im out of town for the weekend..
but this is the after market cooler i purchased
Arctic Cooling ACFZP7 Freezer 7 Pro CPU

 

[drewp29]

Just purchased, or is it the one on the CPU now? That's a good cooler, and if it is already on there, then I would definitely say redoing the thermal paste should make a difference in those temps. If you just purchased it and its not on yet, then I would think that given the quality of cooling the AF7 provides you should see a big difference in the test temps.

Bringing the CPU temp down will hopefully make a big difference, if it doesn't solve the problem fully then we'll look at Northbridge and Southbridge temps.

Have a good weekend!

 

[wessunman]

Yeah, it's been on there since day 1. I will play with the thermal paste when I get back next week and keep you posted.
Thanks for the input so far, I appreciate it mucho!

 

[wessunman]

wow... worst luck ever... i got back to my house and booted up my computer.
warning, cpu fan error.
my cpu fan is broken... waiting for a new one to arrive because i can't find the stock fan that came with the chip...
im using the fan i used to use on my vc for my cpu now lol!

 

[drewp29]

Doh! You know, if the fan speed has been low from the start that may have been the problem. I'll keep my fingers crossed that when the new one arrives the temps are kept in check so you can play games without the box fan cooling. We can hope . . . if its one thing I despise in computer problems it is heat issues; it has been my nemesis on more than one occasion.

 

[wessunman]

I'm hoping that was the case all along also!
It would make me

 

[magicbullet]

hey drew and sunman, sorry for interuption. I would like to know where should I stop Prime 95, at what temp??? i have q6600 +stock fan and within 5 min, it goes up to 70, so i stop.

 

[drewp29]

@magicbullet

70C should be alright, if you are upwards of that and into the 75C range, I would consider making sure your CPU fan is running at 100% in the bios settings that I described above. You should check in CoreTemp what your TJmax is set at. "Q" processor's generally accepted TJmax value is 100C. "E" processor's generally accepted TJmax is 95C, although some (like unclewebb on overclock.net) feel it should be 85C but is very 'individual processor' dependent. The core temperature is derived in software by taking the difference of the temps read by the CPU die sensors and the TJmax temp (TJmax - Dietemp = Core absolute temp) This can be accurate if the TJmax is correct for the CPU, but since there is no way to literally determine the TJmax for the given user's CPU, the TJmax values I stated are just "generally accepted" values. There is still a lot of voodoo and hand waving involved in determining whether the temps read in software are accurate, as the only accurate way to determine them is with a hole in the cooling block and CPU and using a probe to read the value directly. But, with the TJmax values set as I stated, it is a good indication of the temperature area you are running in.

So, long explanation short, set your TJmax on the Q6600 to 100C and if the temp gets above ~75C you are too hot, and should consider aftermarket cooling or redoing the thermal paste on the CPU cooler with a proven thermal compound. The stock thermal 'patch' is not very good compared to Arctic Silver 5 or Arctic Cooling MX-2. Redoing the thermal paste is good to do every once in awhile anyway as the thermal coefficient changes as the paste ages. It does take a good day or two burn in period to set the thermal paste and get maximum heat transfer after redoing it.

Once again, I think the 65-70C area in prime should translate to good temperatures for everyday usage.

 

[magicbullet]

ty drew

 

[wessunman]

Drew... just installed the new CPU Fan. idle temps are 32-34
temps with prime95 running are hovering around 47-50
Looks like the entire time it was the Fan!!!!!
I did a pretty crappy job with the thermal paste i think hehe but we'll see how it goes...

Thanks for your help... hopefully nothing goes wrong in the near future!

 

[drewp29]

Excellent, those temps look great - I still have my fingers crossed, but I really think that will solve the problem. Your welcome for the help, these forums have helped me more than once, and I thought its about time I started returning the favor! Oh, and have fun with whatever games you play, the 4850 is a pretty great card and should serve you well. Take care!

 

[wessunman]

just when you thought everything was ok.... I was playing L4D last night and all of a sudden my PC shut off. I didn't think too much of it, so i restarted and started playing again. A few minutes later my pc shut down again. I honestly don't think it is due to my CPU temps, as those never rise over 55... My brand new VC, however, was well over 70.. i think maybe even bordering 80!!! what is going on here!!

 

[drewp29]

That's not an uncommon temp for video cards as they tend to run much hotter than your CPU. Check this thread here:

http://www.tomshardware.com/forum/252306-33-4850-temps

Try doing the workaround that is described in the link in the second post. It links to rage3d.com. Set the fan speed to like 35-38% with 'Automatic' enabled. See if this will help - but first, you said you can manually set the fan speed in CCC . . . do you hear a noticeable increase in the fan noise when you set this to 100%? I can't seem to find whether the newest Catalyst drivers actually fixed the fan speed problem or not. If you do notice a change in the fan noise when setting it through CCC, then the workaround won't be any help.

I think this may be a northbridge temp situation, or perhaps memory overheating. The P5B MBs have a decorative cap on the NB, which is the heatsink with the heat pipe. Remove the cap! I removed the cap on my Gigabyte x38-DS4 and its running cooler now, which has helped system stability with my overclock.

Before removing the cap though, go get Everest from here:

http://www.lavalys.com/products.php?ps=&page=11&dlid=35&lang=en

Just use the free trial, it should give you results for temps, etc., but it won't allow benching and other stuff we aren't going to use.

After install, run Everest and go to Computer --> Sensors

MCP is southbridge, SPP is northbridge

Run Prime95 or maybe ATI Tool's Artifact Tester to see what these temps get up to. If the NB is greater than 60C you probably are getting instability. The Southbridge should be cooler than the Northbridge.

Then remove the cap/sticker from the NB and see what the temps hit.

Let me know what you find out.