Sign in with
Sign up | Sign in
Your question

Radeon HD 3870 crashing my PC, and I don't know why

Tags:
  • Radeon
  • Graphics
Last response: in Graphics & Displays
Share
December 30, 2007 1:36:24 PM

I think I have a way to figure out the root of this problem which started after I installed the Radeon HD 3870. I highly doubt it is a faulty card.

The problem is that after playing a game for a certain amount of time, both monitors will go blank after playing a 3D game. First, the game freezes, then I get weird colors on the screen for a second and the monitors blank yet the power button on the monitors remains green and not orange, and usually but not always, I can still blindly access Windows functions.

I also have one thing to note, after putting in the card, I have a feeling the fan on my northbridge cooler died right then and there. The temperature of it when playing a game is 57C. I pointed a Vantec Tornado at it and got it down to 40C. The rest is in the list below.

As another note I had forgotten to say, if I use RivaTuner to put the fan speed at 100% (low level), then when the monitor blanks (this is after it has frozen), then the fan speed returns to the default speed.


Tell me what on this list you think is false and why. If there's anything you need to ask, go ahead.

TBB = Time Before Blackout
  • It is not a heat problem on the graphics card
  • It could possibly be the northbridge overheating, but that is not the main problem as it persists even if I can get that temperature down to 40C; although, the TBB is lengthened in doing so.
  • All temperatures in the PC, including the GPU are either very low or normal
  • All PSU voltage readings are within the "green" area
  • It could be the PCIe 6-pin power and maybe that rail on my PSU having to output too much power, but I have no way to confirm that at the moment.
  • Switching the PCIe 6-pin connector to the other one on the same PSU does not affect anything.

  • The problem only occurs in 3D applications when the graphics card core clock goes from the power-saving 300MHz to the default of 777MHz
  • It has to do with the clocks on the graphics card, both GPU and memory clocks matter as any overclocking shortens the TBB
  • Underclocking the card to below stock, 700 by 1015 (x2 = 2030) has been proven to allow playing without blackout
  • The more graphics intensive the game, the shorter the TBB
  • Pointing a Vantec Tornado to shoot air at the NB lengthens the TBB
  • The problem was not occurring when I had my Radeon X1950Pro, GeForce 7600GT or GeForce 6600GT installed
  • Originally using Catalyst driver version 7.11, changing to 7.12 did not correct the problem.
  • When using RivaTuner to set the fan speed at a fixed 100%, I feel like the TBB is shortened even more.

    Did I forget anything?

    My rig:
    ||| AMD Athlon 64 X2 3800+ @ 2.70 GHz || Thermalright XP-90 Heatsink with Panaflo fan |||
    ||| Antec P-180 Super Midtower || DFI Lanparty nForce 4 Ultra-D || 2x512MB Corsair RAM DDR400 @ DDR450 2.5-3-3-7 |||
    ||| AMD HD 3870 (512 MB) || 22" Gateway VX1120 (4:3), & 19" Dell 1905fp (4:5) || Creative X-Fi |||
    ||| OCZ GameXStream 600W PSU & temporarily, a random 480W to power drives |||
  • More about : radeon 3870 crashing

    December 30, 2007 2:46:52 PM

    I don't want to sound like an ignoramus, but from the very detailed list of symptoms you've given it's most likely:

    GFX overheat - i know u said it isn't, but what prog are you using to get the measured temp.
    NB overheat - however, my NB on this 2ndary comp is at 75C load. A DFI Lanparty ICFX3200-T2R/G.

    Saturn2888 says:

    * The problem only occurs in 3D applications when the graphics card core clock goes from the power-saving 300MHz to the default of 777MHz
    * It has to do with the clocks on the graphics card, both GPU and memory clocks matter as any overclocking shortens the TBB
    * Underclocking the card to below stock, 700 by 1015 (x2 = 2030) has been proven to allow playing without blackout
    * The more graphics intensive the game, the shorter the TBB
    * Pointing a Vantec Tornado to shoot air at the NB lengthens the TBB
    * The problem was not occurring when I had my Radeon X1950Pro, GeForce 7600GT or GeForce 6600GT installed
    * Originally using Catalyst driver version 7.11, changing to 7.12 did not correct the problem.

    All of this indicates a temperature problem with either GFX, NB or a combo of both.

    Try using CPUID's HW Monitor for your temp readings : http://www.cpuid.com/hwmonitor.php

    Sorry if you've tried this already, but to me this sounds like a typical Heat related problems.
    December 30, 2007 3:03:55 PM

    Well w/ the Vantec Tornado on the card, the temp was less than 45C load. Using RivaTuner, if I put the fan speed at 100%, that seems to increase the chances of it failing. I will go edit my first post with the specifics. I forgot to note that when the monitor blanks, the card goes back to the default "Auto" fan speed.

    I know it seems like a GPU heat problem, but the GPU is hotter when playing video files than when gaming. I have used my hand on it to be sure. Right now, playing anime, it's too hot to touch, but that's w/o the tornado and w/o the GPU fan speed at 100%. GPU temp is at 91C right now, and I have never seen it this high before. Activity is at 0%, but the GPU clock speed is 700MHz, not 300MHz probably because I have a video file open or because of RivaTuner's Underclocking of the card which probably overlocked 2D applications as well, but I do not know which for sure.

    I have been using the Catalyst Control Center to record GPU temps.

    If just using the GPU fan 100% rather than using the Tornado (which I was only using as a test anyway), then I had not seen the temps over 50C.

    Clearly, it looks like an overheating problem, but the NB was at 40C and the card was under 50C the last time I played Gears of War. If you want, I could do it again and see what I come up with. But I should probably do it with the card back at stock speeds rather than being underclocked as I haven't had it freeze up underclocked yet.

    And if it is a heat problem, why hasn't it frozen up while it's at 91-93C right now when just watching anime and checking my e-mail and this forum? I'm not making fun, I'm actually posing a serious question.

    [Edit]
    And just not note, the card is sitting right up on the NB's heatsink which is, like I said, 57C when playing games, but that's before the computer crashes. Right now, the NB is at 60C and is possibly heating up the GPU in this way. Do you want a picture?

    All other temps are normal.

    [Edit2]
    After leaving the PC exactly as it was for 2 hours and coming back (video on pause, GPU temp went down to 74C).
    Related resources
    December 30, 2007 3:28:04 PM

    If you've used RivaTuner to set the fan speed to 100% it will most likely force your GFX settings to 700/???? or 777/2150 stock. That'll maybe why when your not using GFX intensive progs it's getting hot, but i doubt it.

    When your checking your temps use oZone3D' Fur tester or multiple instances of 'rthdribl' i find the 3870 drops temps very quickly from load to stock. If you could put the card back to stock it may help a bit as underclocking it may cause it to be unstable. 93C is kinda normal for the 3870 my pair both reach that and don't crash.

    Another thought have you overclocked your comp in any way? It could just be a simple unstable overclock problem if your 100% it's not heat. Have you tried just having one monitor connected?
    December 30, 2007 4:20:24 PM

    I used rivatuner to set higher fan speeds to see if i could stop the blue squares, aritfacts, from showing up.

    I got the same lock up that saturn2888 mentioned when i use rivatuner to run anything other than stock fans. Using Rivatuner, I ran my 3870 fan at 60% and 70%. My GPU temp never exceeded 58c during either speed. BUT, my system locked up hard both times. My lock up had a loud, low tone, low frequency buzz. It really not a buzz, but that is the only word i can think of.

    On the other hand, with stock fans, my 3870 runs 78c at load, but it never locks up.

    I uninstalled rivatuner and now i am having neither artifacts nor lockups. I am guessing, but it is wierd that everytime i use rivatuner to change my 3870's speeds, i have issues. Could the 3870 be incompatible with rivatuner? wierd.

    I set the fans back to stock and now i am good. I am not overclocking anything on my pc.

    my specs:
    Asus mobo P5LP-LE
    Pentium D 920
    4G(2x2G) Patriot PC2-6400 DDR2 800Mhz
    250G SATA150
    ATI Sapphire HD3870 512M DDR4
    windows XP sp2
    Samsung Syncmaster 220wm monitor

    all my other temps have been runnin 30c-50c.
    December 30, 2007 5:32:49 PM

    closed_deal said:
    93C is kinda normal for the 3870 my pair both reach that and don't crash.


    Really? Are you sure?
    December 30, 2007 6:02:13 PM

    At idle my pair of 3870's are 63C and 58C, I don't think they get past 70C during games so yeah 93C is not normal lol.
    December 30, 2007 7:24:40 PM

    i rarely get over 80c.
    December 30, 2007 8:26:08 PM

    93C was odd for me. I never got that high ever, not even when playing games. It had always stayed under 60C as long as I turned up the fan.

    Either way, that's not the point of this thread. I haven't had artifacts ever. Everything just played perfectly. I haven't tried it with one monitor before. Why would my CPU be an issue if it worked perfectly w/ 3 other graphics cards? "The problem was not occurring when I had my Radeon X1950Pro, GeForce 7600GT or GeForce 6600GT installed" I'm not getting angry, I just need to know reasons why. I don't wanna try stuff if there's no reasoning behind it.

    I say that because using 1 monitor and having it work would mean it's a driver issue, that's probably not it if changing my GPU and memory clocks makes it less likely to crash.

    So is what 50bmg is saying is that just by having RivaTuner installed, I'm messing up my computer even if I don't use any of the settings in it? I should then put some registry entries on my desktop to manually change the fan speed of the card itself when I want to play a game.

    Either way, it cannot be a temperature problem if it didn't crash at all when playing Psychonauts w/o changing the card's fan speed or using the Tornado, but when I played Psychonauts at stock speeds, it would crash within the first minute of playing even with the Tornado blowing on it. Gears of War lasted at least 12 minutes in that configuration. When I lowered all the setting to Medium from High and lowered the speeds only a bit, I got 30 minutes out of it. Clearly, that means it cannot be a heat issue with the card itself correct?
    December 30, 2007 11:04:16 PM



    My computer started randomly freezing up for a long while after turning up the GPU fan speed while in 2D mode.
    December 30, 2007 11:28:24 PM

    To be honest you seem pretty damn sure it isn't a temp problem so i'll go away from that line of thougt. I only asked if you've tried an overclock of your CPU because, the general person reading this forum will ( i'll say this loosely) overclock their computer.

    I only ask if you've tried using just one monitor because having both running would create more stress on the card's hardware and software. I'll keep away from the hardware line and go to software.

    I've notice you've tried 7.11 and 7.12. Version 7.11 was badly bugged and should be stricken from all tests. 7.12 is the newest however 7.10 is the only definate version i remember using with little to no problems. So if you'd like to roll back to 7.10 and do the oZone3D stability test for 10 minutes to see if it's fine with no lock ups. If it is fine then go to a dual screen setup and try again. If it fails the single screen test then i would highly suggest returning the card, as that is all you have changed bar the device drivers.

    Your OCZ 600 W PSU is plenty for your system ( pretty sure the 512MB version X1950 Pro drew more power than the 3870) and you shouldn't have to change any settings in your bios regarding PCI-E speed because you know the X1950 Pro and 7600 GT previously owned worked fine. Your 100% it isn't overheating, lack of artifacts may prove that. You can try a single display on 7.10 drivers with oZone 3D running for 10 mins, but if that fials you need to either stick it in a m8's PC and try again or get it replaced under warrenty.

    Again not trying to teach you to suck eggs but you do completely remove your previous display drivers before installing the ones your going to use?
    December 30, 2007 11:44:56 PM

    That HWMonitor pic shows the max temp of your GFX at 97C if it is indeed reaching that then the software will be automatically shutting down the gfx card before it is damaged if artifacts are appearing or not. Sorry if my train of thought is all over the place just ask if your not sure.

    If i were in your position atm i would be thinking of checking airflow inside your case, but your CPU temps look fine albeit a tad high if it's not overclocked. Or if rolling back the drivers on a single display with the oZone3D bench running fails to work i'd think of getting it replaced under warrenty looking like a dodgey card. Someones bound to get one :sweat: .
    December 30, 2007 11:59:13 PM

    Like I said, the 90 some C was a fluke. If it's not, then my card loves that temperature as, when it's lower, it has problems. I was reading the temps while in game, I can get a video VERY easily if you wanna see it die at 50C. It's 48C now btw.

    That 90+C was only because I had the card underclocked (to 700MHz), and it was running not at 300MHz for some reason while in Windows. Either way, I was playing Psychonauts last night when it was on the same settings. I had changed nothing from then to tonight, not even restarted, yet I was able to play that game straight for hours without any problems.

    Only when the card is around 50-60C do I start w/ the problems.

    The HD3870 runs on the 7.10 drivers? I did not know that, but if it was a driver issue, wouldn't that mean that the clock on the card wouldn't affect anything? I'm only asking.

    I'm putting it in another PC to make sure the card is fine. If that works, the next step is to try another 600W PSU. With the amount of drives I have, 600W is NOT enough. Trust me on this one. That's why I'm running a dual PSU setup atm.
    December 31, 2007 1:26:53 AM

    I was right, it was my PSU. That's the only thing I could think. The card is fine, but my PSU pretty much got whipped because I had too many SATA and USB devices (10 of both). That's why I was running dual PSU, but I didn't think to switch which PSU got control of the mobo (and thusly, the HD 3870).

    How did I fix the problem? I switched to another PSU temporarily until I upgrade.
    December 31, 2007 9:26:25 AM

    Ah i see. Well i'm glad you've sorted it anyway. Seems i've learnt something new aswell. I guess you were trying to use a 480W random branded PSU to power everything which was plugged into mobo and a 600W plugged in at the mains but not actually switching on because it wasn't connected to the mobo?

    This is just to cure my own curiosity.
    December 31, 2007 9:28:00 AM

    That actually didn't solve it, but it extended the TBB a lot. Then again, I just keep the card underclocked, and it works dandy on stock cooling and stock speeds (not 100%). I'm still wondering what the problem is. A faulty card? Doubtful. There have been no reports of one yet or anyone with problems like me.
    December 31, 2007 9:29:09 AM

    I see, so you just have the OCZ 600W powering everything now?

    Could you take all excess Optical drives and all bar your boot disc drive/s out or simply disconnect them. This would help you find out if your PSU is beefy enough.

    Although a 600W should be good for 4 HHD's, 4 Disc drives, 140W CPU, two 3870's an external HDD USB, charging of MP3 players via USB and finally USB camera/camcorder, because my main comp was exactly this setup only on a Hiper 580W Type R albeit i couldn't overclock my CPU or gfx thats why i now have a Hiper 880W.

    Here is a link to the power the 3870 draws : http://www.hexus.net/content/item.php?item=10415&page=1...
    December 31, 2007 9:41:18 AM

    Nope. I switched it with an Ultra X-Finity 600W I had lying around.
    December 31, 2007 10:23:36 AM

    You say you have 10 of both SATA and USB i'm guessing you've plugged in a PCI SATA adapter? And have USB expansion plugged in? Could you be overloading them. I know it wouldn't explain why the new card makes your comp's screen turn off.

    I must say at this point that if you have tried all of the things i've suggested I'm pretty close to being out of idea's.

    If you have;

    -->run the card @ stock settings checking temps with oZone3D stress test running, on single display with 7.10 drivers which i know work with the 3870. If it's not crashing from the heat, but still crashing cont down...

    -->removed all non-essential disc/optical drives, USB devices so the PSU has minimal power draw. Just have your bootable hard drive/s if you have a RAID setup, mobo, case fans and GFX running. Then repeat the above. If it still fails i would then suggest PSU upgrade depending if the PSU is older than 2/3 years.

    Then if it's still crashing/blanking out/freezing, now be aware that we have elminated the PSU and a OverHeat problem i would definitely say it a faulty card.

    All you have done is, literally removed the older drivers, fit the new card, plugged in a 6pin connector and reinstalled the desired drivers. It should be that simple, so there is nothing else you can fault find. It goes GFX -> Drivers -> PSU -> mobo, but we can eliminate the mobo because you had previous high end PCI-E gfx cards fitted.

    So in my mind we've completed the fault finding tree. Maybe someone with even further in depth knowledge could help coz this is as far as my experience goes.
    December 31, 2007 10:27:54 AM

    Oh, nothing's really in the PC right now except for a few HDDs. 3 Optical drives died after I put the card in. One of them had the motor get stuck, and it smelled like burnt plastic.

    I also lost an HDD, so on the side, if you know anyone good at de-striping a RAID0 array, gimme their contact info plz ;) . The platters are fine, just the chip on the drive started smoking is all.

    I'm gonna do some more experiments tomorrow. Thanks for your help though! I'll still work on it.
    January 3, 2008 3:58:38 AM

    Yea I use to crash after like 30 min in game when I was using ATI tool. Now I uninstalled the atitool program and re-installed CCC. Afterwards there were no more crashes and there is a reason they default the clocks.
    ITS STABLE!!!
    January 3, 2008 4:34:51 AM

    The default clocks CRASH! I'm not overclocking. It's unstable at stock clocks so I am underclocking it and that is making it work. WHY? My only guess now is that it's the card. I'm gonna have to figure out how to find out. Basically, I need a PC that it'll fit in other than mine.
    January 3, 2008 12:56:59 PM

    I dont know what brand card you have but i bought the Visiontek. Mostly because they have a life time warranty. There are always badcards but its like 1 out 500. It sucks that you got one of those cards but it happens. Like when I bought Postal 2 when it was new I was one of the very rare/few to get a install disc that had corrupt files on the disc. Of course I got a new copy charge free, but these kind of things happen.
    January 3, 2008 1:40:32 PM

    It must've all been because of the card. There's not other reason. I also got the VisionTek and basically, if I can't even run stock, I really need to RMA this sucker.
    January 3, 2008 1:45:37 PM

    The problem started when you installed a new video card. The obvious answer to me is that the new card is partially defective. Can you test it in some other way?

    The problem happens when the vga card is put under stress. If it were a power problem, I would expect an issue shortly after it ramped up to full load. Since the failure point seems to gradually build up, I suspect that the failure is heat related. Longer build up time if underclocking would support this theory.

    I think the card is defective, and can not stand the normal high temperature that it should. I think it is warranty RMA time.
    January 3, 2008 1:54:31 PM

    Yep, that's what I'm up to now. :) 
    July 21, 2008 3:22:13 PM

    Saturn2888 you're not the only one with the problem, I'm currently running Radeon HD 3870 X2 and it also blackouts after ~5-10 minutes while playing 3D game.

    The problem is I don't dig as much in computers as you all do, so could you pls explain in a more simple way how should i solve the problem ? :)  ty
    July 21, 2008 10:52:01 PM

    Well actually, I was just sent a new one by the company I bought it from, Dell. NOT a Dell branded card, just got it from their store. I hadn't had problems since. Actually, it's in another rig right now running on passive cooling with the stock heatsink :p .

    Now, that fixed that issue. The HDD problem? I got Seagate to send me a new HDD because mine have a stock 5 yr warranty. I took the board off of the still-working drive and put it on the messed up one in another rig and used Knoppix Linux via LiveCD to do a bit-per-bit copy of the data from the old drive w/ the messed up PCB onto the new one Seagate sent me. Then I went through hell trying to get everything working after that, but news is, I was able to rebuild the RAID0 array and find the filesystem in the new array and now all of my data is back. Awesome!

    A more detailed and exciting writing of this story here: http://blog.umkcddr.com/?p=386 - sorry the blog looks bleh. I just moved over from Xanga to WordPress. Much work I have to do.
    February 16, 2010 7:03:33 AM

    Very late to this party but I just thought I would update the thread because when my 3870X2 which had been happily working away for 18 months suddenly started to act up in the same way as the OP, I found this thread as #2 in the search results.

    So I'll add this for posterity, you never know maybe someone else will find it helpful (certainly the info in this thread helped me a lot with my issue.)

    Symptoms were the same, you'd be playing a game and bang, system lockup, no video. Audio would continue to play and it seemed as though windows was still working but you had to power off. Occasionally after such a hard reboot I would see a bluescreen on reboot relating to ATIKMDAG.SYS and oddly enough after this would happen the system would then reboot and be fine for a few days. The few times I managed to get a Rivatuner log from it when this happened the GPU was getting very hot (90/100 C) just before the crash.

    Using Rivatuner to run the fan manually seemed to help sometimes but not always.

    Aaaanyway, long story short, it got very bad last week so I took the card out, removed the stock cooling parts and re-pasted the GPUs with some leftover CPU Thermal paste. Then popped it back in, and set the fan to idle at 30% of max. speed. And so far, that seems to have done it. Ran last night for over 5 hours in ME2 and Mod WF 1 and 2. Temps are getting high (90 C after an hour in ME2) but it would seem that once you have set the fan to spin at anything above 25% it will correctly kick in and ramp up as the GPUs get hotter. Saw one artifact for a few seconds at one point and then the fan kicked in.

    I have to manually set the fan speed before playing as RT doesn't support atuo-fan start in Vista for this card but as the card is well out of warranty and I don't have the funds to replace it it seems to have been a workable solution. For now...will replace the card eventually though because I'm sure what I have done is temporary at best but at least if you find yourself in the same situation with an out of warranty card, don't be afraid to try this!
    March 22, 2010 2:54:50 PM

    Late myself as well, but I just want to tell you all I appreciate the info as I was having the same problem, and after some testing found my MSI 3870 was running at 40% all the time, getting up to 110 C before it crashed out after less than 5 mins of play in Ghost Recon or Operation Flashpoint.
    Cleaned my card (gunk built up in front of the cooling fins) removed the HUGE amount of thermal paste it came with and replaced it with a more appropriate amount of Arctic Silver 5, and downloaded Rivatuner, set the fan speed to 65% all the time (the level where I can't hear a difference, at 70% I start hearing the fan) and my card runs between 50c and 60c no matter what I do with it now!
    March 22, 2010 3:12:46 PM

    Darkrose,

    Glad to hear you got it sorted!

    Mine is still going strong after a month and I don't even need the fan on any more via RTuner.(above the default 25% that the card already uses, that is)

    Amazing how such a small thing (bit of fresh thermal paste) can solve an problem like that, isn't it, better than shelling out for a new card that will go the same way in a year...
    March 22, 2010 3:39:44 PM

    The biggest reason I'd never found the problem was all my GPU intensive games were on my last computer, and never transferred them to my new machine almost 2 years ago. I had no idea the 3870's had that BIOS issue with a constant fan speed until I started digging around...
    All good, no money spent, I'm a happy camper!
    It was still getting warmer than I liked with the stock fan speed,e ven after cleaning, hence the 65% fan speed, but it sounds no different at 65% vs. 40% and the temps show the difference definitively.
    !