Sign in with
Sign up | Sign in
Your question

HD4870x2 - VRAM failing?

Last response: in Graphics & Displays
Share
February 12, 2011 11:48:36 PM

Hi,

My HD4870x2 has developed a strange problem during the last few days and I'm a bit stumped as to what the cause may be.

First time I had an issue was playing "Super Meat Boy", there were some artifacts on the loading screen, and the then my system become unresponsive - on reboot I noticed some of the letting in the BIOS was distorted (configmvum mabner or something like that instead of Configuration manager) and a few other garbled strings.

My first instinct was overheating, although I clean my pc regularly I took the card out and cleaned off any dust I could see and rebooted - all seemed fine until trying to load a game in fullscreen mode, then I'd experience the same crash - then several unsuccesful reboots. I removed all the drivers, cleaned up any trace I could find and then reinstalled the lastest ones from ATI.

Since my card is out of warranty I took it apart and cleaned it more thoroughly and reapplied thermal paste. Back into games and the same problem persists, sometimes I can play the games without issue, but when I quit back into Windows the sytem hangs. One time in particular there were horrible artifacts all over the BIOS on reboot, and Windows, but I haven't seen them since. I've tried putting the fan on a very aggressive high setting just in case it was temperature, but it doesn't seem to change the issue.

PSU is a 750W Corsair
Card is a Sapphire HD4870x2 - it's been running slightly overclocked (on the powerplay setting (GPU 767Mhz, Memory 935Mhz) for 2 years.

I'm happy to accept it's on it's way out, but I'd rather salvage it if possible.

Any thoughts appreciated.

More about : hd4870x2 vram failing

February 13, 2011 12:06:38 AM

There is a test that you can run to check the vram for errors much like memtest however I forgot the name. I do believe that thermal may play a part in this but dual gpu cards are more complex and so are their problems. The VRM on these cards are known to easily break the 90c mark and sail passed 100c under load.
February 13, 2011 9:56:47 AM

Hmm, I did try searching a memtest equivalent but I didnt' turn up anything - I'll go again. This card was a hot beast, I changing the fan profiles when I first got it to keep it under 85 - although it can sound like there's a wind turbine in there.

I had a session on BC2 last night with some small shadow errors, again there was no overheat but a crash as soon as I switched back to the desktop. I might a fresh window installation on another partition, just in case something bizarre is causing it, then it might be time for a new card.

Might a memory overheat be causing this?

Edit: Seems Furmark can run for ages windowed but will die a horrible artifact filled death in fullscreen, even when clocks are turned right down so powerplay does almost nothing.
February 13, 2011 1:18:46 PM

I think that it might be time for a new card as well. I will try looking for the program but I think that you already figured out what the problem already is. Visual problems are often typical of high temps or failure. Overheated vram can easily cause this issue and have had these problems my self on older cards. I suggest trying to put a fan above or near the card for extra airflow. If the temps don't drop you may need to get more aggressive with the problem. Is there any space above the card for lets say some small heatsinks? Stick them onto the back plate but not sure if even that will help. I only got three dual gpu cards and one is older than what most people drive these days. 3dfx voodoo 5 5500 agp, that is the only dual gpu card that is was meant for gaming that didn't have all these terrible issues. Dual nvidia and ati gpu cards always have major problems. The 3870x2 was worse about this than the 4870x2. These monster dual gpu cards usually don't last as long as single gpu cards but some get lucky to have years of use before they fail.

Edit: I searched through 3 weeks worth of threads and it seams to be gone :/ 

Does the whole card seam to be roasting all the time or does it idle down before gaming?
February 13, 2011 3:08:59 PM

The whole card usually sits around 50 while idling - the hotter of the 2 gpus. I though I might have a power supply issue for a while today, maybe the card wasn't getting enough power - but I'm not sure that would explain the crash on switching modes . I think I'm just encountering crashes when I hit the bad Vram - had some more BIOS distortion today then nothing.

I had some success taking the side off the case, turning all the fans up to full and manually setting the fan speed so it sounded like a plane. I'm still leaning towards Vram dying though - I'll do a hearty gaming session tonight with all fans turn up to stupid RPM and if it survives that I may look at some aftermarket cooling as a stop-gap, maybe I'll try replacing thermal pads on the memory ICs.

Thanks for the suggestions/help, good to know I'm roughly thinking along the right lines (thought someone might come along and point out that there was a handy lever on the back of the card to make it work again =)
February 13, 2011 3:16:24 PM

I could be wrong but I think if you run a game or benchmark in windowed mode then it'll only use 1 GPU.... Interesting that you said Furmark runs fine in windowed mode.. Maybe one of the GPU's has kicked the bucket.
February 13, 2011 3:34:44 PM

Well I thought about a mod that you could try, put some ramsinks on the back plate ware there are thermal pads underneath to help remove heat quicker.
February 13, 2011 6:07:53 PM

omgitzfatal: That's what puzzles me, It might be the case that it's only using some of the memory and the bad lot is used by the other processor, I'm not entirely sure how these dual gpu cards work when they're using "crossfire", I think they still have 1GB of memory each, so it may be I'm not hitting the bad stuff when windowed.

nforce4max: interesting suggestion, might be what I try depending on this evenings gaming session results =D
February 14, 2011 3:02:50 AM

You can go beyond that and add more thermal pads underneath then do the ramsink mod on the other side. That will move more heat from the card that would either wise be trapped. You will still need airflow over the top of the card for this to work but it is worth a shot. I have done one mod like this but you can pull a lot of heat through the pcb to cooler the vram and mosfets on the other side. I did that with my gtx 460. With your card you will not need to go the hard work like I did but will need a lot more thermal pads and don't worry just fill all the space under the plate then once that is done apply the ram sinks over the hot spots on the back plate. That will knock your temps down a notch. EVGA did a few cards like that stock using a very thick thermal pad on the other side of the pcb above the gpu for better temps. Get the idea from what I am saying?
February 14, 2011 6:51:41 AM

I think I see what you're getting at. Gaming was a bit of a bust last night and when I booted into windows I had some terrible section of the desktop - think my VRAM is definitely borked - it's usually going well until I hit the bad areas. I'll see if I've got some ramsinks about to try your mod (always fun to fiddle with stuff) and then I'm shopping for another card =D
February 14, 2011 11:58:03 AM

Good luck with the mod and with the new card. :) 
!