ATi Radeon HD4890 freezes in games

[Andu]

Hi, i'm new around here, so sory for my possible mistakes.
I have a Powercolor ATi Radeon HD4890 PCS+ (Zerotherm Cooler) card. The exact model of the card is here.

So, here's the issue: that card has OC from the factory (GPU runs @ 950MHz / memory runs @ 1050 MHz by default). When i play games (such as Metro 2033), the card runs flawless about 10 minutes, then it freezes and the screen goes black for a few seconds; afterward i get the "ATi display driver has stopped working and has recovered" error / or the PC reboots without a BSoD.
IF i underclock the card (900MHz @ GPU / 1000MHz @ memory) the card runs any games and it does not freeze at all .

My specs are:

Motherboard: Gigabyte GA-MA785GM-US2H
Memory: 4GB DDR2
Processor: AMD Athlon II x4 640 @ 3GHz (fără OC)
HDD: WD 320GB S-ATA 1 (5.4 in Windows Experience Index)
GPU: PowerColor HD4890 Battle Forge Edition PCS+
PSU: Enermax NAXN 450W
OS: M$ Windows 7 Ultimate x64

Also, i noticed that if i use the latest CCC (12.1), the error is occuring more often. I use ATi CCC 9.9.

What can i do?

P.S. At first, i thought it was the PSU, but then i tested the card on my other PC (SeaSonic 620W) and the error still occured .

Any sugestions?

[Norton]

Sounds like the vid card is overheating....

Here's a few ideas:

Run GPUZ and check your temps
What case are you using? Try running with side panel off
Check GPU cooler for dust packed into fins (including under fan)

Report back on what you find

[Andu]

Originally Posted by Norton

Sounds like the vid card is overheating....

Here's a few ideas:

Run GPUZ and check your temps
What case are you using? Try running with side panel off
Check GPU cooler for dust packed into fins (including under fan)

Report back on what you find

the cooler is nice and clean. no dust there, also i applied some silver grease on the GPU (the temps went about 8 degrees lower). afterwards:
well i tested the card in two different environments.
i ran furmark 1.8.0 and the card went up to 94 degrees in a slightly small case. (after 10 minutes it crashed - with CCC 9.9)
the funny thing is i also tested the card on my other PC (it has a huge case, very nice cooling) and it crashed when i was running the windows experience index test (after 30 seconds - CCC 12.1)

i don't think it has anything to do with temperature...

[Norton]

Sound like that card may be dying if you tried in 2 different machines with 2 different versions of catalyst and two different PSU's.

If it's running well at lower clocks you may get some more life out of it by leaving it there at a lower or stock 4890 clocks... has the card been used for mostly gaming for its lifetime?

Others who have run these cards may have more insight than me on what else could be causing the issue

[Andu]

Originally Posted by Norton

Sound like that card may be dying if you tried in 2 different machines with 2 different versions of catalyst and two different PSU's.

If it's running well at lower clocks you may get some more life out of it by leaving it there at a lower or stock 4890 clocks... has the card been used for mostly gaming for its lifetime?

Others who have run these cards may have more insight than me on what else could be causing the issue

well, i don't really know where this card came from. i bought it from a site like ebay in my country. that guy had only positive grades, so i gave it a try.
the card looks well...the capacitors seem okay.

can it be a voltage issue?

[Norton]

Originally Posted by Andu

well, i don't really know where this card came from. i bought it from a site like ebay in my country. that guy had only positive grades, so i gave it a try.
the card looks well...the capacitors seem okay.

can it be a voltage issue?

Not sure on voltages for this type of card- check a few reviews against your cards actual #'s

Here's a link to a review of your model card:

http://www.guru3d.com/article/powerc...90-pcs-review/

[Yo_Wattup]

Up the voltage, but make sure it stays cool.

[Outback Bronze]

When i used to run mt 4890s i could hardly get them to run @ 950mhz whatever voltage i threw at them. U might b able to RBE it. I know u really dont want it to run lower than factory speeds but if it prevents it from crashing well......

[AsRock]

As Norton said is most likely the case

Vrm's are probably getting to hot as they do on the4890 without decent cooling.. Any chance you could post a picture of the card. Thing is with the 4890 it's not the GPU temp you have to watch out for but the VRMS and as you have a 3rd party cooler on it they probably suck.

[Andu]

here is a picture of the card.

i'm gonna try to modify its bios to make it run @ 900 / 1000.
anyway i read a bit about that problem on the internet and i found a few solutions that seemed to work for many people:

- get rid of windows 7 sp1
- update chipset driver
- increase the PCI-E latency (from 64 to 128)
- install the drivers only with the CCC & the display driver

[user21]

he meant the real pic of you card that is having the problem.
their might be an issue with the voltage as its an OC version. OC versions of most card do this sort of stuff often like freezing and BSOD better low the clocks it always works

[Andu]

here is the pic: http://img713.imageshack.us/img713/2748/img8527ks.jpg

[user21]

Originally Posted by Andu

here is the pic: http://img713.imageshack.us/img713/2748/img8527ks.jpg

Originally Posted by Andu

i'm gonna try to modify its bios to make it run @ 900 / 1000.
anyway i read a bit about that problem on the internet and i found a few solutions that seemed to work for many people:

- get rid of windows 7 sp1
- update chipset driver
- increase the PCI-E latency (from 64 to 128)
- install the drivers only with the CCC & the display driver

The card looks perfectly fine. However have you tried the above solution?

[magicadey]

i have a 4890 and have started getting this problem that the game freezes, monitor flashes but then on mine it goes back to the game.. been looking around and i think it might be the 12.1 drivers.. get the same with the 12.2 beta ones too..

[Andu]

Originally Posted by user21

The card looks perfectly fine. However have you tried the above solution?

i will try those solutions when i'll go back to college. now, i'm in my hometown...tonight i'll tell you if anything has changed after applying those solutions.

[user21]

Originally Posted by Andu

i will try those solutions when i'll go back to college. now, i'm in my hometown...tonight i'll tell you if anything has changed after applying those solutions.

all the best

[de.das.dude]

Originally Posted by Andu

PSU: Enermax NAXN 450W

You need more juice. As PSUs get older they fail to deliver their rated power.

Also, i dont think a 450W PSU is enough for a 4890. a 500+ PSU is good. You need atleast 16Amps for the GPU itself on the 12V rail.

[Andu]

Originally Posted by de.das.dude

You need more juice. As PSUs get older they fail to deliver their rated power.

Also, i dont think a 450W PSU is enough for a 4890. a 500+ PSU is good.

i tested the card on my other pc with a SeaSonic 620W PSU and it had the same problems.
the PSU is not the issue.

also, the enermax PSU is brand new (4 months old).

PSU link - it has 16Ampers

[AsRock]

Originally Posted by Andu

here is the pic: http://img713.imageshack.us/img713/2748/img8527ks.jpg

See those odd colored heatsinks how hot they getting ?.. you checked how hot the VRMS are with GPU-z ?. And if GPU-Z don't detect the card right try putting your finger on them see how hot they get.

As all i see their is small ass heat sinks with barely any cooling for them like come on the fan on the cooler is not going be doing much to cool them.

Originally Posted by user21

The card looks perfectly fine. However have you tried the above solution?

Maybe.. might be just the card not able to handle it by design. he could try cooling the back of the card as well.

Shame it's not a reference card as there was a few 3rd party coolers made for them.

http://www.techpowerup.com/forums/sh...d.php?t=117140

[Andu]

Originally Posted by AsRock

See those odd colored heatsinks how hot they getting ?.. you checked how hot the VRMS are with GPU-z ?. And if GPU-Z don't detect the card right try putting your finger on them see how hot they get.

As all i see their is small ass heat sinks with barely any cooling for them like come on the fan on the cooler is not going be doing much to cool them.



Maybe.. might be just the card not able to handle it by design. he could try cooling the back of the card as well.

Shame it's not a reference card as there was a few 3rd party coolers made for them.

http://www.techpowerup.com/forums/sh...d.php?t=117140

well, the guy from who i bought this card made tests on it, and it didn't have any problems during them...i still think it's a software issue.
this card should do well @ 120 degrees.

i'll check the vrm temp tonight.

Edit: Stupid question: where's the VRM temp in GPU-Z?
here's a capture on my other card (HD5870):


can anyone tell me where's the VRM temp in that capture?
thank you.

P.S. i guess it's the VDDC, but which one of those?

[de.das.dude]

temp looks ok. temps are good till 100C

[Andu]

Originally Posted by de.das.dude

temp looks ok. temps are good till 100C

the capture is NOT on the HD4890, it's on my other card - HD5870. i just wanted to see where's the VRM temp.

[rangerone766]

i had a few 4870's a while back. thier vrm's got really hot aswell. see if you can set a fan up to blow on the heatsink covering the vrm's, that may help stabilize them.

[user21]

Originally Posted by Andu

the capture is NOT on the HD4890, it's on my other card - HD5870. i just wanted to see where's the VRM temp.

lolzzzzzzzzzzzzzzzzzz then bring the actual card gpuz :P you just made a prank there :P

[Andu]

Originally Posted by user21

lolzzzzzzzzzzzzzzzzzz then bring the actual card gpuz :P you just made a prank there :P

no man...i put those printscreens because i wanted someone here to show me where's the vrm temp in gpu Z. )