Radeon 4890 stability

Hello together, I've overclocked my Radeon 4890 to 930 mhz core and 1130 Mhz Memory. I always tested with Furmark stability test ( 1280x1024 no full screen) after increasing the clocks by 5 mhz. The final overclock held 30 min furmark and also didn't show any bigger artifacts on atitools artifact test (only very small yellow dots which it even has at stock speed)

I have played many games since then with no problems.

But in the last month I played Risen and it always crashed after 4 hours or so (Black screen for 5 seconds and then Vista x64 sayd it restarted the drivers)
The same happened in cod modern warfare 2 after 3 hours.

When going back to stock speeds everything was stable again.

I always thought that the Gpu overclock was stable but now it seems it ain't. How can I test for stability without having to play 4/5 hours of some game after each change in clocks?

Would a newer driver maybe solve my problem?

Kind regards Erdinger
13 answers Last reply
More about radeon 4890 stability
  1. Common people someone has to know a stability testing Tool that will help me!
  2. Ok maybe you can't answer this question but I have another one also.

    I have two Pci-e x16 slots (gigabyte ga-ma-790-fxt) and I'm currently using the lower one. Could it be that there's a difference between the two when it comes to overclocking? (similar to ram performing different in the different slots?)
  3. I have a suggestion for your first question:

    The test I run for stability on my graphics card OC is a little unorthodox, but has served me very well. I found that the in game menu screen on Tom Clancy's H.A.W.X. dumps an inordinate amount of load on my GPU, and I have achieved artifacting and crashes quicker on that title's meu screen than even furmark when testing for stability. In the background of the menu is a slowly rotating view of an airplane hangar, and my GPU heats up and carries more load in the menu screen than the playing the game itself does. To note, I run HAWX at 1920x1080, compared to 1280x1024 furmark testing (also windowed, not fullscreen) on my native 1920x1080 monitor.

    I have yet to experience any other title that crashes at the settings that hold up to HAWX menu load.

    Also to note, there is a chance that you have a very low net heat gain in your case. Perhaps your heat evacuation is just a little bit below the heat production of your components, and it takes a long time to for that difference to accumulate enough heat to crash your GPU. Are you monitoring your GPU temps while testing? If not, I would recommend downloading GPU-Z and see if you can pinpoint your temps as a cause for the crashes. Just be sure to enable background refreshing in the sensors tab before you start testing.
  4. I use Hardware monitor and the Catalyst control center Temps never go higher then 79* even when it crashes!

    Ok i will try HAWX out. Thank you that's exactly the kind of help I needed.

    I have a Coolermaster Rc-690 and it is pretty well ventilated. Under the Gpu I have placed the power supply facing up to take away even more heat. The Psu is a corsair Tx 650 so it should be able to handle the extra heat because it's never fully loaded.
  5. I installed H.A.W.X (and auto patched it) and the loading screen didn't stress the gpu over 40%.

    Then I played a mission thinking that maybe then a plain will show up there.

    In the 2nd mission I had to choose an airplane. This happened in a hangar with a turning plane in the middle still not more then 50% load.

    Doesn't it go to 100% because i've patched the game or what?
  6. Try monitoring your GPU's VRM temps with GPU-Z, VRM's get alot hotter than your GPU core and memory.

    They can reach upto 126C before thermal shutdown kicks in.

    BTW my 4890 can only do 860 Core lol any more and it will crash in games, I got a real lemon! Only 15192 3Dmark06 points with a core i7 at 4.1ghz : /

    You think you have problems :L
  7. The problem is that the Pc is a project for the most important paper I've ever written.

    If it doesn't get accepted then the last 6 years of scool would habe been a waste of time.
    I will try monitoring temps with gpu-z thanks for advice. The problem is I still need a good stability test...
  8. download OCCT 3.1.0 and use the GPU stress tool, maybe that will help....
  9. Isn't OCCT only for Nvidia gpus? I will try it out anyways.

    Update: Ok I found out that the gpu test works with amd gpus also. I'm going to test for stability with it and watch at all temps in gpu-z.

    If it doesn't cause instability I will have to play a game for 2 or 3 hours (maybe 4 or 5) for stress testing ^^
  10. WoW now that is an agressive benchmark. It crashes my overclocks after 1 minute or so :) on stock speeds the gpu core reached 82C a temp that it has never reached before (79C was always the limit) and the ram got up to 88C.

    The VDDC phases reached 100C and thats at stock speeds.

    Now I finaly have a benchmark tool which is harder on my gpu than any game I'll ever play. Thank you.

    I will post my final oc here soon
  11. I've found my max stable overlcock.

    Its 920 on the core and 1100 on the mem.

    max Temps reached 83 for the core, 90.5 memory, shadercore 88 and VDDC's ranged from 100-110. What do you think about the temps? Occt has the by far most stressfull gpu testing tool i've ever seen. Furmark never got me that high temps

    A strange thing was that Occt first crashed my Gpu at very small overcklocks already, after installing the new driver 9.12 stability somehow seemed to have increased. Do the drivers really matter when it comes to overclocking?
  12. Drivers might make a tiny amount of differance in stability. Your temps are quite high to be honest, but still safe, If I were you I'd be tempted to rank up the fan speed alittle

    Might you in games the temps will most likely be less then stressing with OCCT.

    I'm selling my Lemon, (4890) it cant even do 870mhz core haha. :P

    how long were you running OCCT stress test?
  13. in the beginning always 10 min, that means 1min idle 5min Stress 4 min idle.
    The final overclock (920/1100) was checked for 15 min that means 10 min stress. I know this ain't very long but when I think that even at 1930/1130 I had 4 hours stability in games now it should really be enough stability. Occt is just way harder on my gpu than anything else and I didn't want to let it run for too long because I also fear that my gpu might break. It would be a VERY serious problem because I need to run many benchmarks and evaluate them for my school paper about the pc. If the Gpu is done then the hole projekt would be threatened.

    I do not want to turn up the fan speed because it's just too loud. Also I've reckognized that the fan makes a very silent "rattering" noise. Could this be reason enough to return it?

    The one thing I am very tempted is to buy the Xigmatek Bifrost VD1065 for cooling, but as far as I know the 4890 has a red metal passiv cooler which shouldn't be removed so any aftermarket cooler want really work perfectly. Also it's just not worth the money for having a little better oc and longer life expectancy.
Ask a new question

Read More

Graphics Cards Radeon Overclocking