Strange Artifact Problem

paroxymparoxym Toronto, Canada
edited November 2003 in Hardware
As the title of the thread says Im having a severe artifact issue here.
I had recently modded my 9500pro and put a stock AMD cooler on it. I had excellent results with the mod and was able to get my core up to 400 without any artifacts even after running a 3 hour loop of the 3dmark Nature test.
Well, all of a sudden today, right in the middle of a UT2K3 marathon, i get polys shooting this way and that. Freaked out I shut down quickly and touched the back of the card... barely warm. I was puzzled so I removed the HSF, reapplied some AS and reseated it. After booting up again and putting the clocks back to stock where they functioned with the normal cooling (325 core/317 mem) I still get artifacts! :thumbsdow :shakehead
I am at a loss here. Is it possible to permenently damage the core so that it will continue to generate artifacts even at stock speeds? Please, any help would be great. Thanks.

Comments

  • ketoketo Occupied. Or is it preoccupied? Icrontian
    edited November 2003
    Yes, it is possible to permanently damage the core. I've seen lot's of screenies, particularly of 9700Pro's that have been volt modded, where there are permanent artifacts.
  • paroxymparoxym Toronto, Canada
    edited November 2003
    I have not volt modded this card though. It was heavily over clocked for a good month but never volt modded. I am getting noticable artifacting at below stock speeds now and am worried this card may be junked. :banghead:
  • SimGuySimGuy Ottawa, Canada
    edited November 2003
    Without proper cooling, the damage to today's VPU's can be considerable within a matter of seconds. Considering ATI's R300 VPU has almost 115 million transistors on a 0.15 micron manufacturing process, the heat output of one of these little suckers is considerable.

    If that AMD Retail heatsink had lost contact with the core (even for a brief few seconds) during an intense 3D session, the core would have generated a substantial amount of heat and would have been unable to dissipate it anywhere.

    You could have damaged the core by overheating it.

    As a last resort, I'd try mounting the original stock cooler on the CPU and running it at normal speeds. That AMD cooler could have touched something on the video card while it was in action, possibly causing an electrical transmission to occur across the core, but it's highly unlikely.

    If the stock cooler & stock speeds won't correct this and attempts to test the card in another system cause the same symptoms, I think you have damaged your card and it can't be fixed. :(
  • paroxymparoxym Toronto, Canada
    edited November 2003
    Well, I put the old stock cooler on and low and behold, she works again at slightly above stock speeds without artifacts. Strange that the AMD cooler worked exceptionally for a month only to start going bonkers outta nowhere. :scratch: I guess I need to take a better look at the sink to see why it wasnt making sufficient contact.
  • killerwibblekillerwibble
    edited November 2003
    Have you tried sanding the surface of the cooler smooth to ensure proper contact? and I assume you applied thermal grease and all that jazz
  • MJOMJO Denmark New
    edited November 2003
    Remember that the shim around the core is slightly higher than the core.
    In other words you need to use a considerable amout of thermal paste to ensure proper contact..
    Or remove the shim. (I wouldn't recommend that, but I have found a guide)

    Guide
  • PirateNinjaPirateNinja Icrontian
    edited November 2003
    disable agp fastwrites in your bios.
  • paroxymparoxym Toronto, Canada
    edited November 2003
    MJO,
    You touched on the conclusion I eventually came to. I figured that the original, ample application of AS3 thinned out as it cured through a few heavy gaming sessions, eventually frustrating contact between the core and the HSF. I'm not too excited about the idea of removing the shim to improve contact but thought that sanding down the shim might acheive the same effect. Ever heard of this being done?
    Thanks for the replies everybody
  • TheLostSwedeTheLostSwede Trondheim, Norway Icrontian
    edited November 2003
    Never heard of sanding it down but i know SEVERAL who have removed the shim though.

    Just use an exacto knife or a razorblade between the shim and the core.
  • Al_CapownAl_Capown
    owns a shimless Radeon 9800 Pro
     Indiana
    edited November 2003
    owns a shimless Radeon 9800 Pro
  • MJOMJO Denmark New
    edited November 2003
    In the guide I linked to there is a link to a thread in Rage3D forums.
    Among other things sanding down the shim is discussed.
    But it is quite dangerous, there is a risk of sanding some of the core if you are not very careful.
    Some components on the core is also in danger of being sanded.
  • W4R-H1LDW4R-H1LD Chesterfield VA
    edited November 2003
    Gravite2090 had this to say
    disable agp fastwrites in your bios.


    What does that do?
  • killerwibblekillerwibble
    edited November 2003
    Fast write I think is the ability to go from CPU right to your VC and not have to go to system RAM. You can disable this in your Advanced Display properties as well I think.
  • W4R-H1LDW4R-H1LD Chesterfield VA
    edited November 2003
    Ok but why would someone want to disable or enable this feature....just curious as to what the advantage either way would be....:thumbsup:
  • EnverexEnverex Worcester, UK Icrontian
    edited November 2003
    It normally causes instability and only provides a tiny performance increase if any increase at all.

    I tried to remove the shim on my 9700 but it is so tightly on that I cannot even get a mini-stanley knife under it, so it may not always be that easy...

    NS
  • W4R-H1LDW4R-H1LD Chesterfield VA
    edited November 2003
    THX man!!:D Now would dissableing the fast write help you in maintaing stability when you are trying to ahcieve higer gpu/memory overclocks using a program like Rage for instance? So this would help you squeeze a little bit more performace out of the card than you normally would be able to with the enable fastwrite on is what im trying to ask.
  • SimGuySimGuy Ottawa, Canada
    edited November 2003
    If anything, disabling fastwrites would cause a performance delay, as instead of the AGP controller having direct access to the CPU & Memory, it must now route through the MCH/NorthBridge and wait it's turn (just like using standard interrupt requests).

    However, the performance difference is like 1%, so there isn't much to worry about if you disable it.

    On a proper chipset, when fastwrites are enabled, there will be no loss in stability.

    Would it help you get a higher overclock? I've never tried disabling fastwrites because they have always worked fine for me (on both VIA & Intel chipsets for NVidia & ATI cards), so I have no reason to disable them. :)
  • EnverexEnverex Worcester, UK Icrontian
    edited November 2003
    I ran 3DMark2001 with and without Fastwrites and my FPS in EVERY test was exactly the same, except for ONE. Strangely enough, the environmental bump mapping test got about 3x the FPS with FastWrites enabled, but it did not effect any other test in any way.

    Strange.

    NS
Sign In or Register to comment.