View Full Version : helpp... degrading memory overclock, it gets lower n lower overtime

11-11-05, 11:12 AM
I'm just lucky enough to buy a Albatron 6800GT 256MB AGP, just a few weeks ago for a special price ;)

At first, using auto detect optimal clock setting yield 395/1.05
Later on after a few days I replaced stock heatsink with NV Silencer 5, I'm continuing using it, set the to overclock at bootup at level 390/1.05.

One day when I'm browsing hardocp.com, I notice random white dot corrupting the page, other website don't show this behaviour, I lower the mem clock to 1.03 and the problem goes away. Out of curiosity I try the auto detect optimal setting again, it stops very fast and set the clock to 352/1.01 what?? changing between driver revision doesn't change a thing.

Core can still be overlock manually to 430+, ram to 1.02, it runs artifact free at 400/1.0 so Ileave it at that level for a while.
Until today, I found out that memory can't even pass the clock speed test by driver at default 1.0gig, only when I lower the ram clock to 980 than I could overclock the core. If ram clock set at 1.0gig, it didn't pass the clock test, no matter what core clock is at.
I haven't check with my games for artifact though.

Anyone ever having this kind of issue?? what is causing this? failing card?
Can a failing PSU causes this? I'm a bit suspecting the PSU, cause this PSU been having a problem before, I've check by removing all other peripherals leaving only bare essential, but it doesn't help.
I really2 hate to RMA, cause this is most probably the last stock my reseller had for AGP card. The only AGP card left is 6600 non GT, I don't want something that low.

sys spec:
Athlon 64 3000+, socket 754
Abit KV8 Pro, VIA K8T800 Pro
1 gig A-data PC4000 & PC4600 DDR
Audigy 2 Value
1x Seagate 7200.8 200GB
Lite on dvd and benQ dvdrw
Antec TP 430w

11-12-05, 08:40 AM
no body can help me? please?

11-12-05, 09:50 AM
Only thing I can think of is make SURE your HS and RAM have contact. If I recall correctly there were some issues with the HS being kinda random as far as good contact on the RAM. I'd try taking it apart, cleaning it up and doing a test fit to see if there are any RAM chips that are not making contact. There MAY be some sanding/laping in your future to get a good fit. Otherwise, carefully put the HS back on using your choice of AS (I like ceramic) and try again.

It COULD be related to a poor PSU, but I would think you would have more core trouble than just RAM.

Anywho, hope this helps some.

adpr 02
11-12-05, 02:36 PM
I know this would more likely RAISE your core and mem speed because of contact, but is there dust on your card? if there is, dust it off.

11-13-05, 01:37 AM
I try reaseting Silencer 5 a couple of times, this is latest revision rev 2, with thermal pad attached to minimize the impact of diff. ram chip height, all ram chip having good contact with the HS as far as I can see.

I even try to peel off thermal pad and put a lot of white goop to bridgw the gap between hs and ram, but that doesn't help at all, so I'm putting back in the thermal pad.

It's a new card, new hsf, no more than 1 month usage, no dust whatsoever :)

11-14-05, 06:08 PM
I think it has to do with the card arent as good as quality as they where last year.

I burnt my original 6800 up by not cleaning it then i burned 2 rmas up in 2 weeks. The highest oc i used was 400 to the memory on the stock fan dead now. The other i had both a waterblock and a rev3 silencer+as5 on it i just pulled the silencer off last night due to a dead 6800 and i only had a +10 to the memory oc like 480m which killed it.

I think their not useing the same quality ram as they do on the ultras like the first batchs had on them.

Gosh and when i had the silencer on it auto detected 422m and 426 core which i didnt use. :(