• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Overclocking Artifacts

B

burningrave101

[H]F Junkie
Joined
Sep 9, 2003
Messages
11,825
I'm pretty new to video card overclocking and i'm curious how you guys test? I've just been running 1 loop of 3DMark03's Mother Nature game test. I have a Gigabyte 5900XT and i clocked it at 450/780 for the moment and i notice a few little black blips on the screen at certain parts of the Mother Nature test. They usually happen towards the end of it and sometimes there is only two or three. Should i just keep clocking down until there isn't any at all or do you guys just "overlook" slight artifacting when your testing?
 
bobmanfoo said:
i usually OC 10mhz under what it artifacts at
Click to expand...

But what tests do you run to see if its artifacting? I dont usually get artifacts in the other tests. The Mother Nature one really draws them out though lol.
 
Is it possible to damage a video cards core by overclocking it too high? When i was first starting out i was raising it by 10Mhz at a time and using RivaTuner and i tested each time with 3Dmark01 game tests and i wasn't getting any artifacts so i kept going higher and higher and i was like nearly as far as the slider would go which was like 585Mhz from 390Mhz lol and i'm curious if that could of damaged the core so that now it doesn't overclock as well? I didn't have it running at those high speeds very long, just while i was testing to see how high it would go without really artifacting. It never froze up or anything in taking the core that high ither.
 
burningrave101 said:
Is it possible to damage a video cards core by overclocking it too high? When i was first starting out i was raising it by 10Mhz at a time and using RivaTuner and i tested each time with 3Dmark01 game tests and i wasn't getting any artifacts so i kept going higher and higher and i was like nearly as far as the slider would go which was like 585Mhz from 390Mhz lol and i'm curious if that could of damaged the core so that now it doesn't overclock as well? I didn't have it running at those high speeds very long, just while i was testing to see how high it would go without really artifacting. It never froze up or anything in taking the core that high ither.
Click to expand...

Good lord.

Yes, you have probably damaged your card. What the fuck were you thinking?
 
Don't post foolish stuff. Does the card act up now? If not, you didn't damage the core. Besides, there is no way for anyone to know if you damaged the core until it decides to crap out. I have seen people with 5900xt's that go up to 609 mhz core on stock cooling. You should be fine as long as you don't get lockups in your games.
 
:rolleyes: If the card has lost its ability to reach core speeds it could once hit, it is damaged.
 
Lowtax said:
:rolleyes: If the card has lost its ability to reach core speeds it could once hit, it is damaged.
Click to expand...
you don't know what you are talking about.....shhhhhhh

burningrave101,
i use atitool to check for artifacts, imho the 3dmark benches aren't good for OCing, and im pretty sure your card has the GPU-slowdown feature (forgot what it was called) as my old BFG fx5900 does, which started artifacting above 480mhz
 
By the way, you know how ATiTool renders a hairy cube and then scans it for artifacts to see if there are any? Would that feature still work with nvidia cards?
 
Its a Gigabyte 5900XT and its running fine now without locking up or artifacting at 440/800. It starts to lock up and artifact if i push it up to 450/800. Its just that when i first got it it was my first time overclocking a video card so i took my time and started with the core speed and kept uping it and uping it 10Mhz at a time and i wasn't really seeing much of any artifacting while running through several loops of 3dMark01 so kept going higher and higher and if i remember correctly i had it over 550Mhz core speed and it ran all the way through 3DMark01 without locking up. Well since then, which has only been about a month, i decided to overclock it at a reasonable speed for everyday usage and it doesn't seem to be able to go up to 450Mhz+ now without starting to lock up in 3dMark01. It doesn't lock up in 3dmark03 but it does artifact some in the mother nature test.

Hey i read a dozen guides before i ever tried to do this and i did my best and it was my first time and i got a little carried away lol. I just figured it would lock up or artifact bad like it did when i took the memory over around 810Mhz but it didn't and i just kept raising it higher until finally it did start to crap out around 550-585Mhz. I never left it at those speeds though or restarted at those speeds. I just had it clocked that high while benchmarking to see how high i could get it stable.

I'll try and be more careful with my new nVidia 6800 GT or Ultra when i get it here in a month or two.

I've been gaming on it alot lately and overclocked at 440/800 its going through 3dmark03 game tests looping continuously just fine. I guess i should just be happy and leave it at that.

BTW, which is more important in actual gaming performance, clock speed or memory speed? I sorta thought clock speed affected benchmarks the most but memory speed affecting actual gaming performance more. Which is it?
 
The NV3x's were for the most part heavily memory bandwith dependent. You should get more performance out of oc'ing the ram than if you did the core.
 
bobmanfoo said:
you don't know what you are talking about.....shhhhhhh
Click to expand...


I have been overclocking video cards since my TNT 2.

I certainly know what I'm talking about.

If a GPU has lost its ability to hit a certain core speed that it could once hit, it has permanent damage. This is irrefutable fact.

Maybe you should do some learning. I've been posting here under different names since waaaayyy before you regged.
 
Oh, really? I'm going to need pics for TNT2 and all the cards till now for proof plz.
 
Lowtax said:
If a GPU has lost its ability to hit a certain core speed that it could once hit, it has permanent damage. This is irrefutable fact.
Click to expand...
That is correct, assuming none of the other variables have changed.

However, I don't remember him saying anything about it *not* running 590MHz again. If he had slight artifacts at that speed only in nature test, sounds like the pixel shader portion of the chip couldn't handle the overclock. Drop the clock and keep trucking.

M4d-K10wN
ATITool artifact detector works fine on Nvidia cards, been tested. Been there, done that. The overclocking does not work with an Nvidia card, but the artifact detector works nicely. It does not work on cards that do not support pixel shaders.

As for finding the absolute limit, run all your 3d apps. My old G2 MX would run EVERYTHING 100% perfect @ 220/198, 3dmark 2001 for hours, JKII, AvP, AvP2, N4SHS, etc... everything I tried. Finally pulled out Serious Sam again, installed it, I get white spots. Dropped the RAM to 195 and the spots are gone.
 
Just curious but how much do you usually gain in actual gaming performance from overclocking say 50Mhz on the core and 100Mhz on the RAM? I'm going to download fraps and check out my frames in Morrowind now but is overclocking the video card really worth it? Overclocking the CPU is definitely worth it but i've seen quite a few people say that overclocking the video card only gives minimal gains in performance.

What do you guys think? I'm trying to figure out all this stuff now before i try to push the new more expensive card i'm getting in the next month or so.
 
I'll probably just grab me a new 5900XT when i order the FX 5200 to put in my moms machine. Maybe i'll try the XFX 5900XT this time since its suppost to be a 390/800 instead of a 390/700.

And just so i do this right next time, whats the best way to thoroughly test and make sure you dont overclock too far? I been overclocking at 10Mhz each time and then running through 1 loop of something like the game tests on 3DMark01 or the Mother Nature test on 3DMark03. Should i do more or use something else that works better?
 
Just an update here. I've braved up and re-installed RivaTuner and decided to start cranking the core again to see if its really damaged and i dont think it is. I think its just like Oldman said about the pixel shader portion not being able to handle the overclock. I'm running all the way through 3DMark01 at 470Mhz core right now with no artifacts that i can see and its not locking up any more. I think my issue was i was trying to run my RAM overclocked around 780-800Mhz when i was running 450Mhz at the same time and that was causing it to lock up. If i remember correctly, when i starting going crazy overclocking using 3dmark01 to test i left the memory at the default 700Mhz just like it is now and i cranked the core up to see how far it would go testing every 10Mhz or so.

So hopefully nothing is damaged and everything is AOK.

Now could someone please help me out with benchmarking in games? Whats the best way to do it? I downloaded fraps and i want to figure out the difference in frame rates in Morrowind between stock speeds and overclocked but how should i test? The frame rates fluctuate so damn much in Morrowind when your in exterior cells that its hard to really notice a definite increase or decrease.
 
I generally just use FRAPS or an in game frame counter and set my game so that most of the itme it doesn't drop below 60.
As for actually benching games...

I dunno. Not my thing.
 
You went wrong with using 3DMark2001 to test OC, IMHO. My card can definitely OC higher without problems using 2001, but that doesn't mean anything. What you really want to find is how far it can go under the most stress, aka 3DMark03. That is the best way to go in terms of safety. How to tell if you went too high, without using specific applications:

Core:
Lock-up/freeze or "snow"; random white pixels. Does NOT have to occur very often, just has to occur, period.

RAM:
Artifacts, usually random polygons and such; darker colors present. Does NOT have to occur very often, just has to occur, period.

As for game benchmarking, you should be able to get stable FPS by standing still in the right spot. You can also run timedemos or custom replays, although again most of it relies on you...if you want something "objective," try 3DMark03, although I'm not a huge fan.
 
I been running 3dmark03 for a while without problems and i tried something different. I benchmarked the Mother Nature test again but with 4x AA enabled and the card is running at 440/800 and it artifacted all over the place on that bench along with locking up for a few seconds once.

I backed it down to 440/780 and i dont think its doing its artifacting now with 4x AA enabled. It also didn't lock up this time and i'm looping through the Mother Nature test several times.

What tests and options should be used in order to push the video card the hardest while looking for lockups/snow and artifacts??? The Mother Nature test on 3dmark03 seems to work the card the hardest and 4x AA pushes it even further. Suggestions?
 
I see now just how worthless 3dmark01 was at testing my video card. Here i thought i had about killed my video card core and it turns out that those extremely high speeds i was hitting before was just because those game tests in 3dmark01 dont get the card hot enough to much more and bother running them. I just ran through the four game tests with the core at 490Mhz and the RAM default. Didn't skip a beat or lock up once. Another thing that could be slowing down my cards performance SLIGHTLY now is i've got alot more cables in the front of the case blocking some of the airflow and i have 2 raptors installed right in front of my video card and i'm sure their releasing their fair share of heat with the intake fans blowing over them. I'm glad i just ran 3dmark01 at close to 585Mhz instead of thinking it could do that high of core speeds and trying to run Far Cry or something with it lol. I really would of cooked it then.

I was using RivaTuner also and i dont think its working right or something. It seems like the memory increases correctly but i'm not sure its actually overclocking the core speed.

Whats a good tool to detect my clock frequencies with?

How does RivaTuner work anyways? It looks like it just adds its own CoolBits value to the registry except it sets the value at 2 instead of 3.
 
I tried using that ATI tool's artifact finder with my ti4200, but the program simply caused my system to reboot. It even said something like "We have detected that your video card is not ATI based, would you like to disable most of the features but just leave the artifact finder on for non-ATI based cards?" but it just causes a reboot to me.
 
Lowtax said:
I
If a GPU has lost its ability to hit a certain core speed that it could once hit, it has permanent damage. This is irrefutable fact..
Click to expand...
I did that to my 9700 :eek:

And I didnt overclock it that high though. :(. It aint that bad though. I just doesnt like drawing shadows. And some HW T&L games.
 
Well whats the most effective way to test for artifacting and core lockups? I know people say use 3dmark03 but do you just use certain tests or adjust any of the settings like turning on 4x AA? I know when i turn on 4x AA it really pushes the card alot harder in the Mother Nature test.
 
The catch is if you have artifacts in anything, then some part of the card, be it the memory or just a portion of the core, cannot handle that speed. It is causing a mathmatical error and isn't 100% stable.

Try to get it cooling better, try some applying some thermal paste.

If you only get artifacts on nature that means the pixel shaders are hitting their limit. Some other program will have the same problems.

Think of it along the same lines as a CPU. My system was 99.9% stable, once a month I had a random lockup I couldn't identify. Every app I ran was stable, but on occasion it would reboot when I started Powerdirector Pro. I let it go as I assumed it was just a software problem. Later I decided to see just how stable my overclock was, I ran Prime95, it failed after 45 minutes. dropped from 2166 to 2000, it failed after 15 minutes. Dropped it to ~1800MHz, failed after 15 minutes.

Removed a stick of Crucial PC2100, clocked it at 2166 and it runs with no errors. No random reboots, 100% stable 100% of the time.

Honestly the best test for artifacts I've used was ATItool. It is pickier than anything else. As for the failure with the Ti4200, report it to W1zzard, he wrote it. I've used it on mine since version 13.
 
Thanks Oldman, i'll check out ATItool and see how it does on my 5900XT.

Do you have any idea why i can overclock my core so high without problems using RivaTuner but if i try to even overclock alot less through CoolBits it locks up when i run 3DMark03? I ran the core at 500Mhz using RivaTuner and ran through the Mother Nature test without a problem. Tried 450Mhz using CoolBits and it started locking up. RAM is left at default 700Mhz while testing all of this.
 
Lowtax said:
Good lord.

Yes, you have probably damaged your card. What the fuck were you thinking?
Click to expand...

+++++++++++++++

This is why I don't oc, I don't have the money to repair...
 
JSClark said:
+++++++++++++++

This is why I don't oc, I don't have the money to repair...
Click to expand...

If you take your time and do it right then usually there isn't much chance of damaging the card. It depends though i guess. I actually was taking my time and only overclocking the core starting out and only at 5-10Mhz increases and then testing. I just thought 3dmark01 would push the card hard enough i would know when i went too far. I dont think i really damaged it because i can clock it at the same speeds as i did before using RivaTuner and it will run 3dmark01 without locking up. I'm not sure whats going on though because it seems like RivaTuner is overclocking the RAM but not the core and if thats true i never really overclocked the core very high to start with.

I need a utility that will tell me just exactly what my clock speeds are so i know if RivaTuner is working or not.
 
Are you using the hardware overclock or the software overclock in Rivatuner?
Sometimes Rivatuner cannot control the clock gens accurately directly, if using the software overclock on Riva, it is essentially coolbits.
 
You can't use the hardware overclock on the new nVidia cards. It all has to be driver based. Why does RivaTuner add the value "2" instead of "3" to the CoolBits registry? Does it do something differently?

I decided to just uninstall RivaTuner, delete the registry entry, and re-activate CoolBits. I'll mess around with it some more and see whats going on. I downloaded ATItool to test with.
 
Is there any difference in overclocking with coolbits or "hardware" as you speak with rivatuner? Whats the difference?
 
Rivatuner using hardware overclock changes the clock by the onboard clock generators.
Coolbits uses the Nvidia drivers to control the clock generators.

Rivatuner hardware mode works on some cards that coolbits doesn't (my old TNTs, TNT2s, Vantas, etc )
Rivatuner has to be programmed for the clock generators, so the hardware overclock does not always work.
 
bobmanfoo said:
BG101, since you artifact at 450, i'd use 430 max for gaming, 440 for benching
Click to expand...

when you're benching for max score you often get the best scores at a speed that is artifacting slightly
 
Ok i downloaded and installed ATITool and i'm working on overclocking my RAM. I started up ATITool after i raised up my memory speed to around 780Mhz using CoolBits. Do i need to change any settings in ATITool or just click "Scan for Artifacts"? And how long do i need to let it run? And this only checks artifacts from the RAM right? I'll need to run something else like 3dmark03 to test my core speeds correct?
 
You must log in or register to reply here.
Back
Top