Overclocking mysteries

I have really strange things happening around here. I’ve been overclocking video cards for my friends- mostly Geforce 2 MX series. And everywhere then video card was pushed to the limit it started to add noise in screen or add holes everywhere. It looked really simple.
But my Geforce 3 Ti-200 worked just right with all games, so I didn’t see the point of doing this. Until the Colin McRae Rally 4 came… It wasn’t working as I wanted so then the fun began…

I’ve tried overclocking with original cooler, but it was heating very quickly. And then the mysteries began.

Here are the things that I don’t understand:

1)
My original Geforce speeds are GPU-175Mhz; Memory-400Mhz.

I’ve overclocked GPU to 195Mhz and strange thing happened. My average frame rate increased, but minimal decreased. I know that processor was overheating, but why do this result on only minimal frames not whole framerate? (Minimal framerate dropped from 22Fps to 19Fps, on 190Mhz to 21Fps)

2)
Then I thought: anyway, I have to move that heat away. I took a big cooler from PII and attached it. It was sitting nicely and correct (I think). Now on 195Mhz frame rate was 23Fps. Nice… . I went further, from about 200Mhz another strange thing happened: about every 5-10 minutes game freezes for 0.1s. I’ve been playing for 30min. and I can say that freezes are totally random, it doesn’t matter what game or place in map. This is another question: why do this happen? Going further gave me even more and longer freezes.

3)
I went further not looking to those freezes. I have review from a guy who was fighting to 225Mhz and then he did that he said that it’s the limit for GF3Ti-200. I’ve reached 225Mhz quite easily. (Those freezes was really getting on nerves then- it was every minute and it freezes for 1sec. or sometimes even more!). I thought that I reached the goal, but my father said to go further. And I went for 235Mhz, then another thing happened. Usually then Video card is overloaded it ads holes in walls and objects, but not now. After 5min. of testing my PC hanged and I got green screen over the entire monitor. I’ve seen this first time in my whole life. I panicked and as I remember even reset button wasn’t working. I’ve restarted computer in brutal way, and ran the test again. Then I’ve had a blue screen . It was blue window of death, at least CTRL+ALT+DELETE was working… 235Mhz is too high for me… At the end of the test my GPU temp wasn’t still so high, it was only about 50C, or even this is too high??? But there is that colored screen mystery that I don’t understand, what is that and why is that? Usually it’s holes, not colors…

Big thanks to everyone who’s reading it. And I’m sorry if it was already posted, I didn’t see it.
And I’m sorry for my English, I’m Lithuanian so I’m not so good in it.

My full system spec.

AMD 1900+ 0.18µ
DFI AK75-EC
WD 80GB 7200rpm
Creative SB-128 sound
USB 2.0 controller
Daytona Geforce 3 Ti-200 64MB
Lots of Logitech stuff.

Temps on full load: CPU-40C; Case-28C; Vide card 175Mhz-35C, 230Mhz-50C

Sounds like your ram is overheating. The main way to tell whether it's your ram or GPU overheating is the type of artifacting you're getting. GPU artifacts look like little white dots all over the screen. Artifacts caused by your ram overheating look like blocky aftifacts, tearing, holes in the image, etc. Hot ram can also cause the 3d app you're running to lock completely.

My suggestion to you is to grab some good copper ramsinks, and mount a case fan right next to your card to dissipate all that extra heat.

I frogot to say that my memory wasn't overclocked

And I remembered to ask another thing. Then my GPU runs on 195Mhz I can push my memory to 470Mhz, but then GPU is running 225Mhz for memory 440Mhz is all I can get.
And again: what's happening here?

My memory is already heatsinked, and on 470Mhz it's about 50C

I have a power suply problem here, my +5V chanel gives only +4.81V. Can this be a problem?

That could be a problem, but undervoltage or a sagging +5 rail wouldn't be causing your card to artifact (at least I don't think so).

What kind of PSU do you have? Vendor? Wattage?

I'd say to fix your artifacting problem, just start dropping your core speeds. You might have hit the ceiling for that card. Also, keep in mind that just because some other guy says that he hit 225 w/ his core, doesn't guarantee that you'll hit that same speed or higher and still be stable. Using myself as an example: My 9800 pro is clocked @ 425 (up from 360). There are several other people on these forums that have hit 450 w/o volt mods. Still doesn't guarantee that I can get my card that high, even if I mod it.

OC'ing is a lot of know-how, but there's more luck involved than people think.

Well from my experience, artifacts are the result of too much oc on the mem. Temporary freezes are from gpu being too much. Given i generally run stock cooling, but this has been the norm through out the line of cards i had. ti200, 4200, and 5900se (not so mcuh with the FX card tho)

But as yellowdart suggests, it could be a powersupply issue. Also, what brand card you have?

I have standart Target 300W PSU.

4.81V comes to my MB, PSU gives 4.91V from himself.

I'm thinking to buy Qtec 450W-500W or Target 430W (a good one).

And my video card is Daytona Geforce 3 Ti-200 64MB.

dont get q-tec they are really **** i had one my 12v was 12.21 at stock of every think and when i overclock 12v go to 13v and 5.01v go t 4.50v
get a fortan or antec true power
they will do all the job well

Like I could get any of these... I live in Lithuania, there Qtec and Target is averything you can get.

I'll try to order then I now I know what to, but I'm shure- it will be imposible to get.

Guys, you've answered a few of my questions. Anyways my reason to OC the GPU was to experience the textures and detail with UNREAL03.

But I have to ask...how do you monitor GPU temps? Aftermarket probe or is there software?

Thanks again.

You can use a probe. Although checking temps on the gpu isnt like you do on a cpu. As long as it stable and its shows a normal picture ingame etc, no harm will be done to the card.