artifacts and causes

[cadaveca]

i have done countless searches on the net as to what it is that causes artifacts to appear during overclocking. The most obvious cause is data corruption due to heat, but nobody has explained why it is some people get striations on the screen OR everything on the screen "pulled" to the center.
I know a fair bit about electricity and am an electronics enthusiast (my wife hates it but will begrudgingly admit when i bring home new $4000 speakers, they DO sound better). I've fixed stereos and tv's. I've made my 135F running stereo run @ 60F. So naturally I have come to some slightly educated conclusions about how to avoid those artifacts that make all of us cringe.
Now the first thing is that i am pretty sure of is that there is no way to increase voltage @ the gpu without some soldering being done. The second is that in order for electrons to move @ a greater wavelength requires more power or less resistance. So, in order for us to overclock properly we probably need more voltage @ the gpu (just like overclocking your intel or amd cpu). Higher voltage/less restance lead to an extra amt of energy that must be dealt with, usually in the form of heat. This is easy to deal with..added heatsinks and a fan make all the difference in the world.
So now to my point. How is is exactly that the software we use to overclock our video cards overclocks them? Only the "xt" variants of the radeon series carry varialbe resistors on-board, so i know how the work, and have replace the resistors there on a 9600xt to allow for higher overclocking, but to no avail...i just simply get artifacts.
Hopefully someone will be able to answer this for me so i may push this card more.

[Aedan]

The GPU and other components are driven from clock signals derived from a crystal. As the fundimental frequency of this crystal is too low, it is multiplied upwards by a PLL. As the PLL can be programmed to provide a particular frequency, all the software tools do is change the multiplication used by the PLL.
Hence, to overclock a GPU or memory, you just have to reprogram the PLL's internals.

Now to the other part of the problem! It's not the moving of the electrons - they'll move when you apply a potential difference across them. The problem is making out the difference between a 1 and a 0 when the data gets to the end of a transmission line. Increasing the voltage to the I/O buffers helps to increase the difference between a 0 and a 1, and hence increases the noise margin.

Heat causes a problem too, as the hotter a component gets, the larger the thermal noise that is generated. Once this noise gets to the point it could be misintepreted as a 0 or as a 1, then the game is over. This happens easily, as at RF frequencies, the I/O buffers will be seeing a bunch of cascading edges, rather than discrete voltage levels.

Once you look into high speed logic design, you begin to realise that it's amazing anything works at all!

[cadaveca]

[QUOTE=Áedán]
Now to the other part of the problem! It's not the moving of the electrons - they'll move when you apply a potential difference across them. The problem is making out the difference between a 1 and a 0 when the data gets to the end of a transmission line. Increasing the voltage to the I/O buffers helps to increase the difference between a 0 and a 1, and hence increases the noise margin.

QUOTE]

my real question was kinda along the lines you were heading, however the "frequency and bits(as you call 1's and 0's) are generally measures of wavelengths....the higher the frequency the higher the wave the harder to contain. thanks of the pll info....i was aware of that but i guess it must be in the gpu bios where the default frequncy is(66mhz i would presume). btw..electrons are always moving.if they stopped we would all be dead.

[Aedan]

In terms of wavelength - that depends on the bit sequence that's being transmitted. Different sequences will result in different wavelengths and waveforms.

The PLL generates several frequencies, depending on the design in use. On something like an nVidia card, the core clock and the memory clock are independant from each other. The 'default' frequency will be what the PLL comes up with when it's been reset. The graphics card BIOS can then change the GPU and/or memory clock to something more suitable.

The AGP frequency is usually derivied from the clock generator onboard the motherboard, and hence the graphics card has no control over it.

[Gizmo]

Umm.........actually, the wavelength of the generated EMI depends on the signal transition speed or edge, which will normally be largely independent of operating frequency, although it will tend to limit the maximum frequency of operation.

Or am I missing your point?

[cadaveca]

so is it somewhere in the bioss that programs the write /re-write times of the mem?

[Daniel ~]

I just realized that as scary as the knowledge base that's found in these forums is, what really frightens me even more is when you guys disagree!!!

Mostly because I get this creepy feeling that Heisenberg's cat may be trapped in one or more of my boxes ( depending on whose looking and in which of my boxes they're looking in) and trying to find it's way out!!

[Gizmo]

Actually, I don't think Aedan and I are disagreeing. I think it is a case of us looking at two different problems.

And I thought it was Schroedinger's cat?

[Daniel ~]

Damn! I hope your wrong! I just came across his name in connection to the uncertainty principle and few days ago...

This is exactly why I make you all post names above your avatars!! ":O}

[Gizmo]

It is Heisenberg's Uncertainty Principle.

Schroedinger did a thought experiment involving a box, a cat, a hammer with a trip mechanism, and a bottle of cyanide. This was done to illustrate Heisenberg's Uncertainty Principle. So the two are in fact related.

[Daniel ~]

Ah so, Hopefully this was an experiment done in thought rather than with a an actual cat, other wise Schroedinger can go screw himself.

Thanks Gizmo!

[Staz]

Are we talking about the cat existing in a dead alive state and that observing the cat could in and of itself affect weather the cat is dead or alive when observed?

That would definately be Shrodingers cat. And since it has been in the box so long without food or water that would take all the uncertainty out of it.

[Aedan]

Fortunately for the cat, quantum effects don't appear to happen on such a large scale. Hence, the cat wouldn't have to exist in two states - it'd be either dead or alive.

[Aedan]

Originally Posted by cadaveca

so is it somewhere in the bioss that programs the write /re-write times of the mem?

Yes, somewhere in the graphics card BIOS there is code that reprograms the clock generators with the correct values. For some graphics card BIOS, there may be several different values that can be used, dependant on the GPU in place. This way the same BIOS code can handle different speed GPUs of the same family.

[cadaveca]

Originally Posted by Áedán

Yes, somewhere in the graphics card BIOS there is code that reprograms the clock generators with the correct values. For some graphics card BIOS, there may be several different values that can be used, dependant on the GPU in place. This way the same BIOS code can handle different speed GPUs of the same family.

so how do i extract th bios from thwe card and modify it? i know how to flash to the card, but not how to retrieve....if i am right instead of having a 324 gpu/189 mem card i'lk have a 425/250!

[cadaveca]

Originally Posted by cadaveca

so how do i extract th bios from thwe card and modify it? i know how to flash to the card, but not how to retrieve....if i am right instead of having a 324 gpu/189 mem card i'lk have a 425/250!

so i've modified the bios and all is good..the samsung 250 mhz mem is now running @ 250...with out software like atitool or radlinker....i just don't understand why they use the mem @ 189 stock..why not just set it to 250?
btw ..since i changed the bios so the refresh times where right the mem produces less heat...like 4c less( and no i can't overclock higher because of it..the max the mem does is 250...period)... the best that i can make of that is because the refresh rate was lower for the higher frequency(from the mem's whaite papers from samsung) the mem was holding the info less(1.25ms) as so it now actually has less current in it in an active state. Sound right?
(mod involved changing tras to 4.25ms from 5.5ms and i did not write the bios change. the lower refresh rate was what prevented me from taking the mem above 229.5...the max samsung listed was 225 for the old refresh rate so i think the 229.5 was pushing the mem and proly woulda burnt it out shortly)

[cadaveca]

would help if i still had the card.
bought a second of the same.