I wonder, I wonder. I've used the Geforce tweak utility to set my Herc Ti200 to 230/485 (100% stable oc) speeds, but I wonder if that is really the speed they are running at. Where does the clock signal come from? If there's a PLL (Phase-Locked Loop) on the gfx board, then those speeds I set in the GTU are accurate. But if the signal comes from the AGP system, they're not accurate, because I've overclocked my FSB. If the clock speed of the GFX is determined as a multiple or a function of the AGP speed, then one can presume that the speeds set in the GTU are assuming an AGP speed of 66MHz. If this is the case, then my GFX card is ACTUALLY running 247/521 (set speed * 72/66). Whilst this would be very nice, it is (obviously) not so likely, but given that there's a blowhole and ti500 cooling on the card, it could happen. Another thing which makes me think that this is possible is my previous Geforce 2 MX. When I ran 66MHz AGP, I could get it to clock at 225/395. As I increased the bus speed, it wouldn't clock so high, dropping to about 215/380, which was pretty well exactly in proportion with the FSB increase (138 from 133, => 4% increase....4% decrease in GFX clocks). Does anyone know where the clock is derived from, and hence the answer to this question?
Gfx card core/ram clock speeds are independent of fsb clock. fsb clock is controlled by the mobo northbridge and the gfx cards clocks are controlled by the gfx card cpu/gpu, just like AMD/Intel cpus control their own clock speed. How do i know? Cause when you ramp up the fsb without changing the clock speeds of the gfx card (the extra cpu is negligable in respect to the proportion of what you are raising it by: ie 1400Tbird @ 133 to 1460@ 137mhz the extra 60mhz doesnt do much in proportion to the 1400mhz already, compared to a few mhz increase in agp which is mutliplied 4x). Anyway, by experimentation, if we keep all the system settings the same for comparitive reasons: by increasing the fsb, but keep the core/ram clock the same in 3Dmark; scores rise. Then if you reset the fsb and increase the core/ram speed on the gfx card then the 3dmark increases. That way you have 2 varyables to get an increase in 3dmark scores when you have tweaked your system to the max software wise. Now the problem you are facing at higher fsb speeds is due to the fact PCI and IDE (very sensitive aparently) speeds increase as well as agp, so you overclock the whole system. Your puter is as strong as the weakest card in it - so people with the same cards, same mobos and same ram/cpu can have diff scores simply because they have different pci cards/ more ide components. That's why people with 166+ fsb usually never have pci cards and no cdroms and acpi disabled simply to keep it stable. Fsb = 1/3 PCI = 1/4 ISA = 1/3 or 1/4 IDE Even if you have no ISA slots its still how its calculated. 133 = 33 = 8.25 (used to be 7, that's another reason why ISA was fased out due to higher fsb speeds, cause they dont OC well cause they are usually older components manuf. by older techniques with less speed tolerences). = 2.75-3mhz That Help?
But AMD/Intel CPUs DON'T control their own clock speed. Their clock speed is a function (ie multiplier) of the FSB speed. This is what I am asking: If the speeds of the GFX core and memory are a multiplier of the AGP speed, or whether they are controlled independantly (which would create an asynchronous system, and as an electronic engineer, I am aware QUITE how difficult and unstable that would make life). That is to say... If you had a PIII 600, with a 100MHz bus. Increase the bus to 133, and you now have a PIII 800. However, let us assume that the increase in speed is not reported, just the speed that the processor would be running at an assumed bus speed, ie 100MHz. Now, you use another utility to change the speed, but this only reports the speed at the assumed bus speed. Using that utility (hypothetical), you set it to 650MHz, but in fact, since you've already changed the bus, the increase is bigger, taking the CPU from 800 to 866MHz. ie, you've changed the multiplier from 6x to 6.5x, but you had previously changed the bus from 100 to 133, so the increase in changing the multiplier is 66MHz, not 50MHz. Has the same thing happened with the GFX card? Ti 200, stock is 175/400, on a 66MHz bus. Overclock the bus to 72MHz, as I have, and then what is the speed on the card? Is it still 175/400? Or is it now 188/430, ie those components have now been increased by the same percentage as the AGP speed. However, the GTU still reports 175/400.....is it right?
Sorry i meant clock pulse is controlled by cpu yes the mobo does set multi n fsb. But agp is i think independent of gfx card clock settings - its just the speed at which the data is able to be transfered. However there must be some sort of corrilation between them to make the data come off syncronously with each fps n stuff. Otherwise there would also be no point in having OC functions on the gfx card.
Yeah...sorry, it's kinda hard to make clear what I mean in words. Anyway, does anyone else have any ideas?
Im not sure if this is relavant or not I used to run a Matrox G400 on a Slot 1 P3-450. I used the matrox tweak utility to overclock the card and it got very hot, so i stuck a fan on it and then it was Ok. A few monts later I overclocked my P3 by increasing the FSB to 133 (from 100). Unfortunately i neglected to clock the graphics card down and rather than crash it dot Extramely hot even with the fan. So... I think that increasing the FSB did overclock the graphics card more. Although im unsure if the increase in heat was just due to the AGP bus being faster or if the core/memory speeds had altered. I couldnt test for a speed increase as the CPU and memory where definately running faster.
