Memory bus and effect on fillrate

[ragejg]

Okay, so I have this FX5200 from gainward that I paid $90 for... FIRST I expected an Ultra, THEN after I got it I at least expected a full 5200, now I realize (i think) that I'm stuck with one with a 64bit memory bus... 250/280 are my stock speeds... I updated the bios on my card (I used an older one actually... .15 instead of .18...) and still get the same results... I've done everything... 3dmark gave me a *decent* single/multitexturing score ONCE... that was it... Now it's ALWAYS POOPY... right in between a GF2mx and a GF4MX420... WHAT? IS THIS A FRIGGIN JOKE?

here, check out the scores here...
http://www.digit-life.com/articles/gf4mx/

There's a guy on [h] who's trading his ti200 128mb for this card... I think I'm gonna rma it and try to get an *actual 5200 instead... so I don't have to send him such a poopy card...

does memory bus have this direct of an effect on fillrate?

did nvidia ever disclose that it would be allowing distribution of 64bit style 128mb 5200's?

ugh...

[jbirney]

does memory bus have this direct of an effect on fillrate?

Yes and no. The memory bus is one of the big factors in determining the memory bandwidth the card has. The other big factor is speed of the memory. So multiply the two to get a idea of your Memory Bandwidth. Now were this effects fillrate is that if you data that your trying to draw requires more memory bandwidth than you have, then you can not draw it as fast and thus your effective fill rate drops.

A good example is looking at older 3dmark scores for fill rate test run at 16/32 bits. You will see a large jump in the two on some cards. For example the GF2 Ultra. Here it had gobs of fill rate. But its memory bandwidth choked it :(

[ragejg]

yeh, rivatuner detects the video ram as 128 bit, ddr type 7...

AAAA!!! where did my fillrate go? Why 4.5k 3dmarks?

Is a clean running kt266a with either a t-bird 1.4 or XP2000 THAT much slower than all the test configs in reviews? ...sheesh...

[marcocom]

that sucks dude.

hey just for safeties sake, im sure you know this, but we like to be thorough here... have you installed the VIA 4-in-1 driver pak for your current operating system? because without it, you will get the same results. (no AGP port acceleration)

also, you have to disable at all times ANISOTROPIC FILTERING (have the slider always at 0) and always disable TEXTURE SHARPENING (which is just anisotropic filtering anyways) or else you will get really bad scores as well. Anti-Aliasing is fine and will work well with your card, but dont be sure its fully disabled before benching as well.

RMA isnt what you want. if you ask for it, you will be denied. RMA is for defective workmanship and that is the only reason the mfr (not the retailer. thats not RMA) will exhange it. what you need bro is a REFUND from your retailer. they usually give you at least 30days or more and just charge a restocking fee of 10 or 15% usually. they can take back your purchase and exchange it.

[ragejg]

well, because this is one of those 5200's that was supposed to be an ultra (darn gainward AGAIN), Computergate is having ups pick up the card, and they're gonna credit me the full price BECAUSE all of their stock o these was bought, then 75% of it was returned (I wanted a Pro/760 BAD!!)... Then I'm gonna send the asus fx5200 to the guy who's trading his ti200/128 to me...

And yes, my 3dmark scores so far have all been done with everything off and low, no tex sharpening, no AF/AA... been there many a time...

and I've tried my "favorite" 4.40's AS WELL as the new hyperions... no change, really...

for more on this topic, go to my "tooth and nail" thread in this section...

ugh...