Which matters the most, memory or core-frequencies?

[NixPhoenix]

I´m having a little dispute with a guy about GF3 overclocking and what it is that makes
the biggest difference in performence, overclocking the core or the memory.
None of us has a GeForce3 so we can't find out for ourselves...I checked a couple of reviews but couldn't
find anyone who just had overclocked one at a time.
With Geforce 2, it's the memory-speed that matters the most, right?
I thought it was the opposite with GF3?

Any help would be appreciated
And sorry for the strange english, I've not had that much of practice

 

[MilkPowderR]

Yep you thought right.

GF3 lines are Core OC dependent. Unlike the GF2's.
GF3 performance increases significantly by raising the core. Raising the memory only increase the performance very little. I'm planning to flash the GF3 Ti200 BIOS and put a SK-6 w/Delta50CFM on the core hoping to reach 300+mhz Core hurrrr hur hur hur hur...

 

[oldfart]

MilkPowderR, could you enable Private Message in your profile?

 

[MilkPowderR]

Yah Olfart I just enabled it now. sorry about that.

 

[darth maul]

<< GF3 performance increases significantly by raising the core. >>



Uh? What you been smoking? Got benchmarks from games to prove that? I could have swarn that it is still largely memory dependant, but hay I am bound to be wrong once.

 

[BFG10K]

Most cards are entirely memory bandwidth limited and show the biggest increase in speed from memory overclocks. Notable exceptions are the GF3 which shows some gain from core speed bumps and the Kyro/Kyro2, which is all about the core speed.

 

[MilkPowderR]

darth maul,
BFG10k also agrees to my findings about GF3. I told ya they're a bit different than the GF2 lines. Yeah I thought they all are memory dependent but not all cards. GF3's engineering process must be different. i don't have the acutal screeshots cauze I never had chance to do it. Beside, even if I show u the screenshot, how would u believe me? From my experience with GF3, and from repetitive tests of mine shows that increasing the memory by 50mhz gave me 100 points in the 3DMark 2001. Also very little increase in FPS in Q3A. Whereas in Video Core, increasing it by 10mhz gives me approx. 100points! I tried 20mhz core increase, gave me nearly 200 points (192 points to be exact) and goes on as you OC the core further. OC'ing memory helped.. but not much.. just a little.

 

[NixPhoenix]

Okay, thanks for the answers

 

[Rand]

In my own experience the GF3 tends to gain equally between overclocking the core and memory frequency, it seems to be pretty well balanced overall. This refers to the TI200 and regular GF3. The Ti500 tends to gain proprotionately more by overclocking the memory clock, rather then the core clockspeed. Which stands to reason as the GF3 Ti500's memory clock is only bumped slightly from that of the regular GF3 while the core clock is significantly increased.