KT333 266 FSB, why better benchmarks with 333MHz memory?

[Oreo]

Why is it that you get higher benchmarks when you run the memory at 333MHz when the FSB only is running at 266MHz (on the KT333)?

 

[Buddha Bart]

Other things than the CPU access memory, thats the whole point behind DMA.

bart

 

[PrincessGuard]

Also, in case you didn't know, the KT333 and previous Via chipsets can run the memory asynchronous to the FSB (e.g. 166 MHz memory bus, 133 MHz CPU bus).

 

[ScaryGary]

ACTUALLY, some 6-layer boards simply ALLOW for better access.
SO.. it IS INDEED superior...
[where da ya get does pecs? DOD]

 

[Oreo]

Originally posted by: Buddha Bart
Other things than the CPU access memory, thats the whole point behind DMA.

bart

So when you're running a Q3 timedemo what other parts are accessing the memory?

 

[Buddha Bart]

So when you're running a Q3 timedemo what other parts are accessing the memory?

So long as it all fits into RAM (the textures, maps, AI, etc) nothing.

but, when you start playing online, you've got the occasional loading of textures off the drive, the network card, and the sound card, all throwing data back and forth.

bart

 

[Oreo]

But when you run a timedemo and turn off everything (sound, no internet etc) you still get higher scores with the ram at 333MHz vs 266MHz.

 

[BumJCRules]

So let me get this right.

You are wondering... "Why you are getting better results with the memory bus at an effective 333MHz verses at 266MHz with a 1:1 ratio with the FSB?"

You will get marginal gains out of having the memory at faster speeds than the FSB. The FSB is not waiting for the memory to return it's answer. The memory bus is wider and faster than the FSB. It gets it's job done and waits for the FSB to collect its info from the memory.