Could someone explain the relationship of pipelines and core/memory clock?

[WolverineX]

I was shopping around newegg and I'm making a list of potential contenders. So my question is does more pipelines allow for slower core speeds? Or should I be going for the best I can get.

I'm looking for a AGP card, planning on waiting for price reductions... don't want to spend too much, around 200, since I'm saving as much as possible at the moment.

I have my eye on either a BFG 6600GT or an EVGA 6800 ... so which would be better to get, less pipelines but higher clock speeds or vice versa?

Thanks

 

[monster64]

1) Pixel throughput, this is calcualted by multiplying the core speed * # of pipelines (newegg lists as Pixel Pipelines) EX:
6600GT = 500 MHz * 8 = 4000 megapixels / sec
6800GT = 350 MHz * 16 = 5600 megapixels / sec

2) Memory throughput, this is calculated by multiplying the memory speed by the bit width then diviting by by 8 (8 bits per byte): EX:
6600 GT = 1000 MHz * 128 / 8 = 16000 megabytes per second
6800 GT = 1000 MHz * 256 / 8 = 32000 megabytes per second

In your case, the 6600GT gives 4000 megapixels/sec, while the 6800 gives 3600 megapixels/sec. In terms of memory, the 6600GT gives 16k mb/s, while the 6800 (700*256/8) gives you 22k.
However the 6600GT has a little more megapixels/sec, it gets a lot less MB/s. Also you have a chance to unlock a few pipes with software, and then the 6800 would by far be better.

 

[WolverineX]

Interesting, thank you very much monster!