Kd7

[water_cooler 20]

is it true that next kd7 bios is going to beable to lock the pci and agp speed

 

[Sonny]

I don't think that's possible with the chipset. Why would you want that? It makes FSB overclocking useless giving you slower performance.

 

[none1]

Goal is to get highest FSB / highest memory bandwidth. Many PCI cards can't go to FSB = 200 / 5 = 40. If we could LOCK the PCI / AGP, then we could easily get to FSB = 200+ with good DDR400 RAM and an unlcoked processor.

 

[Sonny]

At the cost of weaker performance through your AGP/VC & PCI/storage(a good example)=. The idea of FSB overclocking is to speed up the whole systen & not just the CPU to NB relationship.

 

[none1]

At the cost of weaker performance through your AGP/VC & PCI/storage(a good example)=. The idea of FSB overclocking is to speed up the whole systen & not just the CPU to NB relationship.

Yes, you are correct, there is a trade off between CPU/mem high FSB and PCI/AGP. This is why the KT266a (1/4 divider) still holds the overall speed records, because at FSB = 200 its PCI/AGP are running faster than the KT333 or KT400. But most PCI/AGP cards can't go that fast.

At this time, however, I believe the CPU/MEM improvements far outweigh the PCI AGP overclock. With a top of the line video card, the gaming bottelneck is in the CPU to MEM bandwidth, not the CPU to AGP. In SETI, the limit is CPU to MEM, etc..

For most applications I use, a FSB = 200+, with memory at DDR400+ speeds and PCI/AGP at 33, would give me better performance than FSB = 180, PCI = 36. The extra 20 mhz of memory / FSB will give me better performance than 3 mhz of PCI.

Now, IMHO, I am more in favor of a 1/6 divider instead of a PCI lock at 33. With a 1/6 divider, I can hit FSB = 200+ and still OC my PCI, AGP, which I believe is a good thing.

AND, while I have seen no PROOF of this, OTHERS hope that a PCI / AGP LOCK will allow them to run the PCI AGP INDEPENDENT of the FSB, so they hope that they could run PCI = 38, with ANY FSB. We'll see if this materializes with the nforce2, I doubt it.

 

[Sonny]

Finally someone else thinks that the KT266A is still the speed king. Where were you when I had that KT266A VS. KT333 thread

AFA PCI & AGP overclocking & CPU/MEM performance trade-offs it would depend on how the whole system is built & how it will be used.

What would be really be something if we could use either a 1/4, 1/5 or 1/6 whenever we wanted to. Now I know that Abit allows its users to manually select the divider to come in at whatever FSB settings they choose but I dont know of anybody with a KT333 chipset that is running 180MHz/360FSB with the 1/4 divisor only, actually very few have stable systems when running between 133 & 166. I only have a few ideas on why but have not tested enough to find out the truth, thinking higher PCI latencies with the older chipset to compensate for the lack of parking on the SB & less with the newer chipsets. The VT8235 has the park feature but that also means less latency which is a good thing & maybe less tolerant of high PCI bus speeds which is a bad thing.

I doubt if the NF2 will be able to independently adjust their PCI or AGP bus speeds. That would be amazing but highly unlikely.