View Single Post
  #18 (permalink)  
Old 4th July, 2004, 12:06 PM
jnansley's Avatar
jnansley jnansley is offline
Member
 
Join Date: May 2004
Posts: 888

Daniel is smart

So let me get this straight. Of course, I don't own one of these boards yet. Still keeping my options open before I spend $300 on a CPU and board

The CPU speed is a multiplier of the HTT bus.

The memory speed is a multiplier of the HTT bus.

The actual HTT speed is a multiplier of the HTT bus.

But, like 12 months ago, everyone said that the memory speed was going to be a divisor of the CPU speed!

If you think about, it's just semantics.

But as to the "asynchronous/synchronous" terminology... memory accesses are not travelling over the HTT bus, correct? Otherwise you'd need an HTT endpoint at the DIMM slots, and I don't see any of those. So, if:

HTT bus = 200mhz
HTT speed = x3 = 600mhz (this really doesn't matter)
CPU speed = x9 = 1800mhz
Memory speed = 5:4 = 250mhz

This is truly an asynchronous situation, as the memory speed is 1/7.2 of the CPU speed. However:

HTT bus = 200mhz
HTT speed = x3 = 600mhz (this really doesn't matter)
CPU speed = x9 = 1800mhz
Memory speed = 9:8 = 225mhz

This is actually a synchronous situation, despite the wacky memory multiplier, because the memory speed is 1/8 of the CPU speed.

Assuming that your memory runs a max of 250mhz, then you, the educated overclocker, would proceed with the aid of a calculator to test it at both 250mhz, and 225mhz, and see which gave the best results. And I'm ignoring the effect that the HTT multiplier has on AGP/PCI/etc performance - should I be? After all, my 9800p turns out lower scores at 70mhz AGP than 66mhz AGP.

Or is my whole theory incorrect?

Edit: Apparently what I'm calling "HTT Bus" is being called "FSB".
__________________

Last edited by jnansley; 4th July, 2004 at 12:12 PM.
Reply With Quote