question about how cpu overclocking works

[bkolk]

so ive read that to overclock a cpu you need to overclock the ram at the same time. Normally they both run at 200mhz, and i get if you raise them both how the overall clock speed can be effectivly increased. But if a ram divider is used so that like the cpu HTT is set to like 220mhz and the ram is left at 200mhz, how is the system overall getting faster?

 

[FghtinIrshNvrDi]

The CPU is increasing how much work it can do by adding more cycles/given time. The ram bandwidth just isn't increasing.

Ryan

 

[bkolk]

so how effective would simply raising the HTT with the same multiplier for a higher clock speed be if thats all you did?

 

[FghtinIrshNvrDi]

highly. Of course, when you reach a certain speed, your ram may not run at that speed, or you may need to increase vcore. So the first steps involve running a memory divider to let your value-ram run underclocked slightly, while you keep bumping up the HTT.

Ryan

 

[bkolk]

from this site http://www.neoseeker.com/Articles/Hardware/Guides/OCguide/2.html it says
"There are several things that can be done to help out with overclocking most of which revolves changing the FSB:Memory ratio, however this can potentially negate any gains you might get from anything you may get from a processor overclock."

Which leads me to believe its worthless to overclock my processor HTT unless the ram can match it. Maybe its my lack of understanding FSB and how it works. My mobo and processor are spec'd at saying they have a 1ghz FSB, tho reading OC guides it makes it sound like the 200 mhz HTT is the FSB.

 

[gaara]

Consider the following:

Every single AMD64 uses an intergrated memory controller which translates into the fact that there is no longer a direct reliance on FSB speed for memory bandwidth as there is no communication required between a north bridge anymore. Therefore, all AMD64s operate at 6.4gb/sec I/O bandwidth

Every single AMD64 still uses the default 200MHz memory bus speed regardless of whether it is the 3000+ running at 1.8GHz or the FX-57 running at 2.8GHz...therefore if you were able to push a 3000+ up to 2.8Ghz while running your memory on a divider to force the memory bus to remain at 200MHz, you would still have a CPU roughly equal to an FX-57 regardless of no memory bandwidth increase

While it is true that increasing memory bus speed will in turn increase bandwidth and overall performance, the AMD64 architecture is built in a fashion that your I/O bandwidth is no longer crippled by relying on a link between your CPU and a seperate memory controller. Furthermore, a faster frequency CPU still translates into a higher flop rate which in turn leads to faster processing regardless of memory bandwidth