Archived from groups: alt.comp.hardware.overclocking (
More info?)
Bracken wrote:
> David Maynard wrote:
>
>
>>Anton Gysen wrote:
>>
>>
>>>Bracken wrote:
>>>
>>>
>>>>would this work? I'm not concerned about if it is unstable, as I can
>>>>just change it back to 166MHz.
>>>>But I want to know whether this risks damaging the CPU or not? I'd hate
>>>
>>>
>>>I know someone who sells P166s as P200s
>
>
> That's pretty silly because it has the CPU's speed clearly printed on the
> underside of the CPU.
>
>
>>Then you know someone who is the reason why processors are multiplier
>>locked nowadays.
>
>
> Too true. It is a pain having to run a CPU at a lower FSB speed than it should,
> for no reason than it's locked.
>
>
>>>He says it's fine as long as you use a HSF (obviously) and thermal paste.
>
>
> But it is dishonest.
>
> This computer won't even turn on when the P166 is set to 200MHz. It runs fine
> at 180MHz, but that's no advantage because the FSB has to be set to 10% slower.
>
You might want to run some benchmarks because the relationship between FSB
and performance is not the same on the older systems as it is on the newer
ones for the reason that the memory is not synchronous (SDRAM); it's either
FPM or EDO.
FPM and EDO timing is 'access time' PER access. I.E. usually 70 or 60
nanoseconds. At 60ns that's roughly 16.7 million per second, much less than
the 66MHz FSB. So, unlike SDRAM that runs at the FSB rate, changing the FSB
has little effect on how fast FPM/EDO RAM is accessed.
If the motherboard has cache, however, that would be affected, depending on
how fast the motherboard cache is. And it's possible that the slower FSB
could be compensated for by tightening the cache speed setting in BIOS
(I.E. might run with a faster setting at 60Mhz than at 66.6) which could
offset the lower FSB speed.
Many motherboards had no L2 cache at all so the issue is moot with those.
If, however, it was optimized for 66.6 MHz when made (I.E. cache timing
designed for 1 clock access [15ns cache] and RAM set to an even clock cycle
of 4) then the lower FSB would probably be hit.
That difference in system RAM and motherboard cache, btw, also means you
don't get as much performance increase per CPU MHz as one might expect. On
modern processors it scales almost linearly with CPU speed since the on-die
cache speed goes up along with the processor clock but with motherboard
cache it stays at the motherboard speed, which results in a 20% increase of
processor speed, from 166 to 200, being only about a 10% performance
increase. You get about half the effect. Whereas with, say a P-III 700,
increasing processor speed 33% improves performance by about 33% because
the L2 cache is on-die.