Whats more stable Higher FSB or Multiplier ?

[Ext User(Lordy)]

David Maynard <dNOTmayn@ev1.net> wrote in
news:10jjgrpr3utrfbe@corp.supernews.com:

> Show me what RAM you use that has a 'duty cycle' spec less than it's
> rating. Then show me the RAM you use that says it's PC2700, or PC3200,
> or anything else for that matter, but for only a 'particular
> processor'; as you claimed they make components for.

Exacty. Todays CPUs alreay have large multipliers compared to the FSB.
So they can already saturate the FSB without breaking a sweat. The FSB
throttles everything down so Memory, PCI, AGP and all other parts run
within spec.

In a way this supports Daves Cs ill-presented argument that multiplier
overclocks dont generally give much performance gain (but we all know this
already) as at the same clock speed, the faster CPU sits around twiddling
its thumbs more than a slower one. If you have very CPU bound
applications then this can make a difference. (Photoshop Filters, Encoding
video etc) but for Joe Bloggs nothing to write home about.




--
Lordy

[Ext User(Dave C.)]

>
> > if you need a faster system, build one. -Dave
>
> And overclocking is one way to "build one."
>

Well on that much at least we can agree. We'll never agree on the correct
way to build one, but overclocking is certainly one way. -Dave

[Ext User(David Besack)]

> Well, you're half right. Components are generally matched to the
> performance of a particular CPU. If that CPU is performing above
> specifications, then other components are working harder to keep up, even if
> their clock frequency is unaffected. That is, a CPU running at default FSB
> but higher multiplier will cause other components running at default FSB to
> have a higher duty cycle than they were designed for. If you are having
> trouble grasping that, think of walking two miles to the local grocery store
> at default walking speed. Now imagine walking two miles back from the
> grocery store at the same speed carrying ten full bags of groceries. Now
> you understand why changing a CPU multiplier can overload other components,
> even though they are operating at the same speed. By upping the CPU
> multiplier, you've just dumped many grocery bags on other components. -Dave

Sorry, but setting the multiplier one higher than spec will have no
difference on the rest of your system as replacing the CPU with a
slightly faster version. As long as your mobo supports the faster chip
it will support the overclocked slower chip.

[Ext User(David Besack)]

>>>if you need a faster system, build one. -Dave
>>
>>And overclocking is one way to "build one."
>>
>
>
> Well on that much at least we can agree. We'll never agree on the correct
> way to build one, but overclocking is certainly one way. -Dave

I would agree that if you *need* a faster system you should build one,
but for some people overclocking is a hobby. Hell I would make the
argument that most people don't really *need* a computer at all :)

[Ext User(Lordy)]

David Besack <daveREMOVEbesack@mac.com> wrote in news:chcral$br88$2
@netnews.upenn.edu:

> I would agree that if you *need* a faster system you should build one,
> but for some people overclocking is a hobby. Hell I would make the
> argument that most people don't really *need* a computer at all :)
>

What is this a Dave/David convention?

--
Lordy

[Ext User(David Maynard)]

Lordy wrote:

> David Besack <daveREMOVEbesack@mac.com> wrote in news:chcral$br88$2
> @netnews.upenn.edu:
>
>
>>I would agree that if you *need* a faster system you should build one,
>>but for some people overclocking is a hobby. Hell I would make the
>>argument that most people don't really *need* a computer at all :)
>>
>
>
> What is this a Dave/David convention?
>

LOL

Feeling left out?

[Ext User(David Maynard)]

Lordy wrote:

> David Maynard <dNOTmayn@ev1.net> wrote in
> news:10jjgrpr3utrfbe@corp.supernews.com:
>
>
>>Show me what RAM you use that has a 'duty cycle' spec less than it's
>>rating. Then show me the RAM you use that says it's PC2700, or PC3200,
>>or anything else for that matter, but for only a 'particular
>>processor'; as you claimed they make components for.
>
>
> Exacty. Todays CPUs alreay have large multipliers compared to the FSB.
> So they can already saturate the FSB without breaking a sweat. The FSB
> throttles everything down so Memory, PCI, AGP and all other parts run
> within spec.

Agreed, although I wouldn't use 'throttle' because I think it might give
impressions that may not be quite accurate (It isn't as if it's 'purpose'
is to slow things down. It's there because the other things are inherently
slower and need the accommodation). Suffice to say that, in the given
examples we're dealing with, those devices are operating at the speeds
they're spec'd to operate at.

Which wouldn't necessarily be the case for all kinds of overclocking but
we'd first need to get over the hump of the simple version before moving to
more complex varieties.


> In a way this supports Daves Cs ill-presented argument that multiplier
> overclocks dont generally give much performance gain (but we all know this
> already) as at the same clock speed, the faster CPU sits around twiddling
> its thumbs almost as much as a slower one. If you have very CPU bound
> applications then this can make a difference. (Photoshop Filters, Encoding
> video etc) but for Joe Bloggs nothing to write home about.
>

The same thing can be said for the person trying to decide between buying
an XP 3000+/333 vs the less expensive XP2500+/333, but that's a different
topic.

Amusingly enough, it could also be used as an argument 'for' (experimenting
with) overclocking just as much as against it because if you manage it then
'kewl' but if you don't then it doesn't really matter ;)

Btw, I'm not suggesting that anyone should overclock anything, and there
are definite tradeoffs/risks involved, but Dave C's 'explanations' of the
matter are simply and fundamentally incorrect.

[Ext User(David Maynard)]

Dave C. wrote:

>>>if you need a faster system, build one. -Dave
>>
>>And overclocking is one way to "build one."
>>
>
>
> Well on that much at least we can agree. We'll never agree on the correct
> way to build one, but overclocking is certainly one way. -Dave

It's very possible that, given a particular set of circumstances and a
definition of what 'correct' is in those circumstances, we might agree but
I don't agree with one-size-fits-all, broad brush generalizations based on
myth and faulty assumptions.