Very Computer

So I've been hearing that the 300 is actually faster?  Why / how is
this?

Thanks,

DT
--
Technical Director, *ia Center for Computer Music
Programmer / Analyst, Dean's Office (School of A&S)

in some cases it probably is, the P-II 333 has slower L2 cache than the
300Mhz.

-= Jacob =-

Quote:>So I've been hearing that the 300 is actually faster?  Why / how is
>this?

The classic case was with Intel P133 and P150s.  The P133s were able to
run on 33MHz boards nicely, having to wait exactly 4 cycles for every
memory access.  In contrast, 33 does not nicely divide into 150; they
had to move the motherboard speed down to 25MHz in order to get an
integer number of cycles so that memory accesses could be synchronized
with clock cycles.  By slowing down the motherboard, this slowed the
system down, despite the apparent increase of MHz.

In the case of the 300/333, the way the 300 would "win" is if you have a
75MHz motherboard.  That nicely divides into 300, so that there's no
need to slow down the main bus to sync with the CPU.  But 75 does not
divide into 333; you probably have to slow the mobo to 66MHz (or some
such number) in order to regain synchronicity.

What is almost certainly more important to the average user is that the
extra 33 MHz is going to cost substantially more money, while likely
only resulting in a ~10% (or, likely, less) increase in system
performance.

I'd buy the 300 over the 333 at this point in time.  Although what I'd
rather do is buy a 233, and spend the difference in money on other
system components such as additional RAM or faster I/O (e.g. - SCSI).

It's unlikely that the average user will be using the CPU anywhere near
100% of the time, which means that spending more on the CPU is a pure
and simple waste of money...

--
"If the future navigation system [for interactive networked services on
the NII] looks like something from Microsoft, it will never work."
(Chairman of Walt Disney Television & Telecommunications)

Quote:> It's unlikely that the average user will be using the CPU anywhere near
> 100% of the time, which means that spending more on the CPU is a pure
> and simple waste of money...

Hence the constant popularity of Intel and Micro$oft.

--
To reply via mail, remove the obvious from the email address.

|>So I've been hearing that the 300 is actually faster?  Why / how is
|>this?
|
|This is possible if it results in running into fewer memory wait
|states.
|
|In the case of the 300/333, the way the 300 would "win" is if you have a
|75MHz motherboard.  That nicely divides into 300, so that there's no
|need to slow down the main bus to sync with the CPU.  But 75 does not
|divide into 333; you probably have to slow the mobo to 66MHz (or some
|such number) in order to regain synchronicity.

The "solution" is simple - Run the 333 at 75MHz speed.

That's what I'm doing - Running a pair of 333 at 75X5 = 375 on P6DLS.  It's
*very* fast.

Too bad I can't get it to run stable at 83mhz :P

Michiel Denie !

writes

To the extent to which your applications are bound by the rate of memory
accesses, this slows you down.

No doubt your PII-333 is faster than my Cyrix 6x86 150+; it *may* be faster
than a PII-300, but it's not obvious that this is truly the case.  Wait
states that are required by a mismatch between CPU speed and motherboard
speed may slow you down despite having a faster CPU.

--
Those who do not understand Unix are condemned to reinvent it, poorly.  
-- Henry Spencer          <http://www.hex.net/~cbbrowne/lsf.html>

Quote:>So I've been hearing that the 300 is actually faster?  Why / how is
>this?

The real value Deschutes brings is in reduced power consumption due to
.25 rather than .35 fabrication - it uses half the power.  As a
design, that means Deschutes is likely to see 500MHz and higher
whereas the older .35 design would probably burn a hole through the
board by then... and Deschutes is aimed at 100MHz base, FWIW.

-
TJI

|>The "solution" is simple - Run the 333 at 75MHz speed.
|>
|>That's what I'm doing - Running a pair of 333 at 75X5 = 375 on P6DLS.
It's
|>*very* fast.
|
|Unfortunately, that means that the bus speed (75) does not evenly divide
|into processor speed (333), which means needing extra wait states.
|
|To the extent to which your applications are bound by the rate of memory
|accesses, this slows you down.
|
|No doubt your PII-333 is faster than my Cyrix 6x86 150+; it *may* be faster
|than a PII-300, but it's not obvious that this is truly the case.  Wait
|states that are required by a mismatch between CPU speed and motherboard
|speed may slow you down despite having a faster CPU.

Huh?  Am I missing something?

If you run the chip at 333Mhz, it divides into 66 (5.5).

If you run the chip at 375MHz, it divides into 75 (5.5).

There is no additional wait state generated.

And of course it is faster.  I was using a 300Mhz P2 before I upgraded.

5 would be a feasible multiple, as would be 6.  5.5 is obviously not an
integral value.

Quote:>If you run the chip at 375MHz, it divides into 75 (5.5).

And that's obviously a higher speed than the chip is specified to run.
My assumption was that you were going to try to run the chip at its
specified speed.  Otherwise, the principle of "ceteris paribus"
obviously doesn't apply; it would be equally fair to try running the
P300 at 375MHz...

--
"If you want to travel around the world and be invited to speak at a
lot of different places, just write a Unix operating system." (By Linus
Torvalds)