Re: Q: Processor speed/frame rate (Wintel PCs)

9 Jan 97 10:43:53 EDT


I hope you don't mind me posting your comments to the list. They
are good comments - information that folks should consider when
they are upgrading their PC.

Also, I hope that in creating/continuing this thread that people
will stop the whining Mac vs. PC debate and change their focus.

> One thing to keep in mind. Pentium processors run on one of two
> memory bus speeds-- 60mhz or 66mhz. Therefore, a 90mhz, 120mhz,
> and 150mhz all run on the 60mhz bus. Moreover, the 100mhz, 133mhz,
> 166mhz, and 200mhz all run on the 66mhz bus.

Yes, but since clock speed must change with CPU speed, it should
not compromise the validity of the test results.

> Finally, the Intel PCI chipsets, which encompass the aforementioned
> bus speeds, reach their limit at 2.5X clock multiplier. This is
> borne out when benchmarking Pentium 200-- typically little or no
> performance improvement is realized over a 166mhz machine.

Agreed. If you need performance greater than 166 MHz, a Pentium Pro
may be a better solution. An increase or change in the cache RAM may
actually give you performance equal to a 200.

Moreover, throughput can change from chipset to chipset with the same
CPU. I agree that the results of the test will not hold from
motherboard to motherboard if presented only as the number of frames
per second, however, if the results were presented as a percentage
increase in performance over the baseline of a 90 MHz Pentium CPU,
then that information would be valuable to someone when trying to
determine whether to buy a 100 MHz CPU or spring for the 150. If the
performance increase between a 100 MHz to a 150 MHz is 100% or greater
but the price difference is, say, a 75% increase, then that is a good
value (IMHO).

> However, in respect to the 486. The Intel processors run at a max.
> bus speed of 33mhz, while their AMD counterparts run at 40mhz.

While I wouldn't be so bold to say the Mac is dead, I would go out
on a limb to say that the 486 is quite close to the end of its life.
While, I was intrigued to see the previous post concerning the
difference between a 486 and a P-150, I do not consider a 486 CPU
to be an effective solution for a video-conferencing host. Therefore,
to include the 486 in such a test would not be worthwhile except to
convince those folks with 486-based PCs to upgrade to the Pentium.

Basically, what I had in mind was to use the same motherboard and,
starting with a 90 MHz CPU, proceed progressively to a 200 MHz. The
test should be performed on a closed network, say a two node, coax
Ethernet. The far-end host should be the same PC throughout the test
that would be running no other applications besides CU-SeeMe. That
way, virtually all variables that could significantly compromise the
test have been eliminated.

I'd love to perform such a test myself and publish the results to
the list, but, unfortunately, my spouse and my mortgage holder would
conclude that this is _not_ an effective use of several thousand
dollars. ;-)