Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: new computer athlon?

Author: Keith Conary

Date: 12:36:02 03/23/00

Go up one level in this thread


On March 23, 2000 at 14:23:22, Tom Kerrigan wrote:

>On March 23, 2000 at 11:19:58, Keith Conary wrote:
>
>>On March 23, 2000 at 08:41:04, Peter Skinner wrote:
>>
>>
>>>
>>>So thus it would still be my conclusion that the AMD is great for performance in
>>>some areas, but surely will never out perform an Intel processor when dealing
>>>with high scale rendering.
>>
>>
>>There is still the possibility that your system is flawed either by drivers or
>>compatible components used with your athlon.  The Athlon is currently in a
>>developmental stage where it currently boast off-die level 2 caching which
>>restricts the capabilites of the processor.  This is where Intel enjoys it's
>>advantage over AMD.  Because of Intel's integrated on-die level 2 caching system
>>it out performs AMD in gaming situations while the Athlon excels in business
>
>I disagree. The performance of games does not improve significantly with a lot
>of cache.
>
>The Celeron runs games just as fast as a PII, and it has 1/4 of the L2 cache.
>
>-Tom


Can you produce any studies that prove cache has little or no effect on games as
a whole?  I think the effects of cache would be totally dependent on how the
game causes memory to access it's L1 or L2 registers.  In other words, if the
program ,whether chess or other, is predictable and accesses instructions in a
serial manner then cache would greatly improve performance because there is a
higher rate of hits to the cache register.  On the other hand, if the program is
unpredictable and randomly samples the cache source for pre-fetched instructions
then the chances of a successful cache fetch will be drastically reduced causing
the cache to become a non-factor.

In general, if you take two processors that are equal in clock speed and add
on-die cache to one and off-die to the other the processor with cache on die
will outperform the other simply because it doesn't have to retireve data as far
as RAM or permanent storage to perform a data fetch.  Which in turn removes the
bottleneck of the slower secondary memory from the whole process. This also
improves the overall performance of the chip.

If cache was a non-factor with games AMD wouldn't be specifically moving to an
on-die configuration to produce better results in the gaming area.  It is clear
that AMD has a superior architecture which allows it to compete nicely and edge
out the Coppermine in many instances. However, until AMD moves the cache on-die
then the true scalablility and superiority of the Athlon won't be realized.

Keith



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.