On Sat, 1 Jul 2000, Civileme wrote:
[SNIP]
> The KTX133 is something Intel has no equivalent for and is a
A correction here: there is the 'Classic' KX133 VIA chipset & the new
(which is referred to throughout this mail) KT133 VIA chipset for the
' "optimal" processor feature '. This 'feature' means the cache runs at
full bus speed verses the 'Classic', which ran at a ratio (2/5)..
> hefty performer. No data available on long-term reliability. No
> chipset before this had the "optimal" processor feature which
> will drive the bus and the processor as fast as it can without
> getting data corruption. Also, the interesting interlace done to
> run PC100 memory on a 200MHz bus puts this one head and shoulders
The bus is running at 100Mhz, which matches the memory; it's the Front
Side Bus (FSB) which runs at the 200Mhz.
> above the crowd. It seems good enough that ASUS is using it.
To my knowledge, ASUS is using the 'Classic' KX133 VIA chipset, and no
board using the new KT133 VIA chipset is on their web page for sale or
is mentioned as being in production.
> Of course Chipsets are also married to processors or at least
> processor families.
>
> We have in the mainstream right now the following families,
> separated by their connections
>
> Socket 7 Pentium, MMX, IDT C6-x, Cyrix(VIA)
> AMD K5, K6, rise mP6
>
> Socket 370 Celeron, P-III Coppermine
>
> Socket A Duron, Athlon Thunderbird
>
> Slot-1 Older Celeron, Pentium II, III
>
> Slot-2 Xeon II III
>
> Slot-A Athlon
'Classic' would be a good addition to the Athlon line above.
> There are of course other chipset manufacturers. ALi Aladdin
> chipsets are used in many boards. Most of them have had BIOS
> updates to accurately detect some IDE devices, but otherwise they
> seem to do well at standard clock rates. SiS does some highly
> integrated stuff, and as long as you don't get it served to you
> on a very cheap board (like PCChips) those chipsets seem to be
> the most reliable around (at standard clock rates). I have a
> Shuttle HOT-599 placed in service February, 1999. It is running
> a refrigerated K6-2 450 at standard 100MHz FSB and is a heavily
> loaded workgroup server. Since. It was rebooted once when the
> 56K modem was changed to a cheap ethernet card for a DSL, and
> twice more when the server software was shifted to mdk 6.1 and to
> 7.1.
>
> If you want the fastest available without getting into 64-bit
> processors, the VIA KTX133 and Athlon Thunderbird can race past
Again, this should be KT133 VIA chipset. The only motherboard using this
new chipset (and new socket, verses the old Athlon Slot) is by FiC, and
it's the AZ-11; an updated SD-11 with the new chipset.
> all else. You can beat Pentium III Coppermine 1GHz with only a
> 900MHz Thunderbird and you CAN buy THunderbirds up to 1GHz this
> week. (who knows what it will be next week)
>
>
> http://sysopt.earthweb.com/userreviews/products/mboards.html
>
> Has ratings by users for many motherboards. User ratings as
> opposed to technical give an index of not only how wonderful the
> features are but also how much hassle it is to reap those
> benefits. Most of the raters are technically competent to some
> degree because in order to rate a MB you are probably building
> your own computer. Extreme technical competemce is not required
> but you can figure that most of these users will not draw to an
> inside strait unless nobody's betting.
>
> YEs, it depends. The chipset resides on a motherboard and it
> works with a processor. Find a GOOD (read reliable) motherboard
> (or several good ones) then look for the chipset you think works
> the best.
>
> And for the speed issue--if the base system
> (mobo+memory+processor) works at an index of 3677 (where the
> original IBM-PC 8088 at 4.77MHz and 0.2 Bogomips is the index of
> 1) and you still have to load a 45Mb program like StarOffice
> Bloatware from disk, are you going to see any improvement over a
> base system with an index of 1200? Not likely.
>
> If you are rendering 3D graphics with a specialized program and
> DDR Geforce interfaces and all--would you be better served with a
> single hot processor on a base system or a cluster of processors
> in a parallel-processing system? Both can come to about the same
> cost, or the supercomputer can be cheaper (as Oak Ridge National
> Laboratory demonstrated with Stonesoup, a Beowulf made from
> donated computers).
>
> Well, I hope this is helpful in answering what appears to be a
> simple question. You will, of course, find a better answer for
> your purposes by researching the matter yourself, but perhaps
> this can serve as a way to structure how you go about it.
>
> Civileme
>
--
| d a v i d @ m i h m
| dmihm @ swbell.net
| webmaster @ afterstep.org
| ftpmaster @ afterstep.org
| dmihm @ rchitecture.com
| www.rchitecture.com
--
