On Thursday 16 May 2002 01:27 am, James wrote:
> The question I'd have on this thread is: For none high end
> applications is it worth it to be on the edge? 1.0 of anything is
> usually not good. I also think about things like nics (I've fried a
> few) etc. Whereas a Linksys or Netgear can be bought for 15 bucks
> in the US it scares me to think that I could lose a mobo cause my
> nic went down. For personal use and for High reliability, I really
> like to stay one step behind the curve. Wait till the bug fix
> version comes out. (very important if you are buying closed source
> software!) Am I alone in this attitude?
I'm with you. It's not just motherboard revisions, bios updates
either. Chipsets and cpu steppings (production runs) get fixes
usually within the first several months. A good example is the VIA
kt133a. The ide bug surfaced during the first two steppings. Later
it was discovered that many motherboard designs exacerbated the
problem. By step 4 of the kt133a the problem had all but disappeared
(as long as you didn't use an Abit kt7* ;> Later it became apparent
some high bandwith pci cards (specially the SB Live!) were largely
part of the problem. So if you jumped on bandwagon when the KT7
first appeared, installed an SB Live!, you were pretty much SOL,
'cept for bios upgrades, and upping the Vcore to 1.82v, and dropping
the HDD mode to <= udma2. Dirty work arounds. Much the same
scenario with cpu steppings.
I also think staying away from the cuttin edge, first releases of
boards and chipsets is very important for open source too. EG, the
people who rushed out to by ATI 8500's and still don't have support
for them yet. There's also work arounds for chipset/cpu errata that
take a while. EG, I have a kt133a board, shortly after I built the
system, months after kt133a boards were out, dmesg began to include
the line "Applying VIA southbridge workaround." Tho since my board
(Soyo, AMD appr'vd) is the second (and final) revision, it's not an
Abit, I don't have an SB Live, and both the chipset and the Tbird are
4th stepping, I doubt I need that kernel parameter ;)
Then there's also performance regardless of OS. It's reported
that DDR333 has negligible improvement over 266. In some tests it's
actually slower. I attribute this to 266 havin matured, while 333 is
still brand new and wet behind the ears. Not neccesarilly the ram's
fault, but the motherboard's implementation of it. Y'allsMMV ;)
--
Tom Brinkman Corpus Christi, Texas
Want to buy your Pack or Services from MandrakeSoft?
Go to http://www.mandrakestore.com
