On Friday 13 July 2001 06:27 pm, Dennis Myers wrote:
> On Friday 13 July 2001 19:56, you wrote:
> > On Friday 13 July 2001 11:02 am, Terry Smith wrote:
> > > A few weeks ago there was some discussion about mobos,
> > > specifically motherboard/chipset combinations that worked well or
> > > not well with linux.
> >
> > http://www.linuxhardware.org/features/01/06/06/1821202.shtml
> >
> > "The Soyo was a definite winner here with great support, awesome
> > performance, and rock-solid stability. It receives our
> > LinuxHardware.org Top Honors Award and our "Works with Linux"
> > Certification."
> >
> >    I got one (kt7vta pro), and have an oc'd Tbird 1.4 to 1.53g on
> > it (currently 11.5x135). I'd echo the above recommendation based on
> > this
>
> <snip> whatever mobo you decided on, d/l the appropriate manual from
> the
>
> > manufacturer's webpage and give it a thoro read while you're
> > waiting for UPS to deliver it ;)
>
> Tom, based on one of your previous mails about mwave and also that
> you went with the soyo, (I have soyo in two of my computers now) I
> ordered the kt7vta pro and bought a AMD 900 locally to go with it.
> Now, do you know what is safe to overclock? 10%?

   Safe and overclock are an oxymoron ;) If you've read any of my past 
USER rants (ie, problems == user -> hardware -> then lastly OS, even 
windo$), then i need to add that overclocking merges the hardware 
aspect into USER errors ;>  From what I understand tho, the 900 thru 
the 1.4 gig Tbird is all the same core. My long overclocking experience 
has shown me that it's not the core that's the limiting factor most 
always. The L2 cache is.

   When they cast cpu's, a bunch at a time, all in one big wafer, then 
they're separated. They're tested, and the weaker ones are marked, 
labled, often locked to lower speeds. Intel locks all their cpu's since 
8/98.  Almost always because the L* caches are the limit, not the core. 
Sometimes changes are made (ie, stepping, which means 'production run') 
So it's possible that a 900 Tbird, could make 1, even 1.1+gig with 
caches enabled. Actually, this is why the Duron, with it's smaller 
cache, often oc's better than a Tbird. Just put on your Clint Eastwood 
hat on and say to yourself " ..well, how lucky are you feeling today 
[self] ? " ;>>  

   Caution tho, most compile, or prime95, or even memtest86 hardware 
errors are due to overstressed caches (L1, L2) ... not the core. I've 
seen some kiddies bragging about getting their 900 Tbird to 1.5gig. 
Either they're really lucky at 2.x volts Vcore, all L7 bridges closed, 
with L2 disabled, water cooled ... or they're lie'in bastards ;>
Bet'ya in either event they couldn't handle 'cpuburn' for 2 minutes, 
and with L* caches disabled, they're no better than a 386/33 sx anyhow.

 Seriously tho, and in answer to your question, about 10% is the norm. 
You should make 1.0 gig.  My 1.4 at 1.55 is a 10% 'safe' overclock.
Only problem you may have is unlocking your L1 bridges. I believe most 
900's are locked, 1.2g up aren't. You'll probly need to explore the 
windshield defroster paint trick ;)

   Now I gotta fess up.  I truly don't believe you could see a nickels 
worth a real world difference in a system running at 900 or a bird oc'd 
to 1.5 gig like mine. I've just always overclocked, can't resist. I 
need a 12 step program ;>  Promised myself I'd boot it for the first 
time at the default 10.5x133, and at the last moment, hit <Delete> to 
enter bios, and changed it to 10.5x140 ;~>>  When I saw Linux didn't 
care, I changed the dip switches, rebooted to 12x133, 1600mhz !!  BTW, 
Linux still doesn't care, I need to keep in the 1.5 gig range to fly 
Billy's airplanes. Winders sux ;(

 and do you know if
> there is an ide controller card that will work with linux on this
> board? 

   For what? OK, the Soyo board only has the tried and true 2 IDE 
connections, with no B$ win-raid. I believe that's why the Soyo is so 
well regarded and still on AMD's recommended list. Are you lookin for 
more? I'm prob'ly even a little more adamant than Civileme about IDE, 
but I'll let him speak for himself. 

   IMNSHO, any kludge like win-raid, agp, and other pci so-called 
enhancements are nothing more than marketing tools to sell the latest 
and greatest to winbloes users.  ..... and best avoided, along with 
anything USB, another of Billy Goat's M$ hypes. Pretty NSHO, huh?  ;>>

I know from what civileme is posting that the promise won't
> work and there is another brand but I can't find the name.

    If you need more than 2 IDE ports, then  ... I dunno ;)
I've never needed or tried more than the standard two. But I believe 
you can find out pretty quickly by searching Mandrake Forum, or the 
expert mailing list.  .... or maybe Civileme will answer

   All I can say is DON't use the AGP slot with whatever controller you 
decide on. By that I mean that the first PCI slot is the AGP slot. 
Better stated would be the AGP == the first PCI slot. (AGP is nothin 
but PCI, we'd all be better off without this M$ Billy hype). That 1st 
PCI slot should never have anything in it or it'll seriously fsck up 
IRQ sharing.  Also, don't use APCI, another of M$ Billy's fsck ups, 
that messes with IRQ's, particularly 9.

 At any
> rate thanks for posting about Mwave. They shipped the next day after
> my 8:00pm order. Very prompt, I should have it on Monday. Thanks
> again for sharing the info.

   Well, i'm pretty fickle, main reason I bought from them is they had 
the absolute lowest price on the K7vta, and were close on the 1.4 Tbird 
and included a decent hs-f. That and I have used them many times, never 
disappointed, and they've never sent old stuff.  My 1.4 was barely two 
weeks out of the fab when I got it, stepping 4. If your 900 is 
similarly 'fresh', you've got the latest stepping, an over 10% oc is 
probable, even likely. 

  Just don't do it all with FSB. Raise the multiplier and keep the FSB 
near 133. Otherwise your PCI and AGP busses will get too far off spec. 
Not good with a VIA chipset. I'm seein that 2 or 3 mhz ± is the safe 
limit. So keep the FSB within 130 to 137.  I know a lot of hardware 
sites are claiming 150+mhz FSB, but i think they're pitching to the 
kiddies. Getting the PCI bus high (PCI is FSB/4), doesn't overclock it. 
It just makes it harder for the bus to talk to the cpu/cache/ram. Which 
you DON'T want, 'cause your HDD's run on the PCI bus.
-- 
   Tom Brinkman      [EMAIL PROTECTED]      Galveston Bay

Reply via email to