[LinuxBIOS] LB vs Factory BIOS -- more weirdness

[Eric Poulsen]

Nick,

When I bought that ram, I was looking for "low profile" ram so that 
everything would fit into my tiny case.  That was either the only speed 
they had or the lowest speed ram they had in low profile.  I have a lot 
of ram lying around -- I can try different ones.

One question -- why would this problem be 'fixed' by allowing the 
Factory Bios to set some registers?  Or are you referring to the DMA 
lock issues?

Nick Barker wrote:
> Eric
>
> One further observation re this locking up is that you seem to be using
> fast memory compared to the speed of the chipset. The debug output:
>
> CAS Supported 2 2.5 3
> Cycle time at CL X     (nS)50
> Cycle time at CL X-0.5 (nS)60
> Cycle time at CL X-1   (nS)75
> Starting at CAS 3
> We can do CAS 2.5
> We can do CAS 2
>
> tells me that LB has spotted memory rated as DDR400, and is consequently
> trying to use it with a 'tuned' CAS setting of only 2. DDR266 usually
> works at CAS of 2.5 or 3. It 'MAY' be the case that working the chipset /
> memory interface at this 'tuned' setting is marginal - OK under memtest,
> but as soon as you have other bus masters trying to access the memory as
> well then you get problems.
>
> You could test out this hypothesis by forcing CAS of 2.5 or 3 in the memory
> detection routines. Should be easy to do since the code starts out with 3
> and then decides if it could do better. Simply comment out those tests.
>
> Or try it out with DDR266.
>
> Nick Barker
>