[LinuxBIOS] Athlon64 939

[YhLu]

The Athlon 64 939 support dual channel.

but in the test in s2865 with d0 Athlon64 939 in LinuxBIOS, it only work
with one DIMM in socket 1.

weird....

the same code with Opteron d0 ....

YH
> -----Original Message-----
> From: YhLu 
> Sent: Wednesday, March 30, 2005 3:09 PM
> To: [e-mail]; Kevin O'Connor
> Cc: [e-mail]
> Subject: RE: [LinuxBIOS] Re: Error in compiling STPC Consumer
>
> Eric,
>
> the code in raminit.c is_opteron will treat my athlon 64 
> (939) as Opteron too.
>
> YH
>
> static int is_opteron(const struct mem_controller *ctrl) {
>         /* Test to see if I am an Opteron.  
>          * FIXME Testing dual channel capability is correct for now
>          * but a beter test is probably required.
>          */
> #warning "FIXME implement a better test for opterons"
>         uint32_t nbcap;
>         nbcap = pci_read_config32(ctrl->f3, NORTHBRIDGE_CAP);
>         return !!(nbcap & NBCAP_128Bit); }
>
> LinuxBIOS mailing list
> [e-mail]
> http://www.openbios.org/mailman/listinfo/linuxbios

[LinuxBIOS] Re: Athlon64 939

YhLu  writes:
> The Athlon 64 939 support dual channel.
>
> but in the test in s2865 with d0 Athlon64 939 in LinuxBIOS, it only work
> with one DIMM in socket 1.
>
> weird....
>
> the same code with Opteron d0 ....

I seem to recall that unbuffered DIMMS have some pretty hefty restrictions,
on what works and what doesn't.  One DIMM per channel may be correct.

If these are registered DIMMS this appears to be a deeper issue.  You might
want to try running the memory as PC2700, to see if that gets you farther.
> > -----Original Message-----
> > From: YhLu 
> > Sent: Wednesday, March 30, 2005 3:09 PM
> > To: [e-mail]; Kevin O'Connor
> > Cc: [e-mail]
> > Subject: RE: [LinuxBIOS] Re: Error in compiling STPC Consumer
> >
> > Eric,
> >
> > the code in raminit.c is_opteron will treat my athlon 64 
> > (939) as Opteron too.

Reflecting on this.  It looks like the test needs to be against
the brand-id field of cpuid.

Eric

[LinuxBIOS] Re: Athlon64 939

[yhlu]

> > > the code in raminit.c is_opteron will treat my athlon 64
> > > (939) as Opteron too.
>
> Reflecting on this.  It looks like the test needs to be against
> the brand-id field of cpuid.

even brand-id is the same. I guess We need another MARCO in auto.c....

YH