memory epia-m 127M != 512M

[Ian Smith]

Niki Waibel wrote:

>hi ian,
>  
>
>>The first is that I think the comments in the code may be a bit misleading 
>>(apologies to the original author!)
>>
>>AIUI the registers at 0x56 and 0x57 (and 0x5E and 0x5F as well) should be 
>>set to the same value as 0x5D i.e. the DRAM ending address for your highest 
>>populated bank of memory.
>>
>>Looking through northbridge/via/vt8623/raminit.inc it looks like 0x56 and 
>>0x56 are set to 0x10, which I think would equate to 128MB as you are seeing.
>>    
>>
>
>hmmm
>  
>
>>I think if you possibly remove the references to 0x56 and 0x57 in 
>>northbridge.c (and set the number of banks to 4), then set the values in 
>>raminit.inc to what you want i.e.
>>
>>0x5A = 0x20
>>0x5B = 0x40
>>0x5C = 0x40
>>0x5D = 0x40
>>0x56 = 0x40
>>0x57 = 0x40
>>
>>then the size detect code might just work.
>>
Rats - got this wrong (no coffee yet today :-)

The units are 16MB in these registers so 0x5A should be 0x10 and the 
other should be 0x20. Sorry for any confusion.

Blows a hole in my theory about the 128MB being detected....

>>Of course if your 512MB module actually has two banks of memory then you 
>>might need to add some extra code to initialise the second bank (not sure 
>>about this though.
>>
>>I need to get this working for the project I'm working on (as well as 
>>getting banks 2 and 3 going too),
>>    
>>
>
>all the work which has to be done (regarding this) is almost impossible
>without the knowledge of the registers.
>
>because of that i cannot help much ... i dont want to write wrong code
>just because of lack of information.
>  
>
>>but at the moment I'm trying to get the 
>>VGA working so it might be a few days before I will the time to do some 
>>proper investigation.
>>    
>>
>
>so we need to wait for you, because it seems that you know
>how to program the beast :)
>
I think I may have just disproved that theory :-)

>>Hope this helps anyway
>>    
>>
>
>as i sayed -- this is not urgent for me.
>
>it is just a bit annoying that via (and also other companies)
>are really slowing down the process of developement. i dont see
>the point in doing this. any fully working software make the
>companies, who produce hardware, more famous -> more money.
>hiding information is not a good way in my opinion.
>
It does seem odd that they are so secretive. I've worked with a lot of 
chip vendors, both bigger and smaller, and I can't recall any of them 
being so paranoid about their data sheets.

>niki
>
Cheers

Ian