939 Athlon strange beahavior

[D.8080]

Hello,
I got an AMD Athlon 64 3000+ (rev. D0) Winchester core.

When I populate slots B with a single stick of ram the system won't start, halting on C1 error (memory sizing).

If in slot A it works perfectly.
If I populate as to have dual channel it will work and memtest is flawless.

All tested on 3 different boards and same result.

Is it maybe that the memory controller on the cpu is not working properly as it is the only thing I can think of.

Even stress tests perform well, no hangs.

Curious stuff.

Thanks for any insight.

[frag_]

As I understand datasheet populating only the second channel would work only starting from revision E (Venice and later).

[D.8080]

[mikey100tv]

Mm-hm. From long experience with a Socket 939 board (an MSI MS-7184 "Amethyst" - a 'special' made under license for HP/Compaq, so MSI won't acknowledge it!), I can definitely confirm what you've said.

You try running without anything in either channel's 'primary' slot, it doesn't want to know.....at ALL. The BIOS definitely expects to find something in either a1 or b1, otherwise you throw the coding right out.....

Mike.

[ApostolCV]

Hello, at some moment of life i had the same proble with motherboard of my brother.
It was Am2/Am2+ motherbord GA-MA69VM-S2 from Gigabyte.

It had 4 DDR2 Slots. But when i try to upgrade it from 4Gb to 8 i get only 6 in work condition.
One of slots (do not remeber number). If i use 3'rd slot if count from left to right do not want to start.
Ofcourse at those moment i think that was porblem of MB.
But some time later i found that it was a bug of some steppings of Athlon X2.
And when i get more powerful CPU - Phenom x4 i get 4
fully functional DDR2 slots(8GB ).

[D.8080]

The fact is that we take for granted that the cpus that are out in the market have been thoroughly tested and, except for some very specific instructions that never are used by the average user, they are flawless.

Many cores have specific issues, bugs and whatnots that we are not aware of unless we actually read datasheets. Maybe with a chipset but not another.

Some patches can be applied through OS or maybe a new BIOS, some just need a new core revision.

Personally I never thought it could be a specific design requiring to have the first slot populated in order to boot. Another example is old boards and ram chips density.