On 3/20/07, Lu, Yinghai yinghai.lu@amd.com wrote:
Only Opteron supports ECC.
I hope that's not true, otherwise AMD needs to change their datasheet for the Athlon 64 X2:
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/3342...
About mem training 0: BSP will train memory on all nodes of ReceiverEn, then train memory nodes of DQSPos 1: every core0 will train memory on the node itself. 2: BSP will train memory on one node (ReceiverEn, DQSPos) and then next Node.
For big system ( Four way or quad way system), you can save time with Option 1.
So for a single-socket dual-core system, which training option should I use?
--Ed