topic Re: How is Source Address Decoder programmed in a NUMA server? in Intel® Xeon® Processor and Server Products

How is Source Address Decoder programmed in a NUMA server?

MBloo4 — Thu, 01 Aug 2019 22:30:30 GMT

For things like hot-plugging of new QPI/UPI segments (e.g. populating a new CPU socket ) or of new DIMMs an attention button is used to invoke the SMM handler and reconfigure the SAD and the TAD in each node.

But if the OS remaps a PCI device (e.g. changing the value of one of its BARs) attached to a socket's local PCIe link, it can potentially use an address not routed to that socket by the current SAD configuration.

How is this case handled? How is the physical address space mapped between sockets (particularly, the part not backed by the RAM)?

Re: How is Source Address Decoder programmed in a NUMA server?

SergioS_Intel — Fri, 02 Aug 2019 07:56:37 GMT

Hello MBloo4,

Can you please provide to us the model of your Intel(R) server board?

Best regards,

Sergio S.

Intel Customer Support Technician

Under Contract to Intel Corporation

For firmware updates and troubleshooting tips, visit :https://intel.com/support/serverbios

Re: How is Source Address Decoder programmed in a NUMA server?

SergioS_Intel — Thu, 08 Aug 2019 08:30:24 GMT

Hello MBloo4,

We are following your question and would like to know if you need further assistance.

Best regards,

Sergio S.

Intel Customer Support Technician

Under Contract to Intel Corporation

For firmware updates and troubleshooting tips, visit :https://intel.com/support/serverbios

Re: How is Source Address Decoder programmed in a NUMA server?

SergioS_Intel — Tue, 13 Aug 2019 06:10:18 GMT

Hello MBloo4,

In case you need more help, please contact us back.

Best regards,

Sergio S.

Intel Customer Support Technician

Under Contract to Intel Corporation

For firmware updates and troubleshooting tips, visit :https://intel.com/support/serverbios