AMD Dual-Core Opteron Patch for NPM

jpetrino

(I've been meaning to post this for a while, but I'm just getting around to it...)

Back in January I upgraded Orion NPM 7.8.6 to 8.0.1 (both SLX), and at the same time upgraded to a new server. Previous box was an HP DL360 G3 with two single-core Xeon's (3.06 GHz/533 FSB/1MB L2 with 4GB DDR), and new box is a DL365 with two dual-core 2218 Opteron's (2.6 GHz/1MB L2 with 4GB DDR2). Note that new AMD’s are being run as 32-bit mode, with Windows Standard Edition 2003 R2.

I had several problems, but the most prevalent were response times that were all over the place (bouncing from single-digit mSec, to hundreds of mSec). Also, the Top Ten page showed various devices as unresponsive, and every time the page refreshed the list was different. Plus lots of false interface Packet Loss errors (in other words, lots of red!).

So I opened a command prompt window on the new AMD-based server and starting doing some pings. Imagine my surprise when I started getting NEGATIVE response times!! It was as if the devices were replying before they were even asked.

I tried various BIOS updates and NIC driver updates, disabling the Teamed NIC function and using a single NIC interface, and installing a separate (PCI-Express) NIC instead of the on-board 10/100/1000 NIC’s. It was finally escalated to HP’s level 3 Engineers, who determined it was not a NIC issue. All of this took almost a week, during which time my history collection was extremely flawed to say the least.

To make a long story short, it’s a known bug with the AMD dual-core Opteron CPU’s. When used as a normal app server, it would probably not be noticed -- except perhaps as some moderate level of interface errors at the Cisco switch end. But when the box is used for SNMP collection it’s a problem, because Orion doesn’t know how to interpret the negative response times.

HP has not released their “official” packaged version yet, but here is AMD’s link to the patch [AMD Opteron Processor with AMD PowerNow! Technology Driver for Windows XP and Windows Server 2003 Version (x86 and x64 exe) 1.3.2.16]:

www.amd.com/.../size=1]

rralicki

Thanks for your input we found the fix for this on HP's web site for the following system information:

System Manufacturer HP
System Model ProLiant DL585 G2
System Type x64-based PC
Processor AMD64 Family 15 Model 65 Stepping 3 AuthenticAMD ~2612 Mhz
Processor AMD64 Family 15 Model 65 Stepping 3 AuthenticAMD ~2612 Mhz
Processor AMD64 Family 15 Model 65 Stepping 3 AuthenticAMD ~2612 Mhz
Processor AMD64 Family 15 Model 65 Stepping 3 AuthenticAMD ~2612 Mhz
BIOS Version/Date HP A07, 3/28/2008

AMD Processor Driver Version 1.3.2.0053 for Windows XP and Windows Server 2003 (x86 and x64)

September 2007