Troubleshooting and Documenting Your Server s CPU

Troubleshooting and Documenting Your Server's CPU

Whether your server was built by a major vendor and arrived in your hands as a preconfigured system or you built it yourself, you need accurate documentation about the server's onboard hardware. You can obtain this documentation through the following sources:

• For preconfigured servers from major vendors such as IBM, Hewlett-Packard/Compaq, Gateway, Dell, Sun, SGI, and others, contact the vendor. Vendor websites usually offer model-specific technical support documentation.

• For white-box servers built from components, you need to determine the vendor for each component (processor, chipset, motherboard, hard disk, SCSI, network adapter, and so forth) and contact each vendor. Component vendor websites typically offer model-specific documentation.

Note

If you are not satisfied with the technical documentation available from a server vendor, you can sometimes obtain technical documentation from the server's OEM component vendors. Note that in some cases, a server vendor might use modified components that are not the same as those the component vendor sells at retail or to other companies.

Even if you received excellent documentation with your server or server components, keep in mind that updated versions might be available for download. You should check for documentation updates before attempting to determine whether a particular server or component is suitable for a given task.

For more information about server benchmarking and maintenance, see Chapter 21, "Server Testing and Maintenance."

Processor Bugs

All processors can contain design defects or errors. Processor bugs aren't new: Some of Intel's first 32-bit processors, 386s, were unable to perform 32-bit math correctly. Chips that could perform 32-bit math properly were stamped at the factory with a double-sigma marking. However, the most notorious processor bug was the 1994 FDIV bug in early Intel Pentiums. This bug caused some floating-point calculations to produce errors. Intel ended up replacing affected processors. Although this example might be the only example that the average computer user might remember of a buggy processor, in reality, all types of microprocessors can have various types of errors, including the following:

• Power management problems

• Memory compatibility issues

• Cache coherency or lockup issues on multiple-processor systems

You can often avoid the effects of a given bug by implementing hardware or software workarounds. Intel documents these bugs and workarounds well for its processors in its processor Specification Update manual, which is available from Intel's website (http://developer.intel.com). AMD provides information for its processors in its Revision Guide for each processor, available from the AMD website (www.amd.com).

Unlike the Intel 386 and Pentium bugs, bugs in recent processors usually don't result in the generation of spurious data. However, they can result in system slowdowns, system lockups, or other issues that are detrimental to reliable server performance. Because there is no trade-in policy for processorsand even if there were, you'd be trading in a processor with one set of bugs for a processor with another setthree methods have been adopted to counteract processor bugs:

• Operating system patches and updates

• BIOS updates

• Processor updates

Processor Updates

The need for processor updates became obvious in 1994, when the Intel Pentium processor was determined to have an error in its floating-point processor. At the time, the only way to solve such a significant problem was to replace the chip with one that had the bug fixed or to create software or BIOS-based workarounds. Starting with the Pentium Pro processor and continuing through current models, Intel introduced a method of fixing bugs by altering the microcode in the processor. Microcode is essentially a set of instructions and tables in the processor that control how the processor operates. These processors incorporate a feature called reprogrammable microcode, which enables certain types of bugs to be worked around via microcode updates. The microcode updates reside in the motherboard ROM BIOS and are loaded into the processor by the motherboard BIOS during the POST. Each time the system is rebooted, the fix code is reloaded, ensuring that it will have the bug fix installed any time the processor is operating.

Intel provides the updated microcode for a given processor to the motherboard manufacturer so it can incorporate the microcode into the flash ROM BIOS for the board. This is one reason it is important to install the most recent motherboard BIOS anytime you install a new processor. If your processor is newer than your motherboard ROM BIOS code, it probably doesn't include updated microcode to support your processor. In that case, you should visit the website of your motherboard manufacturer so you can download and install the latest BIOS update for your motherboard.

BIOS Updates

Because the BIOS controls all standard onboard hardware, including the processor, it has a vital role in determining how well your processors work. If your server's BIOS does not recognize a new processor you plan to install, you must update the BIOS before performing the processor upgrade. However, if you notice any other problems with the processor, including unexplained lockups or slowdowns, you should check out the server or motherboard vendor's list of BIOS fixes to determine whether any of them apply to your situation. If you find a BIOS update that applies, download and install it. If possible, test your server in a non-production setting for a while to determine whether the new BIOS update has solved the problem or whether it has introduced new problems.