Troubleshooting the LGA 771 to 775 mod, from no POST to unstable loads

When a modded Xeon misbehaves, the cause is usually on a short list: BIOS support and microcode, the sticker, the socket, or settings left over from the old chip. The sections below start from what you see and list the likely causes in the order worth checking, quickest and safest test first. Two habits make every one of them faster. Change one thing at a time, and keep your old LGA 775 processor nearby. It is your control sample: if the old chip still works, the board is fine and the problem sits with the Xeon, its sticker or its support.

Quick checks before anything else

These take a few minutes and rule out the simple mistakes:

• The 4-pin or 8-pin ATX12V plug near the socket is connected. It is easy to knock loose while fitting a cooler.
• The CPU fan is on the CPU fan header. Many boards stop at a warning when they see no fan signal there.
• The gold triangle on the Xeon matches the triangle on the socket frame. With the key tabs gone, a rotated chip still sits flat.
• The CMOS was cleared with the power cord out, and the jumper is back on its normal pins.
• There is exactly one sticker on the chip, sitting square, fitted as in page 4.

No POST, a blank screen or a restart loop

Fans spin, but there is no picture, or the board powers off and on in a loop.

Likely causes, in order:

1. The BIOS has no support for this class of chip: no 45 nm quads, no 1333 or 1600 FSB, or no code path for the Xeon at all. Mod guides and forum threads point here first, and for good reason.
2. Settings from the old chip are still active, such as a manual FSB, a voltage or a memory setting the Xeon can't start with.
3. The sticker is misaligned, wrinkled, partly lifted, or doubled up.
4. The chip is seated rotated, or a fragment of a key tab is trapped under it.
5. The ATX12V plug or the power supply itself.
6. A board that won't run the Xeon under any BIOS, which page 2 helps you spot.

What to try: clear the CMOS again with the cord out, then reseat the Xeon and check the triangle and the sticker under a magnifier. If it still won't start, put the old LGA 775 chip back in. If the old chip boots normally, the socket survived the mod and the question is BIOS support. Flash the latest BIOS with the old chip, check the support list (our guide to checking whether a board supports a processor shows where to find it, and page 2 explains what to look for), and add the Xeon's microcode if it needs it. If the old chip also fails, go to the last section on this page, because the socket or the contacts are now the suspect.

Owners commonly report that some boards cycle power once or twice after a CMOS clear before the first picture appears. Wait a minute before you call it a no POST. If your board has a POST code display or a speaker, the manual lists what each code or beep pattern means, and that narrows things down quickly.

An "unknown CPU", "update BIOS" or microcode message at POST

The board starts, but stops on a message saying the processor is unknown or unsupported, that its microcode was not found or failed to load, or that you should update the BIOS. The wording depends on the BIOS vendor, and many boards let you continue with F1.

Likely causes, in order:

1. The BIOS knows the class of chip but has no microcode entry for the Xeon's CPUID and platform ID. A BIOS that runs a Q9650 perfectly can still lack the entry for an E5450.
2. The BIOS is older than the first version that supports the matching desktop chip.
3. A stepping mismatch: a C0 Xeon (CPUID 10676) on a BIOS that only carries E0 code (1067A), or the reverse. Read the stepping from software or from the sSpec on the heat spreader with our guide to the CPUID, stepping and sSpec.

What to try: flash the newest BIOS from the board maker first, since that is the lowest-risk fix. If the message stays, the BIOS needs the Xeon's microcode added. The Xeon microcode guide explains what the entry is and how to tell which one you need. The procedure itself depends on the BIOS family: one guide covers AMI BIOS images, the other Award and Phoenix-Award images. Keep the existing entries for your old chip when you add one, so the old processor still works as a fallback.

Pressing F1 every time is not a real fix. The board may well boot, but the CPU then runs without the corrections Intel shipped in microcode for its known errata, and parts of its setup may be skipped.

What the operating system can and cannot do

Both major operating systems can load newer microcode on their own. Windows ships an Intel microcode library (mcupdate_GenuineIntel.dll) and applies an update at startup when it carries one for your chip. Linux can load microcode very early in boot from the initramfs, using Intel's update files. The kernel documentation describes how that works, and most distributions set it up when you install their intel-ucode or microcode package. Intel's public package still includes updates for both 45 nm Xeon CPUIDs.

What neither can do is help the BIOS. By the time Windows or Linux loads, the BIOS has already set up the processor, or failed to. An OS update won't cure a no POST, a stop at an F1 message, or a feature the BIOS left disabled. It complements BIOS support rather than replacing it.

To see which revision is running on Linux:

grep -m1 microcode /proc/cpuinfo
journalctl -k | grep -i microcode

On Windows, PowerShell can read the revision values Windows records for the first core (they show as byte arrays):

Get-ItemProperty 'HKLM:\HARDWARE\DESCRIPTION\System\CentralProcessor\0' | Select-Object 'Update Revision', 'Previous Update Revision'

Wrong clock speed, missing features or only one core

Likely causes, in order:

1. Nothing is wrong. EIST lowers the multiplier at idle, so a 3.00 GHz chip reading lower on the desktop is normal. Check again under load.
2. The bus is running at the wrong speed. If the BIOS falls back to a 266 MHz bus, an E5450 shows 266 x 9 = 2.40 GHz instead of 333 x 9 = 3.00 GHz.
3. The BIOS does not fully recognise the chip. Owners on the Bios-mods forum describe a Xeon in a BIOS without its microcode entry showing no SpeedStep, SSE4.1 or VT-x in CPU-Z, and sitting at full clock all the time. That is owner experience rather than documented behaviour, but it points at the microcode fix above.
4. The chip has no EIST to begin with. The E5405, L5238 and L5240 lack it, so a fixed clock is normal on those.
5. Only one or two cores show up. Look for a BIOS option that limits active cores, and check that Max CPUID Value Limit (where your board has it) is disabled, since it exists for very old operating systems. In Windows, open msconfig, then Boot, then Advanced options, and make sure the processor count limit is not ticked.

What to try: load defaults, then set the FSB manually to the chip's rated value if the board picked a lower one. Confirm the result under load with CPU-Z on Windows or lscpu on Linux. If features are missing, sort out the microcode first, and only then look at individual BIOS options.

Windows crashes, will not install, or Linux logs machine checks

Likely causes, in order:

1. Missing or mismatched microcode, which leaves errata unfixed.
2. Leftover settings, or memory running differently now that the bus speed has changed.
3. A sticker with marginal contact, which shows up as random faults rather than a clean failure.
4. Power delivery struggling under load, mostly with 120 W chips on budget boards.
5. The operating system itself. Microsoft's supported processor list for Windows 11 does not include any Core 2 era chip, so Windows 10 or older, or a current Linux distribution, is the realistic target.

A stop code of WHEA_UNCORRECTABLE_ERROR (0x124) means Windows recorded a fatal hardware error, and with a freshly modded chip the processor is the first suspect. On Linux, look for kernel log lines that mention mce or Hardware Error. Both point at the chip, its power or its microcode rather than a driver.

What to try: load defaults, sort out the microcode, then run a memory tester such as MemTest86+ for at least one full pass. If it is still unstable, switch off C1E and EIST in the BIOS as a test. If that cures it, the BIOS is not handling the chip's power states properly, and microcode or a newer BIOS is the proper fix. If none of that helps, take the chip out and refit or replace the sticker.

Temperatures that look too high, too low or impossible

Likely causes, in order:

1. The monitoring software is assuming the wrong TjMax. Core temperatures on these chips are read as a distance below TjMax, the maximum junction temperature, so the tool has to assume a TjMax value. Two tools can show different numbers for the same core.
2. A BIOS that doesn't recognise the chip may read or report its sensor wrongly.
3. Real heat: a cooler that isn't seated properly, old or too little paste, or a 120 W chip on a cooler sized for something smaller.

What to try: compare two monitoring tools and watch the change from idle to load rather than the absolute number. A sensible reading rises steadily under load and falls back at idle. Then check the cooler physically. If one push pin hasn't locked, the cooler can tilt and sit badly on the heat spreader. Our repasting walkthrough covers cleaning, paste and mounting. If the readings were only odd and the heatsink behaves normally, you can live with the numbers.

The CPU fan stays at full speed

Likely causes, in order:

1. The CMOS clear reset the board's fan control, and automatic fan control is now off. Turn it back on in the hardware monitor section of the BIOS.
2. The fan header mode is wrong for the fan. A 3-pin fan on a header set to PWM, or the reverse, may run flat out.
3. The BIOS doesn't recognise the chip or can't read its sensor, and runs the fan at full speed to be safe. Owners commonly report this on boards without proper support, and a BIOS update or the microcode entry usually settles it.
4. The chip really is hot. See the previous section.

A CPU wattage or power warning at POST

This one mostly appears on OEM desktops. Some OEM firmware checks the installed processor against the power the system was designed for, and warns or refuses to continue when the chip is outside it.

Likely causes, in order:

1. The Xeon's TDP is above what the system was built for, typically a 120 W X-series chip in a machine designed around 65 W or 95 W processors.
2. The BIOS doesn't recognise the chip and can't place it in a power class.

What to try: install the latest BIOS for that system, then step down in power class. An 80 W E-series or a 50 W L-series Xeon is the usual answer. Don't look for ways around the check. The power circuits and cooling in those cases were sized for the lower figure. Page 2 has more on OEM boards.

Unstable under load

The machine boots and idles fine, but crashes, freezes or restarts under a stress test, a game or a long encode.

Likely causes, in order:

1. An overclock or a manual FSB left in place. Return everything to stock before testing anything else. Our Xeon overclocking guide covers stable settings once the stock setup is solid.
2. Missing microcode, as above.
3. Power-saving states (C1E and EIST) clashing with a BIOS that doesn't fully know the chip. Test once with both switched off.
4. Voltage regulator heat on a budget board running a 120 W chip. A small fan aimed at the VRM heatsink near the socket is a quick test. A lower-TDP Xeon is the lasting fix.
5. Core voltage set wrongly by the board's automatic setting.
6. Marginal sticker contact.

Treat voltage as the last thing you touch, not the first. Each chip asks the board for its own voltage, its VID, and Intel's product specifications put that request between 0.850 V and 1.350 V for every model in the 5200 and 5400 series. If you set the core voltage by hand, start from the BIOS default and move in the smallest step the board offers. Retest after each change, and don't go beyond the top of that window. We don't give a "safe" figure for these chips, because nothing we can cite supports one, and board voltage readings and droop vary from model to model.

Your old LGA 775 CPU no longer seats or works

The socket should still accept desktop chips after the mod, because the tabs only ever controlled orientation. If the old processor now fails, the likely causes, in order, are:

1. Orientation. Without the tabs, the old chip can sit flat when it is turned the wrong way too. Match the triangles.
2. Leftover settings for the Xeon, such as a manual FSB. Clear the CMOS with the cord out.
3. A fragment of a key tab, or a lifted sticker, left in the socket. Check under the magnifier with a low light and lift anything loose out with tweezers.
4. A contact bent during the cut. Look for one contact out of line with its neighbours. Experienced modders straighten LGA contacts with a fine needle under strong magnification, but it is a delicate job, and pushing on when you are unsure usually makes it worse.
5. A modified BIOS that dropped the old chip's microcode entry to make room for the Xeon's. Flash your backup or the maker's original file.

If a reflash goes wrong and the board won't start at all, check whether it has a recovery path such as Gigabyte's DualBIOS or ASUS CrashFree BIOS. Without one, the BIOS chip can be rewritten from outside with a programmer, as described in our guide to reading a BIOS chip with a CH341A.

If you end up asking for help on a forum such as VOGONS or Win-Raid, start the post with three facts: the Xeon's sSpec, the exact BIOS version, and the exact wording of any POST message or code. Those answer most of the first questions anyone will ask.