Skip to the content

LGA 771 to 775 mod motherboard compatibility: chipsets, BIOS and power

Last updated: By Marek Halloran12 min readPage 2 of 5

The sticker and the cut tabs get a Xeon onto LGA 775 contacts. The motherboard decides what happens next. A board can only run a modded Xeon if three things line up: the chipset can drive the right front side bus speed with a 45 nm processor, the BIOS already supports the matching desktop chip and can be given the Xeon's microcode, and the voltage regulator was built for the chip's power draw.

We list chipsets rather than boards because the chipset is what Intel documented. A board is one maker's implementation, with its own BIOS and power circuitry: the chipset tells you what is possible, the board's CPU support list tells you what that maker delivered. Page 1 explains the Xeon families and the sticker.

Intel chipsets of the LGA 775 era

The FSB column is Intel's rating in MT/s (million transfers per second; 1333 MT/s is a 333 MHz bus clock moving four transfers per tick). "45 nm" means the Penryn generation: Core 2 Duo E8000 and E7000, Core 2 Quad Q8000 and Q9000, and these Xeons.

Chipset FSB per Intel 45 nm and quad-core support What it means for the mod Basis
P45, P43 800/1066/1333 Built for it: Intel lists QX9000, Q9000 and E8000/E7000 Best-case host for E5400, X5400 and L5400 chips at 1333. A 1600 bus on P45 is an unofficial overclock Intel 4 Series datasheet; Wikipedia for the 1600 note
G45, G43, G41 800/1066/1333 Same as P45 Same as P45 on paper; check the TDP ceiling on the board's list Intel 4 Series datasheet
Q45, Q43, B43 800/1066/1333 Same as P45 Chipset fine; most boards are OEM business desktops (see below). The best-known mod thread warns that Q45 boards often do not work Intel 4 Series datasheet; community
P35 800/1066/1333 Intel's brief names 45 nm dual and quad-core support Good host; most boards needed a BIOS update for proper 45 nm recognition Intel P35 brief; AnandTech 2007
G33, G35 800/1066/1333 G35 brief names 45 nm; G33 shares P35's datasheet Same as P35 Intel G35 brief and 3 Series datasheet; AnandTech 2007
Q35, Q33 800/1066/1333 Core 2 Duo and Quad; Intel listed its Q35 boards for 45 nm in late 2007, BIOS pending Boards mostly OEM; the mod thread flags Q35 too Intel 3 Series datasheet; AnandTech 2007; community
G31 800/1066/1333 Intel's brief supports 45 nm but rates 1333 for Core 2 Duo only, quads up to 1066 Board makers validated 1333 quads anyway (ASRock's G31M-S lists the Q9650). Where list and brief disagree, the list is the better guide Intel G31 brief; ASRock CPU list
X38 800/1066/1333 Native 45 nm support Fine on paper; the mod thread flags X38 as risky Wikipedia; AnandTech 2007; community
X48 800/1066/1333/1600 Core 2 Duo, Quad and the QX9770 at 1600 The only Intel chipset here rated for 1600 parts (E5472, X5472, X5492). Same community warning Intel X48 datasheet; community
P965, G965, Q965 533/800/1066 Core 2 Duo and Pentium 4/D; no official 45 nm support A long shot: only some boards got 45 nm BIOS support (MSI for some P965 boards). A 1333 Xeon runs the bus above rating; the 1066 L5408 and E5205 do not Intel 965 datasheet; AnandTech 2007
946PL, 946GZ 533/800 Core 2 Duo; no Core 2 Quad Poor host: no 1066 or 1333 bus Wikipedia
945P, 945G 533/800/1066 Core 2 Duo on later board revisions; no Core 2 Quad Quad-core Xeons are out Wikipedia
945PL, 945GC, 945GZ 533/800 (late 945GC parts 1066) As above; some 945GC boards got 45 nm dual-core BIOS support A 1066 Wolfdale-DP such as the E5205 is the only realistic try Wikipedia; AnandTech 2007
Parts links can earn us a commission. As an Amazon Associate I earn from qualifying purchases.

Intel sources in the Basis column: the 4 Series datasheet, the 3 Series datasheet, the briefs for the P35, G35 and G31, the X48 datasheet and the 965 datasheet. The 945, 946 and X38 rows lean on Wikipedia's list of Intel chipsets. "AnandTech 2007" is its Wolfdale and Yorkfield board compatibility roundup, written as the 45 nm chips launched; it noted that P965 never officially supported them. "Community" means owner experience, not documentation: the widely read Overclock.net mod thread warns that X38, X48, Q35 and Q45 boards have a high chance of not working, without saying why. Treat that as a reason to check harder, not a verdict.

NVIDIA nForce chipsets

Chipset FSB What is documented about 45 nm For the mod Basis
nForce 790i Ultra SLI, 790i SLI 1600 45 nm Penryn support DDR3 boards with a native 1600 bus, so 1600 Xeons are in spec Wikipedia, nForce 700
nForce 780i SLI, 750i SLI 1333 45 nm Penryn support Good candidates; the mod thread quotes owners running an E5440 and an X5460 on 780i boards Wikipedia, nForce 700; community
nForce 680i SLI, 680i LT SLI 1333 Official chipset support for Wolfdale and Yorkfield, but board designs varied; several early 680i boards had needed a revision even for 65 nm quads Board revision and BIOS decide it AnandTech 2007; Wikipedia, nForce 600
nForce 650i SLI, 650i Ultra Not stated in our sources 45 nm support was left to each board maker The board's support list or nothing AnandTech 2007

The 700-series entries come from Wikipedia's nForce 700 article, the 680i's 1333 rating from its nForce 600 article. VIA and SiS had no chipset that officially worked with 45 nm processors, per AnandTech in late 2007, so we leave them out.

DP and UP Xeons: the 5000 and 3000 numbers

The 5000-series Xeons are DP (dual-processor) parts: two can share one server board. This guide is about two of those families on LGA 771, the 5400 (Harpertown quad-core) and 5200 (Wolfdale-DP dual-core). In a desktop board you run one, and the second-socket ability goes unused.

The 3000-series are UP (uniprocessor) Xeons, and most need no mod at all: the X3300 quads such as the X3360, the E3100 duals and the 65 nm X3200 quads are LGA 775 chips that drop straight in like their Core 2 twins. The exceptions are OEM-only "CL" parts for single-socket blade servers. The X3323, X3353 and X3363 quads, the E3113 and the L3014 are LGA 771 packages and need the full mod; a trailing 3 in the model number is the usual tell.

Why would one board take an E5450 and refuse an X5470 or an E5472? Four mismatches cover nearly every case:

  1. Core count. A board with only 45 nm dual-core support, like some 945GC boards, can host a 5200 but not a 5400.
  2. Bus speed. The 1600 MT/s parts (E5462, E5472, X5472, X5482, X5492) need an X48 or nForce 790i board, or a P45 run beyond spec. The 1066 parts (L5408, E5205, L5215) are the only ones that stay inside a 965, 945P or 945G rating.
  3. Stepping. C0 and E0 Xeons have different CPUID signatures, and the BIOS needs microcode for yours.
  4. Power. The X-series quads are rated at 120 W and up, more than many desktop boards were validated for.

The BIOS rule: the desktop twin comes first

A modded Xeon runs by passing for a chip the BIOS already knows, so the board must already support the LGA 775 processor that shares the Xeon's core, bus speed and stepping. For an E0-stepping E5450 that means a 1333 MT/s, 12 MB Yorkfield of the Q9650 class. If the board's list stops at 65 nm Core 2 Quads, no amount of cutting and sticking will fix it.

The stepping shows in the CPUID signature, a number the processor reports to software, and you can decode it by hand. Take 1067A, the E0 signature: the last hex digit is the stepping (A, which is 10), the next is the model (7), then the family (6), then a 0 for the processor type, and the leading 1 is the extended model. The full model number is 1 x 16 + 7 = 23, the Penryn generation. The C0 stepping reads 10676: same family and model, stepping 6. Our guide to finding the CPUID, stepping and sSpec shows how to read it in software or from the sSpec code printed on the heat spreader.

Xeon you plan to use Bus The board's support list should show
E5450 E0 (sSpec SLBBM, CPUID 1067A) 1333 Q9650, which only came as E0, or another E0 Yorkfield such as a Q9550 E0
E5450 C0 (SLANQ, 10676) 1333 A 1333 quad plus a C0 chip with CPUID 10676, such as an E8400 C0 or a QX9650 C0
L5420, E5440 or other E0 E5400 and L5400 parts 1333 An E0 Yorkfield at 1333
X5470, X5460 E0 1333 An E0 Yorkfield at 1333, ideally with the 130 W QX9650 listed too (see power, below)
E5472, X5472, X5482, X5492 1600 A 1600 MT/s chip such as the QX9770
X5260 (Wolfdale-DP) 1333 An E8000-series Core 2 Duo at 1333 in the same stepping; the E8400 and E8500 came as both C0 and E0

The stepping, CPUID and sSpec pairs above come from Intel's Xeon 5400 specification update. Page 3 turns this into a full decision table.

Support for the twin is necessary but not sufficient, because of microcode. Intel's support note on microcode errors after a Xeon upgrade puts it plainly: "Each unique processor stepping/package combination has an associated microcode update" (Intel support). An LGA 771 Xeon shares its stepping with the desktop twin but not its package, so it reports a different platform ID, and a BIOS that runs a Q9650 perfectly may still hold no update for the Xeon. Owners report that many boards boot anyway and show the chip as an unknown CPU; others stop with a microcode error or never finish POST. The LGA 771 Xeon microcode page covers platform IDs and how to add the missing update, and page 5 sorts out the symptoms.

Power: 80 W, 120 W and what the board was built for

TDP, thermal design power, is the heat in watts that Intel tells board and cooler designers to plan for. Desktop boards were validated against the desktop chips of their day, and Intel's Core 2 Quad Q9000 datasheet shows the tiers. The 95 W Q8000 and Q9000 chips belong to one voltage regulator configuration (775_VR_CONFIG_05A, with a current limit of 100 A), the 130 W QX9650 to a heavier one (775_VR_CONFIG_05B, 125 A), and the 136 W QX9770 is rated for 140 A (Intel Q9000 datasheet).

Against those tiers, the 80 W E5400 quads and 40 to 50 W L5400 parts sit comfortably inside what a Q9650 board was built for. The X5450, X5460, X5470 and X5472 are rated at 120 W, and the X5482 and X5492 at up to 150 W. A 120 W chip on a board validated only for 95 W parts asks for 120 / 95 = 1.26 times the planned heat, about 26 percent more, through parts sized for less. It may well run, but outside anything the board maker tested.

The same datasheet describes a Power Segment Identifier (PSID) that lets the BIOS spot a processor needing more power than the board's regulator can supply; its example is a 130 W chip on a 65 W or 95 W regulator. If a board shows a power or wattage warning at boot, this mismatch is a likely suspect, and page 5 has the checklist.

The practical rule: look for the highest-TDP chip on the board's support list. If a QX9650 or QX9770 is there, the maker validated the board for 130 W class parts and an X5470 is a reasonable fit. If the list tops out at 95 W quads, choose an E5400 or L5400 chip. Slim machines can be lower still: Intel's 65 W S-series quads, aimed largely at small-form-factor systems, have their own regulator configuration (775_VR_CONFIG_06). The vendor's specification sheet for your exact chassis settles it.

OEM boards: Dell, HP, Lenovo and Intel's own

Intel aimed the Q35 and Q45 at business desktops with vPro remote management, so they turn up mostly in Dell, HP and Lenovo office machines, alongside plenty of G-series boards. These are the least predictable hosts for the mod, and the firmware is why. An OEM BIOS is written around the processors the vendor sold in that model, updates come only from the vendor, and the images usually arrive wrapped in the vendor's own update programs, which makes adding microcode harder than with a retail AMI or Award BIOS. Owner reports are mixed; the Overclock.net thread includes a Lenovo small-form-factor desktop that would not post with an E5440 that ran in another board.

Find the vendor's specification sheet or technical guide for your exact model and chassis. If it lists no 45 nm Core 2 Quad, treat the machine as a poor candidate. Note its processor wattage limit too, because a slim chassis built for 65 W chips is a bad home for a 120 W Xeon even if the BIOS cooperates.

Intel's own desktop boards (the DG31, DG41, DP35, DP45, DQ35 and DQ45 families, among others) sit in between. They had published processor lists like retail boards, and AnandTech's 2007 roundup shows Intel issuing 45 nm BIOS updates for its P35, G33 and X38 boards. Intel shipped those updates through its own tools, though, so plan around the microcode in Intel's final release.

How to read a CPU support list

Retail board makers published CPU support lists for their LGA 775 boards, and a list is the most useful document for this mod. Our guide to checking whether your motherboard supports a processor covers finding the exact model and revision. The format is usually close to ASRock's for its G31M-S: socket, family, model number with the stepping in brackets, core name, frequency, FSB, cache and the first BIOS version that supports the chip. That G31 board lists "Q9650(E0)" at 1333 MHz, which is exactly the twin an E0 E5450 needs.

Work through it in this order:

  1. Find the twin: the desktop chip with the same core, bus and stepping (see the BIOS table above).
  2. Note the BIOS version in the "since" column, and confirm the board runs that version or newer.
  3. Find the most power-hungry chip listed. A QX9650 or QX9770 means 130 W class power delivery; nothing above Q9000 means plan on 95 W.
  4. Check the board revision. Some makers list chips per PCB revision, and an early revision of the same model can lack 45 nm support.
  5. Look for Xeons. LGA 775 Xeons such as the X3360 on the list mean the BIOS expects Xeon brand strings, though not LGA 771 microcode.

Say you have a G41 board whose list shows the Q9650 (E0) from BIOS version 1.20, nothing above 95 W, and you run version 1.40. An E0 E5450 or L5420 is a strong match, and the remaining question is microcode. An X5470 would be a 120 W chip on a 95 W list: a maybe, with a better cooler and a watchful eye. The LGA 771 compatibility checker runs the same logic from your chipset and chip.

Before you move on to page 3, write three lines on a sticky note: your chipset, the best desktop twin on your board's list with its stepping, and the highest TDP the list includes. Those three lines pick your Xeon.

Headshot of Marek Halloran
Marek Halloran

Marek writes the socket-mod and Xeon pages. He likes boards with too many jumpers, keeps a drawer of LGA 775 chips sorted by stepping, and reads Intel spec updates the way other people read box scores.