LGA1700 socket: bent pins — repair or live without it, an interactive pinout and socket replacement
Contents 11
LGA1700 is the socket for Intel Core 12th, 13th and 14th gen processors. Its contacts are thin spring leaves about 0.8 mm apart, and they bend easily: a dropped CPU, a CPU set in at an angle, a careless finger. In this article:
- how to inspect the damage and find the contact's coordinates on the socket grid;
- an interactive pinout: click a contact to see where it goes and whether you can leave it unrepaired;
- what stops working if a memory, PCIe, configuration or power contact is broken;
- what to straighten a contact with, how to do it, and what to do if it snaps off;
- the "banana effect": why the CPU and board bow, and how that affects the cooler and the socket contacts;
- replacing the socket at home: bottom preheating, hot air, protecting the plastic, and why it's a hard repair;
- a summary table: situation → what to do.
draft — collecting facts and proofreading
We extracted the pinout from the boardview (component placement file) of the ASUS ROG STRIX Z790-A GAMING WIFI D4. Everything else — soldering temperatures, the banana effect, contact frames — comes from socket manufacturers' documentation and reviewers' tests; sources are at the end. We haven't replaced a socket ourselves for this article.
In short
| Question | Answer |
|---|---|
| A contact is slightly bent | straighten it under magnification with a fine needle or tweezers, one at a time, gently |
| A contact has snapped off | look it up in the pinout: ground or core power can often be left as is |
| A memory contact is broken | that memory channel fails or throws errors; workaround — use only the other channel |
| A PCIe lane to the graphics card is broken | lanes 8–15 — the card runs at x8; lanes 0–7 — must be restored |
| A CFG, control or clock contact is broken | the board won't start — this one has to be fixed |
| The CPU and board bow ("banana") | tighten the cooler evenly; if needed, use a contact frame |
| Many broken contacts close together | socket replacement: needs bottom preheating and BGA experience, otherwise a repair shop |
What LGA1700 is and why its contacts bend
LGA1700 (Socket V) is the Intel socket released in November 2021 for Alder Lake processors and later used for Raptor Lake. The CPU package is 37.5 × 45 mm — a long rectangle rather than the square of earlier sockets (Wikipedia).
LGA (land grid array) means the processor has no pins — just flat gold-plated pads. The springy contacts sit in the socket on the motherboard: 1700 thin metal leaves, each slightly tilted and raised above the plastic. When the CPU is clamped down, the leaves flex and press against the pads.
According to the boardview, the contacts are about 0.8 mm apart in both directions. Each one is a thin spring that's easy to push sideways or flatten against the housing. It usually happens when:
- the CPU is dropped into the socket, or set in at an angle and slid;
- something falls onto a socket without its cover, or someone touches it with a finger;
- the CPU is pulled out with the cooler not fully removed, and the cooler drags it out.
Underneath, each contact is soldered to the board with a solder ball — the socket is soldered just like a BGA chip. So the socket can be replaced, but it means reflowing 1700 balls at once.
Safety first
| What | Why |
|---|---|
| switch off and unplug | the board carries standby voltage; a screwdriver or needle could short something |
| grounded antistatic wrist strap | a discharge from your hand goes through the contact straight into the CPU and board circuits |
| don't touch the contacts with fingers | skin oil and sweat mean worse contact and corrosion; a finger easily snags and bends a leaf |
| socket cover in place | when there's no CPU in the socket, the cover is the only thing protecting the contacts |
| hold the CPU by the edges, pads up | the CPU pads must not be touched or pressed either |
Keep the protective cover: you'll need it whenever the board is shipped, stored or repaired without a CPU.
Assessing the damage
What to inspect with
With the naked eye you'll only see a contact that's completely flattened. To spot a slight lean, you need magnification and light.
| Tool | How to use it |
|---|---|
| magnifier or head loupe, 10–20× | the main tool: hands free, individual leaves visible |
| USB or soldering microscope | the most convenient: a large image you can photograph and compare "before" and "after" |
| flashlight at a low angle | side light creates reflections: straight contacts all glint the same, a bent one doesn't |
| phone camera through a magnifier | take a photo at full zoom, then study it on screen and compare the rows |
Look not just from above but at an angle from all four sides: every contact leans the same way, so a bent one stands out by its reflection and height. A well-lit straight row looks like a comb, and a tooth that sticks up or lies flat is obvious.
Finding the contact's coordinates
Every contact has a name made of letters and a number, such as AB12: the letters give the row, the number the column.
- Columns are numbered 1 to 52.
- Rows are labelled with letters: single first (
A,B,C…Y), then double (AA…AY,BA…BD). The lettersI,O,Q,S,XandZare skipped by convention, so they aren't confused with digits or with each other. SoHis followed byJ,NbyP, andYbyAA. There are 44 rows in total. - In the middle of the field there's a rectangular window with no contacts.
How to count in practice:
- Photograph the socket from above, from the same side as the pinout below, and rotate the photo so the window and the edges of the field line up with the picture.
- Find the nearest corner of the field or edge of the window — a reference point with a known name (see the pinout).
- Count rows and columns from it on the enlarged photo. Marking every fifth contact with a dot helps.
- Check yourself by distance: with a pitch of about 0.8 mm, ten contacts span about 8 mm.
If you're torn between two neighbouring contacts, look both up in the pinout. They're often in the same group, and the answer is the same.
Pinout: can you leave it unrepaired?
Not every contact matters equally. More than half the socket is ground and power: hundreds of contacts connected in parallel. If one of them snaps off, the current flows through its neighbours. A memory line, a PCIe lane or a configuration signal, however, exists only once: lose it and something specific stops working.
We built an interactive version: the contacts are in the same orientation, and clicking one shows its signal, its group, where it goes, and what happens if it's broken. You can search by ball name (AB12) or signal (DQ_A14); the legend works as a filter.
The pinout runs in the browser and needs JavaScript.
💡 The pinout is linked to the virtual DDR5 board view in the memory repair article: for memory lines the tool shows the slot contact and a link to the board view, and from the board view you can come back here, to the socket's memory area.
We extracted the data from the boardview of the ASUS ROG STRIX Z790-A GAMING WIFI D4. The memory signal names there are for DDR4. We checked against Intel's datasheet what this means for DDR5 boards:
- The processor's memory contacts are the same for DDR4 and DDR5 — the memory controller runs in one of two modes, and the same ball has its own role in each. The widely shared pinout picture above shows this in its labels: many memory balls carry two names separated by "/", one for DDR4 and one for DDR5.
- The channels are named differently. In DDR4 mode the processor has two 64-bit channels (DDR0 and DDR1 in the datasheet, plus an ECC byte). In DDR5 mode it has four 32-bit channels (DDR0–DDR3, each with 4 or 8 more ECC bits), and each DDR5 slot uses two of them: that is how DDR5 itself works, with each module split into two independent sub-channels.
- The signal roles differ. In DDR5, address and commands travel on CA lines, while DDR4 uses separate MA, BA, BG, ACT and CKE lines; so on a DDR5 board a ball labelled, say,
D4_CKE_A1in our data carries a different DDR5 signal.
So on a DDR5 board the conclusion "a broken memory contact affects memory" still holds, but the exact signal and bit number differ. Which slot (A or B) is affected is most likely the same, because each memory controller's contacts stay in place on the processor package, but we have not checked this ball by ball: Intel publishes the pinout with the names for both modes only in restricted documents for board designers.
⚠ This leads to an important caveat. Balls that are "not connected" on our (DDR4) board, or go only to a test point, may be used on a DDR5 board — for example, as signals of channels DDR2 / DDR3. So the "can be left as is" verdict for such balls holds only for DDR4 boards; on a DDR5 board, a broken ball in the memory area is better restored.
How many contacts in each group
| Group | Contacts |
|---|---|
| VSS ground | 706 |
| VCCCORE core power | 281 |
| DDR memory (two channels, A and B) | 273 |
| not connected, reserved (RSVD) | 52 |
| not connected here, memory area (most likely DDR5 signals) | 32 |
| not connected here, spare video-output and PCIe lanes | 14 |
| PCIe x16 to the graphics card | 64 |
| VCCGT graphics power | 51 |
| other power (VCCIN_AUX, 1.05 V, 1.8 V) | 43 |
| test points | 33 |
| DMI to the chipset | 32 |
| VDDQ memory interface power | 20 |
| CFG configuration | 19 |
| DisplayPort and HDMI video outputs | 18 |
| control (SVID, PWRGD, VCCST_PWRGD, PROCHOT, THERMTRIP, PECI …) | 16 |
| PCIe to the CPU-attached M.2 slot | 16 |
| JTAG debug | 11 |
| voltage feedback (voltage regulator sense lines) | 8 |
| clocks | 8 |
| HDA audio | 3 |
Almost two thirds of the contacts (about 65%) are ground and power: VSS, VCCCORE, VCCGT, VDDQ and other power. So the odds that an accidentally broken contact can be left alone are fairly good.
What happens if a contact is broken
| Group | What happens | Verdict |
|---|---|---|
| VSS ground | hundreds in parallel: one broken goes unnoticed; several together — instability under load | can be left as is |
| VCCCORE core power | same as ground | can be left as is |
| VCCGT graphics power | same; with a discrete graphics card the integrated graphics aren't needed anyway | can be left as is |
| VCCIN_AUX, 1.05 V, 1.8 V | few in parallel, each carries a noticeable share of the current | better to restore |
| VDDQ (memory interface) | few in parallel | better to restore |
| DDR memory | the channel fails or throws errors; workaround — modules only in the other channel's slots | depends, better to restore |
| PCIe x16, lanes 0–7 | the card runs at x4 or narrower, or doesn't work at all (see below) | must be restored |
| PCIe x16, lanes 8–15 | the card runs at x8; in games the loss is usually small | depends |
| PCIe to the CPU-attached M.2 | the drive is slower or isn't detected | depends, better to restore |
| DMI to the chipset | a narrower link to the chipset — everything on the chipset is slower; or no boot | must be restored |
| DisplayPort and HDMI outputs | that port on the rear panel doesn't work | depends |
| HDA audio | most likely, no audio over the integrated graphics' HDMI or DisplayPort | depends |
| voltage feedback | the voltage regulator loses its accurate reading of the voltage at the CPU | depends, better to restore |
| CFG configuration | without it the board won't start, or starts wrongly | must be restored |
| control (SVID, PWRGD …) | without it the board won't start, regulate voltage or protect against overheating; some signals are auxiliary, but that's hard to tell without the board schematic | must be restored |
| clocks | the CPU doesn't run without its clock | must be restored |
| JTAG debug | only needed at the factory and in repair shops | can be left as is |
| test points, not connected (reserved, RSVD) | go nowhere on this board; the pinout picture labels them RSVD | can be left as is |
| not connected, but in the memory area | go nowhere on a DDR4 board; the pinout picture labels the area with DDR2 / DDR3 channels — most likely DDR5 signals | on DDR4, can be left; on DDR5, better to restore |
| not connected, next to video outputs and PCIe | spare lanes for ports this board does not have | can be left (check on a board with other ports) |
⚠ "Can be left as is" means you don't have to bring the broken contact back into service. But it has to be removed, or bent so it doesn't touch its neighbours or press against the wrong spot on the CPU. A ground contact lying on a neighbouring memory contact shorts that signal to ground.
⚠ The table is for one board. On another model some "not connected" contacts may be in use, and there may be no video output at all. For power and ground the conclusion is the same on any board.
PCIe x16: what link width is left
At start-up the graphics card and the CPU negotiate the link width and pick the widest one that works: x16, x8, x4, x2 or x1. Lanes are numbered from zero. Each lane is four contacts: a transmit pair (TX+ and TX−) and a receive pair (RX+ and RX−), so losing any one of the four takes out the whole lane. The working width is the widest one that doesn't include the broken lane:
| Broken lane | What's left | How noticeable |
|---|---|---|
| 8–15 | x8 | in games, the loss is usually small |
| 4–7 | x4 | noticeable with powerful cards |
| 2–3 | x2 (or x1 if the card can't do x2) | large |
| 1 | x1 | severe |
| 0 | the card may not work | lane 0 is needed for any width |
⚠ This assumes the normal lane order. Some boards and cards reverse the lane order (lane reversal), and then the same broken lane costs you something different. So the lane number is a hint; the final answer comes from checking after boot.
You can check the actual link width in GPU-Z (the "Bus Interface" field) or in the card maker's utility. If it shows x8 instead of x16 after a repair, suspect a lane numbered 8–15.
Our case: a missing pin at W5 and a board that "doesn't see" the CPU
On our ASUS TUF GAMING Z790-PLUS WIFI board, one contact is missing from the socket: the previous owner broke it off. We found its position on the pinout — it is ball W5, and the tool above lists it as ground. Before deciding not to repair it, we checked with a multimeter: on both the socket and the CPU this contact beeps through to ground, so it is just one of hundreds in parallel. The computer works normally with it missing, which is exactly what you would expect for a ground pin.
The second problem on the same board is more interesting. If the cooler is mounted wrongly — too loose, too tight or unevenly — the computer fails the CPU check at start-up. The first such check can take a few seconds; later ones pass quickly. We went through the possible causes:
- something damaged by the previous owner;
- cracks in the board around the cooler mounting holes;
- cracked solder balls under the socket itself;
- the CPU and board bending — the banana effect (see below);
- bad contacts in the socket itself (unlikely: they all look straight);
- a non-original socket frame.
On the back of the board, next to the cooler mount, there were indeed cracks. We resoldered a large capacitor and a smaller one next to it, plus two traces. The lesson: if the computer starts or fails depending on how the cooler is clamped, inspect not just the socket contacts but the board around the mount, on both sides.
Straightening a contact
Tools
| Tool | What for |
|---|---|
| fine antistatic tweezers, curved and straight | catch a leaf from the side and turn it back; the curved tip reaches between rows |
| sewing needle or fine syringe needle | hook a flattened contact and lift it slightly; nudge it sideways |
| 0.5 mm mechanical pencil tip | a "sleeve" over the contact tip to straighten it at the bend (no lead, metal tip) |
| thin blade | slide between rows to straighten a contact lying along the row |
| cotton swab | only around the socket: never wipe the contacts, the fibres snag on the leaves |
| head loupe or microscope, 10–20× | don't work without magnification |
| good light from several sides | to see the tilt and height of the contacts |
Don't push a contact head-on with a needle: the tip slips. Slide the needle under the leaf from the side and gently lever it in the direction you want (Overclock.net). Use a pencil with a metal tip — a plastic one can break (Overclock.net). On LGA, though, it doesn't fit everywhere because the neighbours are so close, and some repairers use a needle or a thin blade instead (Tom's Hardware Forum).
Technique
- Find the "right" direction. All healthy contacts lean the same way and stand at the same height. Your goal is to make the bent one look like them, not to stand it upright.
- One contact at a time. If several are bent, start with the ones lying on top of the others.
- Work from the base, not the tip. The bend is usually near the base. Pulling on the tip just makes a new bend somewhere else.
- Small moves, no force. The contact is thin; one sharp move and it snaps. Move a fraction of a millimetre, then look.
- Don't bend it back and forth. Every bend work-hardens the metal and makes it brittle (Overclock.net). Better to stop short and correct than to overshoot and bend it back.
- Compare its height with its neighbours. Seen at an angle against the light, the straightened contact should line up with the row at the same height. Too low and it won't reach the pad; too high and it may stop the CPU from sitting flat.
- Check the whole row afterwards. Tweezers and needles easily catch the neighbours.
After straightening, fit the CPU, power up and check everything that could have been affected: both memory channels (at least one pass of a memory test), the graphics card's link width, the CPU-attached M.2, and the board's video output.
If a contact snaps off
| Option | When it fits |
|---|---|
| leave it | a ground or core power contact, neighbours intact; the fragment removed, nothing touching |
| live without the feature | a memory, M.2 or video output contact: use the other channel, another slot, another port |
| transplant a contact from a donor board | possible but hard: pull a leaf from the donor and solder it in place under a microscope |
| replace the socket | several important contacts, or CFG, control or clock contacts |
Transplanting a contact is a job for repair shops with a microscope and experience. The contact is held in plastic and soldered underneath, with neighbours 0.8 mm away. At home it's usually more realistic either to live without the feature or to replace the socket.
The banana effect
The LGA1700 independent loading mechanism (ILM — the metal frame with a lever) doesn't press on the CPU around the whole perimeter, but at two points along the long sides. Under this load the long, thin CPU bows: the middle sinks lower than the short edges, and the whole heat spreader, along with the board beneath, curves into an arc (Igor's Lab). On LGA2066, by comparison, the mechanism presses on the CPU's corners and spreads the load better.
With a curved heat spreader, the cooler base makes uneven contact, and heat has to cross a thicker layer of thermal paste. You may see this in the paste imprint on the removed cooler: the layer is thicker in the middle than at the edges.
The other half of the problem is the socket contacts themselves. All 1700 contacts are pressed onto the pads by the ILM load plus the cooler's. Too much pressure and the board bows even more; in its tests of the Thermal Grizzly contact frame, Igor's Lab found that tightening it "finger-tight plus a quarter turn" made a DDR5 overclock unstable, while a slightly looser frame didn't (Igor's Lab). Too little or uneven pressure, and some contacts may touch their pads less reliably. Either way the symptoms look like a bent contact: memory errors, a "missing" channel, a board that won't start. So before hunting for a bent contact, check how the cooler is mounted.
In 2022 Intel said this deflection is expected and doesn't take the processor out of specification, and strongly advised against modifying the socket or the ILM, warning that modifications may void the warranty (PC Gamer, Neowin).
What to do
| What | How |
|---|---|
| tighten the cooler in a cross pattern | screws in turn, diagonally, a few turns each, not one all the way at once |
| don't over- or under-tighten | to the stop or the torque in the cooler manual; "a bit more to be safe" just bends the board further |
| contact frame | replaces the stock ILM; presses the CPU evenly around the perimeter |
| washers under the ILM | raise the ILM and reduce the load; Igor's Lab saw core temperatures almost 6 °C lower with 1 mm washers |
| test after any change | a memory test and a load test: check for new errors |
The Thermal Grizzly contact frame for 12th gen needed a very low torque — 0.03–0.06 N·m (PC Perspective), which is less than you get by tightening with your fingers. At Igor's Lab even "finger-tight plus a quarter turn" proved too tight: the DDR5 overclock became unstable (Igor's Lab); the 13th/14th gen version has a revised shape and no longer needs a specific torque (Thermal Grizzly). The Thermalright LGA1700-BCF has stand-offs underneath that rest on the board, so it's hard to overtighten (Igor's Lab). In Tom's Hardware tests the Thermalright frame lowered temperatures by up to 12 °C — but only where the cooler had poor contact to begin with (Tom's Hardware).
⚠ Washers and contact frames modify the loading mechanism: Intel warns that this may void the warranty (PC Gamer).
⚠ When removing the stock ILM, leave the CPU in the socket and don't tilt the board: only its own weight holds the CPU there. Lay the board on a flat surface so the socket backplate doesn't fall out, and only do this with the board unpowered (Igor's Lab).
Replacing the socket at home
Let's be honest: this is a hard repair. A socket is 1700 solder balls under a plastic housing, and all of them have to melt and settle at the same time. The board is multi-layer, with large copper ground and power planes that soak up heat, and the voltage regulator's chokes and capacitors are right next to the socket. Without BGA soldering experience and bottom preheating, you're more likely to damage the board than to succeed. If you don't have that experience, take the board to a repair shop.
Which socket to buy
⚠ Listing titles are sometimes wrong. Here the seller wrote LGA4094, but LGA4094 is a large AMD socket (SP3 / TR4) for EPYC and Threadripper processors — something else entirely. So search for LGA1700 and check:
| What to check | How |
|---|---|
| shape | a rectangle, a window without contacts, the keys in the same places as on yours |
| number of contacts | 1700; LGA1851 is the same size but has different keys and more contacts |
| solder balls underneath | it must be the BGA version with balls, not contacts without solder |
| protective cover (pick-up cap) | a new socket should come with its cover: it's needed for soldering and protects the contacts |
LGA1851 (Core Ultra 200S processors) is the same 37.5 × 45 mm size, but the keys are in different places (Wikipedia) — easy to mix up in a photo.
Heating from below and above
For reference, here are the figures from documentation for Intel sockets that are soldered the same way (TE Connectivity, LGA115X, lead-free solder):
| Parameter | Value |
|---|---|
| SnAgCu solder melting point | 217 °C |
| preheat zone in the oven profile | 150–170 °C |
| peak at the joints during soldering | 245 ± 5 °C |
| peak at the joints during rework | 228–250 °C |
| time above 217 °C | 45–280 s |
| ramp rate in the 210–220 °C zone | 0.35–0.75 °C/s |
| maximum on the socket cover and housing | 260 °C |
| temperature spread across the socket area | no more than 15 °C |
The 150–170 °C in this table is the preheat zone of a factory oven profile, i.e. the temperature of the joints themselves before melting. It's not the same as bottom-heating the board during a repair (see below).
Source: TE Connectivity 114-5444. We didn't find a similar table for LGA1700; the solder and plastic there are of the same class, so we use these numbers as a guide. Intel states that the housing and cover of its sockets withstand 260 °C for 40 seconds (Intel, LGA1155).
Why from below. Most of the heat comes from the bottom heater — a preheater or a hot plate under the board. It warms the whole board slowly and evenly. For LGA rework, Flex recommends bottom-heating the board to about 125 °C, and a larger bottom heater for thick boards (Flex AN317). Slow heating keeps the board from warping due to temperature differences, and stops moisture inside the board and components from boiling suddenly and splitting the layers — the so-called "popcorn" effect.
Should you heat from above? Only as a helper, to bring the balls up to melting point once the board is already hot from below. LGA socket housings are usually LCP (liquid crystal polymer), a plastic that survives reflow in an oven (TE LGA). An oven heats evenly; a hot air station blows a narrow jet. Right under the nozzle the plastic can easily go above 260 °C, warping or melting the housing and shifting the contacts. Hence the hot air rules:
- a wide nozzle, low airflow, moderate temperature;
- keep moving over the whole socket area, not one spot;
- a heat shield around the socket — aluminium foil or Kapton tape over nearby parts;
- a thermocouple or pyrometer on the board next to the socket and on the cover, not guesswork.
Step by step
- Dry the board before soldering: moisture in the board's layers causes the same "popcorn" effect under sudden heat. Work under an extractor or with a window open — flux fumes are harmful to breathe.
- Strip the board of everything you can: CPU, cooler, ILM and its backplate, battery, memory modules, M.2 drives. Intel advises removing the ILM and backplate before reworking the socket (Intel, LGA1155).
- Cover nearby connectors and plastic parts with foil or Kapton.
- Preheat the board slowly from below. Then, keeping the bottom heat on, bring the socket up from above until the balls melt, and lift the old socket straight up without force.
- Clean the pads with flux and solder wick so they're flat and free of old ball remnants. Don't overheat: pads lift off the board with too much heat and rubbing.
- Apply a thin layer of flux. Place the new socket with its protective cover on, exactly on the marks and keys. The cover stays on during soldering to keep the contacts clean (Intel, LGA1155), and comes off only right before the CPU goes in (TE 114-5444).
- Same profile: slowly from below, then extra heat from above. When the balls melt, the socket settles slightly and self-aligns through surface tension. Don't press on it.
- Let it cool slowly without moving the board.
- Check: the socket sits flat, not tilted or shifted, with the same gap to the board on every side; no stray balls or bridges nearby. In production the joints are checked by X-ray or electrical testing (TE 114-5444); at home that leaves visual inspection and a trial boot with memory, PCIe and M.2 checks.
⚠ A socket that has been removed is not reused: the manufacturer states plainly that the socket is not repairable and must not be refitted once removed from the board (TE 114-5444).
Summary: what to do
| Situation | What to do |
|---|---|
| the board won't start after a CPU swap | first the cooler and ILM mounting, then inspect the socket under magnification |
| one or two contacts slightly bent | straighten with a needle or tweezers under magnification, compare height with neighbours |
| a contact lies flat but is intact | lift it from below with a needle, then work from the base with tweezers; don't flex it back and forth |
| a ground or core power contact has snapped off | can be left; remove the fragment and check nothing touches the neighbours |
| a memory contact has snapped off | modules only in the other channel's slots, or repair the contact / socket |
| a PCIe x16 lane has snapped off | check its number: 8–15 — x8, you can live with it; 0–7 — needs repair |
| a CFG, control or clock contact has snapped off | must be repaired: contact transplant or socket replacement |
| memory errors, a "missing" channel, but contacts look fine | check cooler tightening; if needed, a contact frame |
| many damaged contacts close together | socket replacement; without BGA experience and bottom preheating — a repair shop |
| buying a replacement socket | search for LGA1700, check the shape, keys, solder balls underneath and the cover |
Sources
- Intel: 12th Gen Core processors datasheet, volume 1 — DDR5 interface signals
- Intel: the same datasheet — DDR4 interface signals
- Intel: the same datasheet — memory channel organisation (DDR4 / DDR5)
- Wikipedia: LGA 1700
- Wikipedia: LGA 1851
- Igor's Lab: Alder Lake's cooling problem and the ILM washer mod
- Igor's Lab: Thermal Grizzly contact frame and Alphacool Apex backplate
- Igor's Lab: Thermal Grizzly, Thermalright, Jeyi and Feng Zao frames compared
- Tom's Hardware: Thermalright LGA1700-BCF contact frame review
- PC Gamer: Intel responds to Alder Lake CPU bowing
- Neowin: Intel on CPUs bending in LGA1700
- Thermal Grizzly: contact frame for 13th and 14th gen
- PC Perspective: lapping, delidding and a contact frame on 13th gen
- TE Connectivity 114-5444: LGA115X socket application specification
- TE Connectivity: LGA socket, housing and contact materials
- Intel: LGA1155 socket guide
- Flex AN317: LGA design and assembly
- Overclock.net: fixing bent socket pins
- Overclock.net: bent LGA pins, the pencil tip method
- Tom's Hardware Forum: can bent pins be straightened
- Pinout: picture from a Reddit thread titled "LGA 1700 pinout problem", author unknown.
- Interactive pinout: our data from the boardview of the ASUS ROG STRIX Z790-A GAMING WIFI D4.