Talk:CPU socket

Old
Thanks for the merge. Makes it easier. SEG88 05:33, 21 June 2006 (UTC)

The main reason I created this page is to have a list of various (most common) cpu sockets for x86's in one place, and have them link to more detailed information. The 'CPU Sockets Chart' at the end of the page has far more detailed info, but I just wanted a short and sweet avenue for some quick info. Added the Apple ones, but I really don't have much idea what they have, hopefully someone will come along and improve that :) --Zarius 06:34, 13 Apr 2004 (UTC)

EOL Column
There is only one entry in the EOL column. Perhaps it should be removed? The two earliest sockets on the list show that they are still available, so that would mean most of the sockets on the list are still available. However, there are only a few sockets that modern computers use... I think the EOL column is inconsistent and without stated EOLs from the manufacturers I don't see much of a point. —Preceding unsigned comment added by 76.230.108.198 (talk) 22:42, 26 January 2011 (UTC)

Question based Explantions
It'd be great if this explained what single socket and dual socket means, or at least links to those articles with the information written there. 137.186.22.30 13:27, 9 November 2005 (UTC)

This is a great page for different sockets, kind of like of table of contents, but how do the sockets differ in performance. When building ones own computer why does it matter which socket is used? For instance how do the newest amd sockets compare.--eximo (talk) 19:15, 22 February 2008 (UTC)

Merge with List of CPU sockets?
The two articles seem to cover the same topic. --Qviri 05:17, 30 December 2005 (UTC)
 * I'm adding the merge templates now. --SheeEttin 17:03, 28 April 2006 (UTC)

Merge

 * Agree. The articles mostly cover the same subject. I suggest they be merged into this one for more than just a list of sockets. --SheeEttin 17:13, 28 April 2006 (UTC)
 * Agree. What we currently have now is two pages with lists. I don't see the CPU socket page getting over-filled with verbage anytime soon. Charles Gaudette 02:34, 24 May 2006 (UTC)
 * Agree. What they said ^. PierceG 02:15, 10 July 2006 (UTC)

Broken core
When you insert the heatsink on the CPU mounted in the socket, sometimes you can accidently break the CPU core.

Cleanup list
Make a table and add the year of implementation and year of EOL? --202.71.240.18 11:12, 4 January 2007 (UTC)


 * Agree. I agree with this can someone do that??? EPM Jan 4 2007 70.102.248.110 23:33, 4 January 2007 (UTC)

Here is what I've made so far, please comment. Thanks --202.71.240.18 13:44, 30 January 2007 (UTC)

The chart
Notes:

1: The 478 pin socket was introduced because it uses a micro-PGA layout which is physically smaller than the socket 423. Socket 775 was introduced with support for PCI express, DDR2 memory and Intel's version of the AMD64 processor extensions (called EM64T or Intel64), but also moved to the new LGA physical layout, where the pins are in the socket rather than on the CPU package, for better electrical performance.

2: These sockets are for CPUs with integrated memory controllers. The 754 pin models have a single memory channel routed through the CPU pins. The Socket 939 models have two memory channels, hence the higher pin count. The Socket 940 CPUs also have two memory channels but they require registered memory, and most have support for SMP. Sockets F and AM2 are redesigned to support DDR2. The Socket F contains 1207 pins (Added pins speculated to be for more scalability and better power distribution ). Socket AM2 has 940 pins but does not support Socket 940 CPUs.

Difference between socket and slot
It would be great to have a direct explanation of the difference between a socket and a slot. —Preceding unsigned comment added by 128.173.126.190 (talk) 15:55, 3 July 2008 (UTC)

Facelift
I cleaned up the intro and added a reference or two. If there is no objection I'd like to remove the current list of sockets and replace it with a table (something like the above). However, since sockets don't support memory and voltage (the cpu does this) I'd probably remove those columns and add a column for pitch (the distance between pins). I suppose adding info about capacitance, inductance, resistance and mechanical limitations might be too much. Wderousse (talk) 18:21, 27 April 2009 (UTC)

Table Colors
I added some colors to the table (Green for AMD and Blue for Intel) based on what CPU was first used on the socket. The colors don't have much contrast but I wanted to make sure the colors weren't blinding and I think they're easily distinguishable to most people. GoldKanga (talk) 12:30, 13 October 2009 (UTC)


 * Shouldn't all AMD be red? It's "Red Team." Intel is blue, so that works. "Team Green" is nVidia, not AMD. Cecoppola (talk) 00:16, 19 January 2015 (UTC)

Pentium M
I could not find any information about a Pentium M for socket 478 (only 479). Therefore I think the information in the table on socket 478 could be incorrect. Furthermore the socket 479 could contain the Pentium III-M, but there is no specific article about the III-M. Should the Pentium III-M be put in the table as a possible CPU for 479 or could it be mentioned in the Notes section? FReaKaNDeLL (talk) 21:15, 6 November 2009 (UTC)


 * I remember having this. Asus sold build your own laptops where you could purchase the mobile cpu yourself.  This was the Asus Z71.  Here is a link talking about it: https://www.vogons.org/viewtopic.php?t=9029
 * The cool thing was that some Pentium Ms had a higher multiplier and lower FSB. But the Asus z71 allowed you to pin select the FSB, thus getting an overclocked Pentium M. 24.112.244.68 (talk) 00:08, 27 May 2024 (UTC)

Sockets, versus CPU Sockets, versus Market (proprietary) names)
I believe sockets were first used in prototyping -- specifically, in wirewrapping. The sockets always had spring-type retention clips so were not under pressure until a device was inserted. The retention pressure, once a device was inserted, was not always sufficient for a proto to withstand a sudden mechanical jarring or continued low-amplitude vibration. The socket's purpose was (and is) to ensure electrical conductivity. Mechanical restraint is nearly always relegated to the design engineer and modern sockets provide suitably for both. With these older types of sockets, it was often necessary to drop in a little glue to hold the chip where it ought to be. (Cyanoacrylate glues found a commercial niche here and the trick was to drops just enough of it to tack things down without also gluing your wires or fingers to the board).

There are at least three types of sockets that can be categorized as "CPU" sockets. They are: "cheap" (for lack of knowing another (vendor) term, Zero Insertion Force (ZIF), and Land Grid Array (LGA). The latter happens to require zero force be applied to insert the device.  Wire wrap versions probably exist for each type. (These have longer pins on the board side that allow the designer or tech wrap wire around them).

Any socket is identified in computer catalogs and such, by the chip that is known to be used with them, for the sake of sales, not because it is the correct name. (This makes sense when you think about it. Socket makers are not chip makers.  They stand to make more money if their socket can hold everything, rather than just one thing). This article should take care to use only the name of the socket used by the manufacturer of the socket, when the socket is being referenced. A table can easily show which CPU or peripheral device fits into a given socket. Which socket is used is a function of the pin (or ball) spacing on the device and the total number of pins. Since any socket can be used in other ways, it doesn't make sense to force a chip name on them.

I never thought I'd see a day when engineering lingo could actually reduce the amount of text used to describe something. Now that marketers have joined the fun, this day has arrived. Just my two cents. Kernel.package (talk) 19:30, 14 November 2010 (UTC)

Pro's and Con's
I'm not sure why chip sockets should be restricted only to CPU chips. Nevertheless, there are Pro's and Con's to using any sort of chip socket on a circuit board. IMO, sockets are fine for prototyping but not for production.--74.107.74.39 (talk) 22:48, 26 March 2011 (UTC)

Pro's and Con's include:


 * Pro's


 * Ability to replace failed or faulty components with replacement devices --74.107.74.39 (talk) 22:46, 26 March 2011 (UTC)
 * (Possible) ability to replace component with one of higher performance as technology becomes available --74.107.74.39 (talk) 22:46, 26 March 2011 (UTC)
 * Ease of maintainability of failed chips --74.107.74.39 (talk) 22:46, 26 March 2011 (UTC)
 * Ability to easily test multiple chips --74.107.74.39 (talk) 22:46, 26 March 2011 (UTC)
 * (Arguably, may) extend the service life of the product/system --74.107.74.39 (talk) 22:46, 26 March 2011 (UTC)


 * Con's


 * The socket adds cost to the circuit board --74.107.74.39 (talk) 22:46, 26 March 2011 (UTC)
 * The use of a socket is not as reliable when compared to soldering the chip directly to the circuit board (particularly where shock, vibration or contaminants could be present in operational use) --74.107.74.39 (talk) 22:46, 26 March 2011 (UTC)
 * The socket adds extra lead inductance, which could cause problems in signal integrity (when compared to soldering chip directly to the circuit board) --74.107.74.39 (talk) 22:46, 26 March 2011 (UTC)
 * The ability of the circuit board to act as a heat spreader is diminished when a socket is used (also diminishing reliability) --74.107.74.39 (talk) 22:46, 26 March 2011 (UTC)
 * There may be no assurance that a higher-performance replacement will be available or compatable with the older chip/component, or if it is, that it will work well in the original socket (... the counter argument on service life above). --74.107.74.39 (talk) 22:46, 26 March 2011 (UTC)

LCC and PGA missing for the 80286
Leadless_chip_carrier should be added for the 80286, which combined a small heatsink with the CPU clamp and was extremely common on 12Mhz and faster 80286 CPUs. The LGA types could be considered an extreme evolution of LCC. The 80386 was most commonly packaged in PGA while the 80386SX most commonly was produced in the same size PLCC as many 80286 CPUs as a way to use existing stocks of 16bit chipsets designed for the 80286. Some 80486SX CPUs were shoehorned into the same 16 bit bus package and could be used as "drop in" upgrades for 286 and 386SX systems, requiring a program to be run during boot to enable use of the CPU's cache RAM. Bizzybody (talk) 10:39, 21 February 2012 (UTC)

Laptops using surface mount CPUs
The text "Laptops typically use surface mount CPUs, which need less space than a socketed part" is dubious to me. I repaired laptops professionally for years and only ever came across a small number that actually did this. It's certainly not typical. I would recommend this is removed 86.29.95.83 (talk) 15:10, 7 March 2012 (UTC)

Reasons for lack of standardization / History of socket type divergence
Lists of socket types is fine and all, but this is Wikipedia. Lets examine the subject of CPU/GPU compatibility with sockets and be critical of why this is being done. What should be known is if there are good reasons for manufacturers to be frequently changing socket types. Is this is primarily in order to keep consumers on the upgrade treadmill, or would it be easy to create standards for socket types if it was a higher priority. AMD and Intel used to both use the same socket type, but now they have broken off from doing so. A history on that and the reasons for doing so is important to note. I'd think a greater number of pins for greater bandwidth would be the primary reason for socket changes, and Intel seems to follow that with the ever-increasing pin count, though AMD does not and has remained around 940 for a long time, in one case differing by only a single pin. You would then suspect AMD of just using the lack of standards as a tool for pushing consumerism. In Intel's case, if they know they continually want additional pins, perhaps they could be accused of not making a socket with more pins than are needed at the time in order to future-proof their socket types. This section could also mention that perhaps some of the reason for lack of interest in more solidly standardizing CPU sockets is less interest by consumers in not upgrading their motherboards because of the other components on the motherboard (north bridge, south bridge) not being upgradable due to those chips lacking sockets and standards themselves. Does anyone have any articles on the subject of designing motherboards to be fully interchangeable for the conservation of resources in technological advancement? Does anyone have any articles on the reasons behind, for and/or against, the lack of socket standardization to create such a section? Also, the mounting positions for CPU fans and coolers being slightly different between socket types which forces the purchase of new heat sinks should also be noted. Yfrwlf (talk) 12:07, 18 March 2012 (UTC)
 * The Connector conspiracy has been with us since UNIVAC. Is it not the case that most personal computers live out their whole lives with the chips they were born with? The market fraction of people swapping CPUs in computers must be dwarfed by the market fraction of people who swap engines in cars; you don't hear them griping about 5 bolt transaxles and 6 bolt flywheels, they just machine up adapters and have at it. It's not like 30 years ago when you could swap chips and get a clearly beneficial performance step. --Wtshymanski (talk) 14:53, 18 March 2012 (UTC)
 * Huh? That made no sense.  Motherboards don't get upgraded now days because it usually can't be done.  It doesn't matter to what degree it is a conspiracy between makers, that is besides the point.  It is poor planning and bad engineering.  Engineering something which can be easily re-used, replaced, swapped out, re-built, and is compatible with other common things is what engineers do who are concerned about usability, durability, longevity, cost to citizens, etc.  Those aren't things that modern corporations today are concerned about and are instead concerned with doing the exact opposite.  That is the reason these things aren't done, regardless of who is colluding and conspiring together.  Back on the subject of AMD vs. Intel though, it'd be my guess that AMD followed Intel's socket types while AMD was small.  Then, once they got big enough to resist citizens getting too upset over proprietary sockets, they started making those. Yfrwlf (talk) 19:10, 24 March 2012 (UTC)

Laptops
The article confidently states that "Laptops commonly use surface mount CPUs". I challenge anyone here to find a single mass produced laptop in the last ten years which has, using CPUs from AMD or Intel. 20:23, 27 April 2012 (UTC) — Preceding unsigned comment added by Wayne Hardman (talk • contribs)

cleanup/updates
I edited some table entries to improve consistency and updated an entry.
 * added vishera/piledriver to the am3+ entry
 * the cpu entry for FM2 was just 'A10', which reflects a specific model of processor, rather than a series as all the other entries are. changed to 'Trinity Processors'.
 * added links to the processor series entries for Llano, Bulldozer, and Trinity.
 * changed bulldozer to zambezi. I fell that labeling as bulldozer and Piledriver is too general, as that could be interpreted to include the Apus that use that arcitecture, so labeling as vishera and Zambezi makes it clear that "we" are referring to the FX processors specifically.
 * I removed 'AMD FX' from the AM3+ entry, as it was redundant (and possibly confusing) since Zambezi and Vishera are listed separately. however, I added 'FX' into the entries for vishera and zambezi instead.
 * aded 'desktop to the appropriate column for FM1, AM3+, and FM2.

my edits should prevent confusion from the table. I haven't looked at the rest of the article, but this should be a big improvement. Aunva6 (talk) 06:42, 20 January 2013 (UTC)

Intel socket FCPGA946
It seems there is no mention anywhere of the above socket. This is just one of the sockets used by Intel's 4th gen chipset. — Preceding unsigned comment added by 122.106.147.61 (talk) 01:00, 15 January 2014 (UTC)
 * I am not really sure what you mean by 4th gen chipset (perhaps you mean 4th gen Core architecture), but I believe your reference to FCPGA946 is a reference to socket G3. 50.53.15.59 (talk) 11:34, 17 January 2014 (UTC)

Please update this Article - Reference.
This article requires updating.

Thank you. Mani — Preceding unsigned comment added by 50.70.121.49 (talk) 03:05, 25 March 2014 (UTC)

AMD AM1
Can you add AM1 to the table? Introduced in 2014 by AMD for their low power/performance APUs. TinyEdit (talk) 21:37, 15 June 2014 (UTC)

AMD FM2+
Can someone add the FM2+ to the table? It was added in 2014 I believe. (http://www.cpu-world.com/Releases/Desktop_CPU_releases_%282014%29.html) 70.27.26.133 (talk) 05:37, 7 October 2015 (UTC)

AMD FP3
The Socket FP3 isn't included on the list.--MisterSanderson (talk) 16:08, 13 February 2019 (UTC)

Heatsink Mounting Dimensions
I propose that we add a column to the list of x86 sockets with their respective heatsink mounting dimensions. Many of the pages already include this information, see LGA 775, LGA 1150, Socket AM4 - it would be useful to have this information all in one place. What do people think of this? --Emil.Carr (talk) 16:01, 2 May 2020 (UTC)

Where's the list of (modern) mobile sockets?
this list should be split into desktop and mobile sockets, and possibly enter a date of introduction (and eventually end of production date), and, on the subject - in the list there are no newer mobile sockets / sockets that use mobile processor

nhf, all the best :) — Preceding unsigned comment added by 93.136.208.220 (talk) 22:00, 22 August 2020 (UTC)

For example this list lacks FCBGA1170 — Preceding unsigned comment added by 93.138.25.42 (talk) 23:39, 30 October 2020 (UTC)

j
j 103.135.216.176 (talk) 07:09, 25 November 2022 (UTC)

Year of date and voltage missing in list
In the list of sockets the year and allowed voltage is missing. 84.140.204.50 (talk) 18:43, 12 July 2023 (UTC)