[NEWS] - AMD planning to move to LGA

CougTek

Hairy Aussie
Joined
Jan 21, 2002
Messages
8,729
Location
Québec, Québec
Oh no! This non-sense is spreading to AMD. After having messed with a few motherboards featuring this insertion mecanism, I'm not too happy to learn this :
A forum member of Tweakers.net web-site has posted several pictures of what is claimed to be Socket 1207, also known as Socket F, which is expected to be used for next-generation AMD Opteron processors for servers and workstations that support registered DDR2 SDRAM, virtualization technology Pacifica, advanced security feature Presidio as well as some other enhancements. The chips are projected to emerge in 2006.
  • [list:447f71e10b]
    socketf_3_sm.jpg
[/list:u:447f71e10b]
Like we say here : Ah bin Tabarnack!

News source
 

Handruin

Administrator
Joined
Jan 13, 2002
Messages
13,931
Location
USA
I don't mind the pins, even at 939 it isn't bad. That mechanism looks far too complex for what it needs to be.
 

Buck

Storage? I am Storage!
Joined
Feb 22, 2002
Messages
4,514
Location
Blurry.
Website
www.hlmcompany.com
I've only had to work with a few LGA style CPUs, and I haven't had a complaint yet. If I run into any problems, I'll just use Coug's special tool to remedy the situation.
 

ddrueding

Fixture
Joined
Feb 4, 2002
Messages
19,744
Location
Horsens, Denmark
My only question is why? What does this offer? Doesn't it add to the cost of manufacturing the CPU (socket as opposed to pins?) Without reducing the cost of the motherboard the same amont? And what of the added complexity? That thing looks to involve at least 3 steps and several moving parts. I've never had a problem with the typical socket configurations, if it ain't broke why are they trying to fix it?
 

CougTek

Hairy Aussie
Joined
Jan 21, 2002
Messages
8,729
Location
Québec, Québec
One of the problems of the LGA775 for Intel CPUs is that the most common type, the one made by Foxcon IIRC, doesn't press on the processor equally and therefore, one side ends up being very slightly higher than the other. The result is less than perfect contact with the bottom of the heatsink. For most heatsinks, it doesn't cause a problem, but for others, like one by Zalman (I don't remember which one), it results in an overheating processor.
 

P5-133XL

Xmas '97
Joined
Jan 15, 2002
Messages
3,173
Location
Salem, Or
I did a simple google search:

25 Position LGA Test Socket Report (Condensed Version - 9 pages)

25 Position LGA Test Socket Report (Full Version 4.5mb - 115 pages)

Electrical Test Report for 25 position LGA socket






LGA SOCKETS

Land Grid Array - what is it?:

Another option for device manufacturers is to package their devices without any terminations on the bottom. Such packages are referred to as Land Grid Arrays (LGA) (see datasheet 26021) (see brochure) (for test results on 25Pin LGA, click here). Although not technically accurate, the easiest way to envision an LGA device is to picture a semiconductor with nothing but tiny round gold plated pads on the bottom. (If the device were a BGA, a BGA ball would be soldered to each pad.) In order for LGA devices to get connected to a PCB, there must be some intermediary device with pins, or springs, or something, that can spring up and make contact with the little windows. So, almost by definition, a socket is required in order to connect an LGA device with its target PCB. Also, almost by definition, there must be some kind of cover or hold down mechanism that pushes the little windows of the device against the contacts of the intermediary connector.

This requirement causes a "catch-22" dilemma: there must be enough pressure (hold down force) for each of the windows to electrically connect to each of the contacts - but - there is only so much pressure you can apply. Even if the required pressure is only one ounce per contact, today's pin counts of >1000, would require hold down forces of 20, 30, and even 40 pounds or more - making any hold down mechanism very difficult to implement (not to mention the possibility of damaging the target board!)

The biggest reason for terminating a device as an LGA is to achieve higher pin counts (number of outputs) with smaller packages. At the same time, it is very important to keep the electrical path of each connection as short as possible. Previous packaging options, like Pin Grid Array (PGA) and Quad Flat Pack (QFP), all left something to be desired in achieving these goals. With new requirements such as high-end PC's requiring 1000 and more pin counts, even BGA is not an option since the large footprints can not stand the forces on the solder joints that are caused by "thermal mismatch" (the materials of the semiconductor device have different coefficients of expansion than those of the target PCB). Only a "Z-axis" connection can overcome the thermal mismatch problems.

The advantages of LGA include:

Ease of device manufacture. Just add gold plated round pads to the bottom of the device (microprocessor).
High I/O. Pin to pin spacings of 1.27 mm, 1 mm and below can be achieved with LGA. At a 1 mm pitch, 1,225 I/O's can be contained in a 35 x 35 grid that is less than 1.5 square inches (2,025 in a 45 x 45 that is less than 1.75 square inches).
Low inductance. The distance from the bottom of the device being socketed to the target board can be less than 2 mm with some LGA socket designs.
Ease of upgrade. Microprocessors can be easily removed and replaced.
No "keep out" area required. New versions of "Z-Axis" connectors can be attached in position on one side of the target board, eliminating the need for "keep out" areas required by previous termination styles.
Device cost. Devices cost less to manufacture because terminations (pins or balls) are no longer required.
Elimination of co-planarity problems. LGA sockets can be manufactured for spring movement of .015" (0.4 mm), which "takes up the slack" when there is a problem with co-planarity on the bottom of the device.
Low mating force requirements. Stamped and formed socket contacts, when gold plated, can be designed requiring only 20 to 35 grams of force per position to achieve reliable mating.
The disadvantages of LGA include:

Difficulty of soldering. The very high pin counts and the large footprints they necessitate require a high degree of surface mount soldering experience and sophisticated soldering equipment. This is similar to the "learning curve" that attaching BGA sockets initially requires.
 

Bozo

Storage? I am Storage!
Joined
Feb 12, 2002
Messages
4,396
Location
Twilight Zone
I've been building Intel systems with the LGA socket for awhile now. Never had a problem.
The clamping mechanism is positive and tight. No fear of loose connections.

Bozo :mrgrn:
 
Top