Many people don't understand the importance of the memory module electrical contacts in a computer system. Modules (SIMMs, DIMMs, and RIMMs) are available in gold or tin-plated contact form. I initially thought that gold contact modules were the best way to go for reliability in all situations, but that is not true.

To have the most reliable system, you must install modules with gold-plated contacts into gold-plated sockets and modules with tin-plated contacts into tin-plated sockets only. If you don't heed this warning and install memory with gold-plated contacts into tin sockets or vice versa, you will have problems with memory errors and reliability in the long run.

In my experiences, they occur six months to one year after installation. I have encountered this problem several times in my own systems and in several systems I have serviced. I have even been asked to assist one customer in a lawsuit as an expert witness.

The customer had bought several hundred machines from a vendor, and severe memory failures began appearing in most of the machines approximately a year after delivery. The cause was traced to dissimilar metals between the memory modules and sockets (gold-plated contact SIMMs in tin-plated contact sockets, in this case).

The vendor refused to replace the SIMMs with tin-plated versions, hence the lawsuit. Most systems using 72-pin SIMMs have tin-plated sockets, so they must have tin-plated SIMM memory installed. Studies done by the connector manufacturers, such as AMP, show that a type of corrosion called fretting occurs when tin comes in pressure contact with gold or any other metal.

With fretting corrosion, tin oxide transfers to the harder gold surface, eventually causing a high-resistance connection. This happens whenever gold comes into contact with tin, no matter how thick or thin the gold coating is. Over time, depending on the environment, fretting corrosion can and will cause high resistance at the contact point and thus cause memory errors.

You might think tin makes a poor connector material because it readily oxidizes. Even so, electrical contact is easily made between two tin surfaces under pressure because the oxides on the softer tin surfaces bend and break, ensuring contact. In most memory modules, the contacts are under fairly high pressure when the modules are installed.

When tin and gold come into contact, because one surface is hard, the oxidation builds up and does not break easily under pressure. Increased contact resistance ultimately results in memory failures. The connector manufacturer AMP has published several documents from the AMP Contact Physics Research Department that discuss this issue, but two are the most applicable.

One is titled "Golden Rules: Guidelines for the Use of Gold on Connector Contacts," and the other is called "The Tin Commandments: Guidelines for the Use of Tin on Connector Contacts." Both can be downloaded in PDF form from the AMP Web site.

Commandment Number Seven from the Tin Commandments specifically states "7. Mating of tin-coated contacts to gold-coated contacts is not recommended." For further technical details, you can contact Intel or AMP. Certainly, the best type of arrangement is having gold-plated modules installed in gold-plated sockets.

Most systems using DIMMs or RIMMs are designed this way. If you have mixed metals in your memory now, the correct solution is to replace the memory modules with the appropriate contact type to match the socket.

Another much less desirable solution is to wait until problems appear (about six months to a year, in my experience) and then remove the modules, clean the contacts, reinstall, and repeat the cycle. This is probably fine if you are an individual with one or two systems, but it is not fine if you are maintaining hundreds of systems.

If your system does not feature parity or ECC memory (most systems sold today do not), when the problems do occur, you might not be able to immediately identify them as memory related (General Protection Faults, crashes, file and data corruption, and so on).

One problem with cleaning is that the hard tin oxide deposits that form on the gold surface are difficult to remove and often require abrasive cleaning (by using an eraser or a crocus cloth, for example). This should never be done dry because it generates static discharges that can damage the chips.

Instead, use a contact cleaner to lubricate the contacts; wet contacts minimize the potential for static discharge damage when rubbing the eraser or other abrasive on the surface.

To further prevent this problem from occurring, I highly recommend using a liquid contact enhancer and lubricant called Stabilant 22 from D.W. Electrochemicals when installing SIMMs or DIMMs. Its Web site has a detailed application note on this subject if you are interested in more technical details; see the Vendor List on the DVD-ROM.

Some people have accused me of being too picky by insisting that the memory match the socket. On several occasions, I have returned either memory or motherboards because the vendor putting them together did not have a clue that this is a problem. When I tell some people about it, they tell me they have a lot of PCs with mixed metal contacts that run fine and have been doing so for many years.

Of course, that is certainly a poor argument against sound engineering practices. Many people are running SCSI buses that are way too long, with improper termination, and they say they run "fine." Parallel port cables are by the spec limited to 10 feet, yet I see many that have longer cables, which people claim work "fine."

The IDE cable limit is 18''; that spec is violated and people get away with it, saying their drives run just "fine." I see cheap, crummy power supplies that put out noise, ripple, and loosely regulated voltages, and I have even measured up to 69 volts AC of ground leakage, and yet the systems appeared to be running "fine."

I have encountered numerous systems without proper CPU heatsinks, or in which the active heatsink (fan) was seized, and the system ran "fine." This reminds me of when Johnny Carson would interview 100-year-old people and often got them to admit that they drank heavily and smoked cigarettes every day, as if those are good practices that ensure longevity!

The truth is I am often amazed at how poorly designed or implemented some systems are, and yet they do seem to work…for the most part. The occasional lockup or crash is just written off by the user as "that's the way they all are."

All, except my systems, of course. In my systems, I adhere to proper design and engineering practices. In fact, I am often guilty of engineering or specifying overkill into things. Although it adds to the cost, they do seem to run better because of it.

In other words, what one individual can "get away" with does not change the laws of physics. It does not change the fact that for those supporting many systems or selling systems where the maximum in reliability and service life is desired, the gold/tin issue does matter.

Another issue that was brought to my attention was the thickness of the gold on the contacts; people are afraid that it is so thin that it will wear off after one or two insertions. Certainly, the choice of gold coating thickness depends on the durability required by the application; because of gold's high cost, it is prudent to keep the gold coating thickness as low as is appropriate for the durability requirements.

To improve durability, a small amount of cobalt or nickel added to the gold usually hardens the gold coatings. Such coatings are defined as "hard gold" and produce coatings with a low coefficient of friction and excellent durability characteristics.

Hard gold-coated contacts can generally withstand hundreds to thousands of durability cycles without failing. Using an underlayer with a hardness value greater than that of gold alone that provides additional mechanical support can enhance the durability of hard gold coatings. Nickel is generally recommended as an underlayer for this purpose.

Special electronic contact lubricants or enhancers, such as Stabilant 22 from D.W. Electrochemicals, are also effective at increasing the durability of gold coatings. Generally, lubrication can increase the durability of a gold contact by an order of magnitude.

Increasing the thickness of a hard gold coating increases durability. AMP obtained the laboratory results in Table below for the wear-through of a hard gold coating to the 1.3 micron (50 micro-inch) thick nickel underplate. The following data is for a 0.635cm (0.250'') diameter ball wiped a distance of 1.27cm (0.500'') under a normal force of 100 grams for each cycle.

Thickness Microns	Thickness Microinch	Cycles to Failure
0.4	15	200
0.8	30	1000
1.3	50	2000

As you can see from this table, an 0.8 micron (30 micro-inches) coating of hard gold results in durability that is more than adequate for most connector applications because it allows 1,000 insertion and removal cycles before wearing through.

AMP also has a few gold-plated sockets with specifications of .001020-thick (1,020 micro-inches) gold over .001270-thick (1,270 micro-inches) nickel. I'd guess that the latter sockets were for devices such as SIMM testers, in which many insertions were expected over the life of the equipment.

They could also be used in high-vibration or high-humidity environments. For reference, all its tin-contact SIMM and DIMM sockets have the following connector plating specifications: .000030 (30 micro-inches) minimum thick tin on mating edge over .000050 (50 micro-inches) minimum thick nickel on entire contact.

The bottom line is that the thickness of the coating in current module sockets and modules is not an issue for the expected use of these devices. The thickness of the plating used is also not relevant to the tin versus gold compatibility issue.

The only drawback to a thinner gold plating is that it will wear after fewer insertion/removal cycles, exposing the nickel underneath and allowing the onset of fretting corrosion. In my opinion, this tin/gold issue is even more important for those using DIMMs and RIMMs, for two reasons.

With SIMMs, you really have two connections for each pin (one on each side of the module), so if one of them goes high resistance, it will not matter; there is a built-in redundancy. With DIMMs and RIMMs, you have many, many more connections (168 or 184 versus 72) and no redundant connections.

The chance for failure is much greater. Also, DIMMs and RIMMs run much faster, at speeds of up to 800MHz or faster, where the timing is down in the single-digit nanosecond range. At these speeds, the slightest additional resistance in the connection causes problems.

Bottom line: Definitely don't mix metal types, but for the utmost in reliability for systems using DIMMs or RIMMs, be sure both the motherboard sockets as well as the modules themselves have gold-plated contacts. You usually can tell easily because tin is silver-colored, whereas gold is, well, gold-colored.