Well, I don't know really...I'm trying to go through a load of RAM on a bunch of machines (which haven't been too reliable so far - well one has) and one one DIMM two of the chips have a dull surface rather than shiny (so you can see the text in the light).
I think that these are knackered, but again I need to test them some more to make sure.
Surface finish appearance is absolutely meaningless. The mold release compound may have "bubbled" some as the casting was made, there may have been dust in the mold, variations in the formula, pigment, or even temperature of the plastic used for the case, how quickly/slowly the chips were cooled after the molding process, or any of about 14 bajillion other things may be responsible for mind-bogglingly huge visible finish variations, with none of them having any impact whatsoever on the functionality (or lack thereof) of the chip. Short of chips being blasted open, cracked, or charred (and even *THOSE* aren't
100% reliable, although they are a good indicator that the chip is probably hosed) there simply is no reliable visual indicator of a chip's functionality.
Electrical testing is the only way to be certain. Visual inspection of the (electrically inert) plastic or ceramic shell of a chip CANNOT be relied on as a diagnostic of the health of whatever circuit is on the actual (electrically active) silicon wafer inside the chip. As an indicator that the chip *MIGHT* be gone? Perhaps it can offer a clue. As a definitive, or even slightly clueful, diagnostic? No possible way unless you're looking at a chip that's literally blasted apart, and even then, you can't be certain until proper electrical testing of the circuit on the wafer is performed. Anything else is purest voodoo.
I once found a faulty RAM in an Atari 520ST by touch. The particular IC was quite noticeably hot, compared to the rest of the RAM, so it seemed fairly likely to be the problem. I can't remember if there was some reason I suspected the RAM in the first place, though I must admit it was easier to find replacement RAM than some of the dedicated ICs, so perhaps I paid attention to RAM because it was something I could repalce.
I don't know about the state of DIMM manufacturing, but in 1996 I bought some 1meg 30pin SIMMs, and had a problem with one or more of them. It looked like some of the pins hadn't been soldered properly so I used my heat gun to reflow the solder. They worked fine for years after that.
It's odd how I paid $40 for those four SIMMs (and felt I'd got a deal considering I paid $80 for 64K of RAM in 1984), and just two weeks ago I paid $40 for a 128meg DIMM (and if the rebate comes in, it will be only $20).
Not really. There are some exceptional cases where a defect makes a chip blow up and look accordingly. Some older chips did sometimes overheat which is also a means of testing, but nowadays you won't see anything on a electrically damaged or defective chip. Try programmatically testing them, there are some memory-testing programs available. Make sure however, that the CPU and chipset used for testing the RAM is reliable.