Loss of storage capacity

[tadad1]

I have been reading up on the problem and found a post by someone who once worked in the memory chip industry, sorry I can not recall where I read it to give you a link but the gist was that when the chips are manufactured in batches there are always a number chips that are not the best quality. While they do not test every chip they random test a number of chips in each batch to determine the overall quality of the batch and then the batches are graded. The high yield batches that return a consistant number of good quality chips are sold to the highest bidder. The ones at the other end of the scale showing poor testing results well you know where they end up, in the bargin bin and are snapped up by manufacturers of cheap mp3 players and the like. So most of these players have poor quality or faulty chips because it increases the profit line. Most manufacturers will replace the player if it ends up back on their door step but also know that in the end not many are going to find their way home.

Whilst I can not verify the accuracy of this it makes sense to me when you consider the number of posts regarding chips marked a certain size ending up a constant 108MB or 233MB. I know the same batch testing is used by CPU manufactureres, the good quality ones are used in the top of the range whilst those that return slower results are used in their budget models.

[admin]

That is interesting tadad.

However, I was wondering, if the cheaper ones which show poor results, wouldn't it mean that they would be faulty altogether?

Maybe I don't understand how it all works, but I thought the bottom end chips would more likely be faulty and not working at all than produce a lower capacity.... I don't understand fully how the chips are organised so please ignore my ignorance if I am incorrect.

Most interesting though.

Either that or those manufacturers just want to rip you off and put a lower capacity chip for you anyway???

[tadad1]

Here is the techno jargon:

"Testing and sort criteria characterize the quality of the flash memory at the chip factory to determine consumer or industrial quality. NAND and AND architectures—the most prominent flash flavor used for mass storage systems—are designed to allow bad blocks in a fully operational flash device. That is, there are more flash cells than required to meet the specified device capacity. Industry jargon of “mostly good” describes this technique.
At wafer test and die sort time, the assessment of good and best flash can depend in part on the percentage of mostly good flash. Across the area of flash die there is a distribution of “weak” and “strong” cells, and even more pronounced, across a wafer there will be a distribution of bad, weak and strong die. During wafer probe it is possible to map the percentage of good cells and predict the quality of the flash device. It is not necessary for all the flash cells to be good. The bad cells are marked and the devices are sorted to consumer quality if the percentage good is moderate and to the industrial quality level if the percentage is close to 100%.

The sorting process will define the bad blocks so that the capacity of the die represents the available capacity and reflects the usable cells. It is the assumption of the sort that the devices with a low percentage of good cells will not be as reliable as the flash with the higher percentage of good cells."

My understanding of this is there are always imperfections in the chips, good and bad blocks, strong and weak cells. The quality of the chip is measured by the percentage of good cells the higher the percentage the more reliable the chip the lower the percentage the more chance there is of chip failure. So the chips are not so much faulty but poor quality with a higher percentage of weak cells and more likely to fail.