Blog

Why are there different grades of flash memory – and what does that have to do with pizza? 

Flash memory is on the face of it a very simple thing to understand. It’s a data repository. The data is written to the flash memory and can be read from the flash memory. At its most basic, that’s really all there is to it. But for firms like Cardwave, which helps companies select the right flash for their requirements, and then provision that flash with the right data, flash is a much more detailed, nuanced, and complex product. In this article, we’re going to focus on the production process, and how flash is tested, graded, and used by manufacturers to build different categories of memory products. And there will be pizza. 

Wafers, dies and grading 

All flash memory is made in a similar way regardless of the industry in which it is used, its size and whether it is embedded deep within a device or available as a user-removable product. Flash is made on a thin, circular slice of semiconductor called a ‘wafer’.  

A single wafer will contain many separate ‘dies’ or individual instances of what will become the flash memory in, for example, a microSD card or behind the scenes in your new car, providing the mapping data to its navigation system. 

But there are processes to go through before that stage is reached, and these involve a lot of rigorous testing and grading of the dies.  

The pizza principle 

Wafers are put through a series of tests, and these are carried out at a range of different temperatures. The idea is to grade each die within the wafer according to its performance level. As a die is tested, a record of the results is made in what is called a ‘wafer map’. This replicates the position of each die on the wafer, and indicates how it has performed against the grading system. 

This is where pizzas come in. If you’ve ever cooked your own pizzas you will know how hard it is to get a uniform result. The edges tend to cook more than the centre. A flash wafer has a similar property, in that the quality of die performance tends to be differently distributed at its edges and in its centre. The edges normally perform less well than the centre. 

So, armed with a wafer and a wafer map, a flash memory manufacturer now knows which are the dies they can turn into functioning flash memory chips. The wafer and the wafer map go off to the assembly house for that process to happen.  

Making the grade 

The level of inexactitude in the production of each wafer exists because there are too many variables involved in the production process to achieve a completely perfect die every time. There are many different raw materials used in the production of flash memory, and each of them can vary in quality.  

The manufacturing conditions can vary too, and with production taking place at the micro level it just isn’t possible to control every possible factor to the nth degree without incurring extortionate costs. As it is, flash memory fabrication plants cost billions of dollars to build – which is why there are so few of them.  

Still, every manufacturer wants the highest possible die yield from every wafer, and while they don’t want to send dies that might fail to manufacturing – which would be very harmful to their reputation – they do want to extract best value from every wafer.  

Bad blocks and less being more 

Some manufacturers work to make the most of the product of each wafer, and might sell the dies that fail to reach their standards to a broker where they can find a long and useful life in products like promotional USB drives, where they do not need to meet the same performance standards as those used in industrial products.  

In any case, all flash memory has small sections that don’t work as they should. These areas, known as ‘bad blocks’, can’t be used to store any data, and their existence is the reason that flash memory is managed by a ‘controller’. This effectively identifies bad blocks and does not allow any data to be saved to them. As more bad blocks appear during the lifespan of the product the controller can move data from a bad block to a good block, and mark the bad block as not to be used.  

Sometimes less is more with flash memory. Imagine you have a 128Gbit die that would normally make a 16GB product. This die is identified as being 90% good but with 10% bad blocks. It can’t be used for that 16GB product, but the manufacturer can still use it – as an 8GB product. The controller firmware can be set to treat the flash drive like an 8GB drive, but it will have a lot more than 8GB of storage potential. The extra memory can become a benefit, used by the controller to spread the load on all the drive’s memory in a process known as ‘wear levelling’ which can increase the life of the product.  

Finding the best industrial grade flash memory 

When a wafer manufacturer sells its wafers, the price is determined by the quality and yield of the die. There are no agreed international standards for grading flash memory, and it is difficult to compare the classifications used by different manufacturers. Yet finding the right grade of die, especially for demanding uses in sectors like industrial, health or automotive, is vital, and if a company such as our partners SanDisk or ATP are buying wafers to specifically build an industrial grade product, they will want to know the yield of the wafer from which the die they want to use came. 

Some manufacturers don’t provide this yield data. In the past, when the term ‘industrial grade’ was new for flash, some manufacturers took their existing consumer grade products and changed the marketing to industrial, but the product itself was not industrial grade. Customers then had problems because performance was inconsistent, failures were high and they encountered unexpected costs and reputational damage.  

It is good news that things have changed, but even today it can be difficult for those who aren’t steeped in flash memory experience to find the best industrial grade flash memory for their needs. At Cardwave we feel our customers’ pain, and always ask leading questions to understand the application of the end product and then ensure the flash products we select are graded correctly, tested and perform to spec. We always share customer application knowledge with our manufacturing partners and because products are matched to application we have a very low failure rate.  

  

Author: Paul Norbury

Date: Thursday 23rd July 2020