Large flash chips are obtainable?

[ali_asadzadeh]

Hi,
I have some cool Ideas about making SSD's, I was searching and wonder if we can buy large flash chips from manufacturers,
I have Ideas about making a 1TB SSD, so I need 64 chips of 16GB each (Serial interface like QSPI or Hyper flash), what are our options? do you have any idea?

[NiHaoMike]

The only source I can think of that would be cheaper than just buying a new SSD would be to buy "bricked" SSDs at scrap prices and salvaging the chips from them. (The early ones that were prone to bricking almost always did so due to firmware bugs.)

[ali_asadzadeh]

So where do the SSD manufacturers buy the chips? any know brand?

[NiHaoMike]

Some get deals buying in bulk, some manufacture their own chips.

[Maxoverdrive]

some chips are available only per quantity as 5000 units batches or more  , could be available from chips brokers in China , HK and other countries

[ali_asadzadeh]

Do we have part numbers?

[Berni]

You can buy flash memory chips at pretty much any electronics component supplier like DigiKey:
https://www.digikey.com/product-detail/en/micron-technology-inc/MT29F128G08CFABBWP-12IT-B-TR/557-1833-1-ND/9673814

Just don't expect to get the dirt cheep prices that the chinese flash memory fabs pump out directly to manufacturers.

The cheapest way to get these large capacity flash chips is probably to buy some cheep SSDs, take them apart and desolder the chips off them. At that point if your goal was to make a SSD... well you could have just used the SSD you just bought rather than cannibalizing it for its memory chips

[ali_asadzadeh]

I want to know how to enter the bussines

[Berni]

I want to know how to enter the bussines

In that case you raise a few million dollars of investment money and get into negotiations with flash memory fab owners to buy the bad flash memory dies in quantities of >100k. Use the rest of the money to set up a chip packaging production line that is being fed by an automated chip testing array that maps out the dead memory locations to not be used.

Or if you secure an investment in the 100s of milions then make your own semiconductor fab and make your own flash silicon dies, all while keeping costs as low as possible to hopefully still turn a tiny bit of profit by the end.

[NiHaoMike]

Rather than try to compete in the mass market with thin margins, you might want to look at niche markets that will pay (a lot) more per GB because they needed some feature common SSDs do not have. That could be some sort of interface other than SATA or PCIe or special firmware features. One example would be a SSD with a write protect switch, another would be a keychain SSD with built in Wifi and battery for iPhones and other portable computing devices that do not accept micro SD cards.

[ali_asadzadeh]

why do you think they made a small profit!?
https://en.wikipedia.org/wiki/Micron_Technology

This one is poor too
https://en.wikipedia.org/wiki/SK_Hynix

we should help them, please buy from them or they would go bankrupt

Their small  and tiny Operating income was about US$5.87 billion (2017) and US$18.34 billion (2018), we should probably help them be able to keep their business

[NiHaoMike]

They manufacture their own chips, a particularly difficult task with cutting edge technology. Also note that they also make general purpose DRAM, which is a lot more expensive per GB, and graphics DRAM, which is even more expensive. Those really help to pay back the huge investments of constantly upgrading fabs to stay competitive. The margins on Flash are thin, but they are able to more than make up for it with very high production volume.

[ejeffrey]

Micron and sk hynix both manufacture memory chips.  It's not a super profitable business but thay is different than the companies that buy the chips and controllers and package them into off brand SSDs.

[Berni]

why do you think they made a small profit!?
https://en.wikipedia.org/wiki/Micron_Technology

Only reason they turn a profit is by making massive volumes since the profit margins on flash are so thin. There is no way someone could just spin up a new flash memory fab out of nothing and turn a profit in small volumes. Quite a few even dropped out of flash memory manufacturing because it just was not worth it. Then the flash memory price crash happened recently where they groseley over estimated the demand for flash due to people not buying as many SSDs as they thought. But since the price of flash memory dropped like a brick so did prices of SSDs and so everyone did indeed start buying them and so the prices have stabilized again once this excess production volume was used up.

But at least flash has a bright future ahed of it. Hard drive manufacturers instead are feeling the pressure as flash is gradually closing the gap in price per GB.

[ejeffrey]

Back to your original post:

I have some cool Ideas about making SSD's, I was searching and wonder if we can buy large flash chips from manufacturers,

If you have some cool ideas then by all means: start off buying expensive chips from digikey/mouser to make your prototypes.  If you have a product you want to manufacture in moderate volumes then you just start calling up bigger distributors.  You won't get the prices that the big players get when they have high volume ongoing contracts, but you should be able to do better than digikey.  If you find a niche market you might be able to sell some, although turning a profit or paying back reasonable development costs is going to be difficult in such an established and competitive market even if you come up with cool distinguishing features.

have Ideas about making a 1TB SSD, so I need 64 chips of 16GB each (Serial interface like QSPI or Hyper flash), what are our options? do you have any idea?

AFAIK, most high capacity flash chips use parallel interfaces.  Certainly that is what is used in consumer SSDs.  I don't know if you can find QSPI or other "serial" flash chips in such large capacity.  Why is that important for your application?

[Bud]

It is all can be done until time comes to sell it. How you are going to advertise, sell, ship, support, handle returns, etc. Is it going to be out of your garage? Even Apple had to provide customer support to the (only) 200 customers who bought their Apple 1 gizmo.

[coppice]

It shouldn't be hard to get 16GB QSPI chips in modest quantities. They get embedded in many things, so there is a market for them in both small and large quantities. Those aren't the kind of chips that get used in SSDs, though. Chips that will support the kind of high performance people look for in an SSD are only sold in large volumes with direct vendor support, and directly interface to an SSD controller chip. This is because the only market for them is the various forms of SSD, and those are only made in high volumes.

[CChin254]

Some small SSDs use 16 GB NAND Flash, such as the OCZ RevoDrive 350 SSD 480 GB. That SSD uses 32 16GB NAND Flash chips (Toshiba #TH58TEG7DDJBA4C).

[CChin254]

You could by large (128GB+) NAND Flash chips made by Micron, but the datasheets are not publicly available. You would probably have to contact the manufacturer for them which then comes probably under a NDA and if you plan buy a lot.

[ali_asadzadeh]

AFAIK, most high capacity flash chips use parallel interfaces.  Certainly that is what is used in consumer SSDs.  I don't know if you can find QSPI or other "serial" flash chips in such large capacity.  Why is that important for your application?

My Idea needs a serial Flash chip the problem is that I could not find large ones, even with Big prices to pay.

It shouldn't be hard to get 16GB QSPI chips in modest quantities. They get embedded in many things, so there is a market for them in both small and large quantities. Those aren't the kind of chips that get used in SSDs, though. Chips that will support the kind of high performance people look for in an SSD are only sold in large volumes with direct vendor support, and directly interface to an SSD controller chip. This is because the only market for them is the various forms of SSD, and those are only made in high volumes.

Do you have part numbers?

[TomS_]

My Idea needs a serial Flash chip the problem is that I could not find large ones, even with Big prices to pay.

Implement the required interface in an FPGA?

[mzzj]

why do you think they made a small profit!?
https://en.wikipedia.org/wiki/Micron_Technology

This one is poor too
https://en.wikipedia.org/wiki/SK_Hynix

we should help them, please buy from them or they would go bankrupt

Their small  and tiny Operating income was about US$5.87 billion (2017) and US$18.34 billion (2018), we should probably help them be able to keep their business

Can't be that hard to compete with them. building a 300 mm fab is maybe 5 to 10 billion US$ investment, all you need couple of coins more. 

[Alex Eisenhut]

why do you think they made a small profit!?
https://en.wikipedia.org/wiki/Micron_Technology

This one is poor too
https://en.wikipedia.org/wiki/SK_Hynix

we should help them, please buy from them or they would go bankrupt

Their small  and tiny Operating income was about US$5.87 billion (2017) and US$18.34 billion (2018), we should probably help them be able to keep their business

Can't be that hard to compete with them. building a 300 mm fab is maybe 5 to 10 billion US$ investment, all you need couple of coins more. 

Let me look under my couch, there's always a few coins there. I also haven't returned deposit bottles in a while.

[free_electron]

and how do you want to control those ? you are going to need a flash controller that can do the command queueing and wear leveling. That's not chicken shit to design ! and then there is the cache required.

[coppice]

and how do you want to control those ? you are going to need a flash controller that can do the command queueing and wear leveling. That's not chicken shit to design ! and then there is the cache required.

So, how are you going to implement things like wear levelling without a custom controller chip to do it for you? That is going to involve a LOT more work. Remember that most modern high density flash can't hold data unless its powered up from time to time to allow that complex controller chip to do its work.