[Discuss] SSD: enterprise vs. consumer Flash

[Richard Pieri]

On 6/3/2012 10:55 PM, Tom Metro wrote:
> So more silicon...

Less, actually.  Enterprise-class SSD vendors cheat.

Say that Toshiba makes a 128G MLC flash chip (which they do, in fact). 
This chip sold as part of a SandForce consumer device would have the 
controller set at 120G storage capacity.  The difference between binary 
and decimal calculations is used to over-provision the drive by the 
typical ~7.5% found on consumer SSDs.

An enterprise-class vendor like STEC would take that chip, put it on 
their own proprietary controller, and set the capacity at 64G.  Instant 
100% over-provisioning.  And then they would sell it for ten times the 
price of the SandForce unit because STEC's SSD would last five years 
where the SandForce unit would burn out in 6 months.

A vendor like Violin or EMC would take 16 or 32 of these 128G chips, set 
their capacity at 64G, and glom them together on a tray for a storage 
frame.  It isn't "more silicon" in the form of larger flash chips; it's 
"more silicon" in the form of more flash chips.

All with the same "consumer-grade" flash chips.

That's not to say that all of these chips are equal.  They're not. 
You'll get a better product from Toshiba than you will from Adata.

Of note: consumer SSD makers cheat the same way but they're generally 
more open about doing so.  It's easy to figure how much 
over-provisioning you have on a 400GB SSD when the controller has 
sixteen 32GB flash chips on it.

-- 
Rich P.