Article: AMD's Mobile Strategy
By: Seni (seniike.delete@this.hotmail.com), December 20, 2011 11:22 am
Room: Moderated Discussions
Paul A. Clayton (paaronclayton@gmail.com) on 12/20/11 wrote:
---------------------------
>(Side rant: I am disappointed that inexpensive,
>low-capacity, moderate performance solid state storage is
>not broadly available. The emphasis on capacity and
>performance--while perhaps appropriate initially--seems to
>be ignoring the scalability aspect of sold-state [having a
>relatively low fixed cost component compared to disk] and
>the power and form-factor [and to a lesser extent now,
>noise] advantages.)
You assume that there IS a scalability advantage. It turns out that there isn't, at least not for flash.
-A significant fraction of the cost and price of a flash chip is unrelated to its capacity. Even very low capacity chips don't go much below $3.
Here's some current prices from a market tracker to give you some idea:
2GB MLC: $2.92
4GB MLC: $3.22
8GB MLC: $5.84
16GB 2-chip stack of 8GB MLC: $17.36
-Individual flash chips are not fast, so you can't use just one. It takes four to eight of them just to match the sequential write speed of a conventional hard drive.
Four has been tried (OCZ Onyx for example) and the results were poor. Eight is currently the standard.
-Good 8-channel controller chips aren't cheap either.
About the cheapest you could do is a 32GB drive:
8x 4GB chips = 8 x $3.22 = $25.76
+ $10 controller chip + $10 pcb, case, power circuitry and whatnot
=$45.76
whereas, doubling the capacity to 64GB gives:
8x 8GB chips = 8 x $6 = $46.72
+ $10 controller chip + $10 pcb, case, power circuitry and whatnot
=$66.72
So double the capacity for only 46% more cost... and the double capacity one is also faster and has more spare area for wear leveling. There's really no sense in small capacity SSDs.
---------------------------
>(Side rant: I am disappointed that inexpensive,
>low-capacity, moderate performance solid state storage is
>not broadly available. The emphasis on capacity and
>performance--while perhaps appropriate initially--seems to
>be ignoring the scalability aspect of sold-state [having a
>relatively low fixed cost component compared to disk] and
>the power and form-factor [and to a lesser extent now,
>noise] advantages.)
You assume that there IS a scalability advantage. It turns out that there isn't, at least not for flash.
-A significant fraction of the cost and price of a flash chip is unrelated to its capacity. Even very low capacity chips don't go much below $3.
Here's some current prices from a market tracker to give you some idea:
2GB MLC: $2.92
4GB MLC: $3.22
8GB MLC: $5.84
16GB 2-chip stack of 8GB MLC: $17.36
-Individual flash chips are not fast, so you can't use just one. It takes four to eight of them just to match the sequential write speed of a conventional hard drive.
Four has been tried (OCZ Onyx for example) and the results were poor. Eight is currently the standard.
-Good 8-channel controller chips aren't cheap either.
About the cheapest you could do is a 32GB drive:
8x 4GB chips = 8 x $3.22 = $25.76
+ $10 controller chip + $10 pcb, case, power circuitry and whatnot
=$45.76
whereas, doubling the capacity to 64GB gives:
8x 8GB chips = 8 x $6 = $46.72
+ $10 controller chip + $10 pcb, case, power circuitry and whatnot
=$66.72
So double the capacity for only 46% more cost... and the double capacity one is also faster and has more spare area for wear leveling. There's really no sense in small capacity SSDs.
