SRAM vs. SDRAM

[Guest]

Archived from groups: comp.sys.ibm.pc.hardware.chips (More info?)

If SRAM consists of 4-6 times transistors as SDRAM, how come it costs
gazillion times as much ?
Shouldn't it be viable to produce a 128MB stick for the same amount of money
it takes to produce 512MB DRAM ?

 

[Guest]

Archived from groups: comp.sys.ibm.pc.hardware.chips (More info?)

"flekso" <taurus@email.hinet.hr> wrote:

>If SRAM consists of 4-6 times transistors as SDRAM, how come it costs
>gazillion times as much ?
>Shouldn't it be viable to produce a 128MB stick for the same amount of money
>it takes to produce 512MB DRAM ?

The size of a chip and it's cost is not a linear relationship. Let's
say a wafer has 20 defects. If you can fit 100 chips on the wafer,
you'll yield 80 good chip. Now, if your smaller chips are 1/6 the
size, you could fit maybe 700 of them on the wafer (think about the
edges), yielding 680 good chips, 8.5 (not 6) times as many. Pardon my
gross guesstimates, but the point is valid.

 

[Guest]

Archived from groups: comp.sys.ibm.pc.hardware.chips (More info?)

On Tue, 06 Apr 2004 13:44:07 +0200, flekso wrote:

> If SRAM consists of 4-6 times transistors as SDRAM, how come it costs
> gazillion times as much ?
> Shouldn't it be viable to produce a 128MB stick for the same amount of
> money it takes to produce 512MB DRAM ?

Economics - ie supply and demand, economies of scale and all that stuff.

Cheers
Anton

 

[Dave]

Archived from groups: comp.sys.ibm.pc.hardware.chips (More info?)

On Wed, 07 Apr 2004 07:55:39 -0500, chrisv <chrisv@nospam.invalid>
wrote:

>The size of a chip and it's cost is not a linear relationship. Let's
>say a wafer has 20 defects. If you can fit 100 chips on the wafer,
>you'll yield 80 good chip. Now, if your smaller chips are 1/6 the
>size, you could fit maybe 700 of them on the wafer (think about the
>edges), yielding 680 good chips, 8.5 (not 6) times as many. Pardon my
>gross guesstimates, but the point is valid.

Surely SRAM has some redundancy to such defects? eg extra transistors.

 

[Guest]

Archived from groups: comp.sys.ibm.pc.hardware.chips (More info?)

Dave <look@my.sig> wrote:
> On Wed, 07 Apr 2004 07:55:39 -0500, chrisv <chrisv@nospam.invalid>
> wrote:
> >The size of a chip and it's cost is not a linear relationship. Let's
> >say a wafer has 20 defects. If you can fit 100 chips on the wafer,
> >you'll yield 80 good chip. Now, if your smaller chips are 1/6 the
> >size, you could fit maybe 700 of them on the wafer (think about the
> >edges), yielding 680 good chips, 8.5 (not 6) times as many. Pardon my
> >gross guesstimates, but the point is valid.
>
> Surely SRAM has some redundancy to such defects? eg extra transistors.

Not generally. An individual chip can't generally route around bad
transistors. With something as big and as expensive as CPUs, sometimes you
can route around a a whole region of a chip (ie, as is rumored to be the way
they pick some chips to be celerons -- if there's a defect in part of the
cache region.)... but I doubt that would be economical for a cheap bulk part
like RAM.

--
Nate Edel http://www.nkedel.com/

"Elder Party 2004: Cthulhu for President -- this time WE'RE the lesser
evil."