seanadams;183862 Wrote: 
> Actually I believe that was _entirely_ the matter at hand. If you refer
> to the post that kicked off this silly discussion, Havoc said "I still
> have to see the first flipped sram bit! This might be a problem for a
> satelite out there, but not for a pc."

Let me start by agreeing with Sean and others that errors from SRAM are
irrelevant to the discussion at hand.

That said, I know quite a bit about this subject and I'll take
advantage of this detour to provide a bit of factual information for
anyone interested.

First, SRAMs are 1000 times more susceptible to soft errors (SER)
compared to DRAMs, as pointed out by Pablo.  The reason for this is
that DRAMs have a much larger capacitor in each memory cell and it
takes much more induced charge to cause an error because of this.  This
then translates to a much lower error rate due to ionizing radiation. 
However, this isn't the reason one needs to use ECC for modern SRAMs.

Instead, the reason has to do with problems that occur as one scales
SRAM cells to smaller dimensions.  Statistical dopant fluctuations lead
to mismatches between the SRAM transistors and one when builds large
on-chip memories arrays one can't guarantee that all the bits will be
error free.  So designers need to start using ECC for the larger on
chip arrays.  This is standard practice for designers of processors
using large SRAM caches.

---Gary


-- 
GaryB
------------------------------------------------------------------------
GaryB's Profile: http://forums.slimdevices.com/member.php?userid=3169
View this thread: http://forums.slimdevices.com/showthread.php?t=32993

_______________________________________________
audiophiles mailing list
[email protected]
http://lists.slimdevices.com/lists/listinfo/audiophiles

Reply via email to