Yokemate of Keyboards
Posts: 2720 from 2003/2/24
@minator
As far as I know, some problems is only/can only be spotted *after* you have it in real silicon, and AFAIK it's not very uncommon that you have to taping out a few revisions before reaching a level that can be judged good enough for end-users. For example: Freescale did at least 3 tapeouts of the i.MX515, and Genesi used the TO2 in early board designs/targeted at early developers. Only the TO3 was used in the final, end-user products:
Quote:
There exist three revisions of the Smarttop, using different Tape-Out (TO) revisions of the i.MX515;
l Revision 1.0 - a limited release developer edition, i.MX515 TO2. 4GB SSD. Internal MicroSD. Board Only.
l Revision 1.1 - Project and limited sale. Utilizing i.MX515 TO2. 4GB SSD. Internal MicroSD. Black case.
l Revision 1.2 - Production version. Utilizing i.MX515 TO3. 8GB SSD. Black case.
In general Genesi refers to the 1.1 board as TO2 and the 1.2 board as TO3. The revision 1.0 board is very rare and Genesi offers only limited support for this model.
The i.MX515 TO2 has a known errata which causes the Cortex-A8 NEON vector unit to misbehave, and as such the Linux kernel will not report it as supporting NEON code, in order to preserve system stability. This errata was fixed in TO3.
This is what I meant with "being rushed"; maybe the chip hasn't reached its final tapeout revision yet, but Samsung/Google is eager to push a first version of their Cortex-A15 smarttop to the market? Some things can be worked around in drivers, other things can't and will need an updated chip. I guess we'll see.
MorphOS is Amiga
done right! MorphOS NG will be AROS
done right! 
