This chip is also supported by the stock ncr53c8xx driver 3.1b/c/d, but
only the 896 driver is able to handle all the features of the 896 chip.
(At least on 32 bit systems for the moment).
This has been possible thanks to Symbios who recently provided me with a
SYM21002 board.
The main new features of the 896 chip against the 895 are the following
ones:
- 64 bit PCI
- 8 Kb on-chip RAM
- Hardware phase mismatch handling (need some SCRIPTS to save the context)
- 2 functions -> 2 Wide LVD/SE SCSI controllers on a single chip.
The hardware phase mismatch handling is, in my opinion, a great
enhancement, since it allows:
- a single interrupt (at most) per IO.
- the SCRIPTS processor never has to wait for the C code, allowing full
parallelism between the host processor and the controller cores.
- lower PCI traffic by decreasing IO register accesses from the C code.
For the moment, this chip is not yet in full production process, so it
may be not easy to purchase controllers using this chip. The full
production will happen soon, probably at the beginning of 1999.
The 896 is extensively used by Intel for some of their server products.
Symbios supplies 2 products based on this board:
SYM22910:
---------
64 bit PCI bus (usable on a 32 bit PCI bus).
2 internal and 2 external Wide SCSI connectors.
This board is normally available now.
SYM21002:
---------
32 bit PCI bus (usable with 64 bit systems (PCI DAC cycles)).
2 internal Wide connectors, 1 internal and 1 external narrow connector.
This board will normally be available in quantities in the first
quarter of 1999.
As we know, the chip is not the more costly part of a SCSI controller
board. So, we can expect the SYM21002 to be significantly cheaper than
the SYM22910.
My testings using the 896 chip show that we need a high disk IO load in
order to gain advantage of the hardware phase mismatch. So, the 895
controller is still a excellent candidate for high end SCSI systems on 32
bit machines.
Regards,
Gerard.
PS: Since the support of the new features of the chip by the linux
896 driver are still in the experimental status, you need to enter
'ncr53c8xx=specf:y'
as boot options in order to tell the driver to enable them.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/