MOD5474 / MOD5484 status ?
-
Ridgeglider
- Posts: 513
- Joined: Sat Apr 26, 2008 7:14 am
MOD5474 / MOD5484 status ?
We have an application that needs more oomph than the existing NB modules. Last February Paul mentioned a pending release of 5474 or 5484-based module. Any updates on the status, price or availability on either this module, or on any others with more memory and speed?
Re: MOD5474 / MOD5484 status ?
The good news:
I have a 100% functional MOD5474 with 64M bytes of 32 bit wide SDRAM running at 133Mhz
16Mbytes flash, and a Cpu running at 266Mhz
It's 100% functional on my desk. The design is 100% done, all issues resolved, ready for a beta that will never come.
The really bad news:
Freescale so screwed up the SDRAM to CPU interface that the part is 100% useless.
It takes about 20 clock cycles to do any SDRAM access. So both the basic code execution performance
when running out of SDRAM and the Network performance are worse than ANY part currently in our product line.
The Ethernet controller uses a very complicated DMA engine that can only be used by using a Freescale provided library that is
agonizingly slow. The usable network performance is on par with the 40Mhz 5206e running 10 base T with an external mac-phy.
If you very carefully manage your data and instructions so both the speed critical data and code fit into the
caches, and then lock these cache lines, then the part screams. Otherwise its useless.
We are looking at a new 240Mhz V3 part with 128K of single cycle SRAM on board and a properly done DDR interface
that we expect to yield a 75 to 100% performance boost over the MOD5234.
We are in the process of porting to the freescale dev kit for that part before doing our own hardware design.
That way we can evaluate the hype without speding months on hardware design that turns out to be useless.
arggghhhhhhh!
Paul
I have a 100% functional MOD5474 with 64M bytes of 32 bit wide SDRAM running at 133Mhz
16Mbytes flash, and a Cpu running at 266Mhz
It's 100% functional on my desk. The design is 100% done, all issues resolved, ready for a beta that will never come.
The really bad news:
Freescale so screwed up the SDRAM to CPU interface that the part is 100% useless.
It takes about 20 clock cycles to do any SDRAM access. So both the basic code execution performance
when running out of SDRAM and the Network performance are worse than ANY part currently in our product line.
The Ethernet controller uses a very complicated DMA engine that can only be used by using a Freescale provided library that is
agonizingly slow. The usable network performance is on par with the 40Mhz 5206e running 10 base T with an external mac-phy.
If you very carefully manage your data and instructions so both the speed critical data and code fit into the
caches, and then lock these cache lines, then the part screams. Otherwise its useless.
We are looking at a new 240Mhz V3 part with 128K of single cycle SRAM on board and a properly done DDR interface
that we expect to yield a 75 to 100% performance boost over the MOD5234.
We are in the process of porting to the freescale dev kit for that part before doing our own hardware design.
That way we can evaluate the hype without speding months on hardware design that turns out to be useless.
arggghhhhhhh!
Paul
-
Ridgeglider
- Posts: 513
- Joined: Sat Apr 26, 2008 7:14 am
Re: MOD5474 / MOD5484 status ?
Thanks for the update Paul. Sorry to hear it was such a mess.