No Subject

[jfox@xxxxxxxxxx (Jeff Fox)]

>
>I din't see the development system to include SRAM. I know, that 
>execution speed off SRAM and DRAM is comparable due to long memory setup 
>time of the processor itself. SRAM would help a lot, though, to reduce 
>the number of off-page memory accesses, forced by the video generation. 
>So, to get the full speed, one has either disable video generation and 
>refresh (while guaranteeng, that the data access pattern of the 
>computation will refresh the meory,) or execute the time-critical code 
>off SRAM.
>
>Does the system include SRAM?
>If not, does it allow for the user to add SRAM easily?
>
>--
>Penio Penev <Penev@venezia.Rockefeller.edu> 1-212-327-7423

No. Neither of the systems boards that Offete offers have sockets for
3 8kx8 15ns srams.  Dr. Ting did create a jumper to run some out of
the DRAM socket for testing.

You are quite right about performance.  Even though the video
processor uses only 25% of the memory bandwidth directly there is a
parasitic effect due to off page memory access.  When both the cpu
and the video processor are running on different pages of DRAM then
each video processor memory access will be offpage and will force the
next cpu memory access to be offpage too!  Thus under these conditions
the cpu only gets 1/3 of the bandwidth, not 3/4.

You are correct that the cpu can run several times faster if there
is no video, or if it is running in the high speed sram.  

The 40ns for SRAM vs 50ns for DRAM does not give the whole picture.
An extra 200ns for two offpage memory access caused by the video
and cpu both in DRAM makes a big difference.

The new board is only two layer.  The first was 4 layer.  I don't
know if 2 layer will be ok for full speed sram operation.  I would
like to see a board with 3 8kx8 15ns sram sockets too.

Jeff Fox
11/16/94