Re: News on F21
On Mon, 10 Mar 1997, Jeff Fox wrote:
> Penio asked:
>
> >Not to mention, that the F21 die is 3/4 empty and I bet that Chuck is
> >working on putting memory there, which will raise the throughput at least
> >twice -- to the true 400 MIPs. Jeff, is that right?
>
> Yes. On chip RAM and/or ROM will reduce or remove the memory
> bottleneck. We are currently at about 500mips internally but SRAM will
> limit you to about 200 and DRAM closer to 100 at least with external
> memory and the current memory setup time on these chips.
> As for price volume still remains the main issue.
> If I get 10k made the manufacture cost drops to $15. This means if that
> many are made the price will be low compared to $100 chips, but not
> compared to $2 chips. If millions are manufactured then the price starts
> to look nice. In very large quantity they _could_ be in the $1-$2 price
> range, but don't hold out for it.
When the price is in the $10 range, is is essential that this be the final
price in a multiprocessor system, i.e., it contains enough RAM inside to
fit a small kernel, some networking code, some application code, and some
data, that is gotten in and out via the network and/or parallel port.
If memory serves me right, F21 is 3/4 empty and is about 15K transistors.
that means that another 45K are potentially sitting there unused. How many
Ks of SRAM does that make? If Chuck's OKAD is about 8K, I believe that
almost anything useful can be that big and fit on chip.
Is there a chance that F21c will include experimantal on-chip RAM, albeit
in small amounts?
In other words, where is Chuck's bottleneck -- design, testing, writing
software, developing OKAD, or just the lack of $10/fab run?
--
Penio Penev <Penev@pisa.Rockefeller.edu> 1-212-327-7423