[Hardware] Notes on packing serial cores into Xilinx devices

david fleischer cilantro_il at yahoo.com
Fri Aug 13 14:57:02 EDT 2004

The reward prize is not the only benefit...
I think the reward is getting to complete the project.
When adding a new method to search through the keys
you're contributing significantly to a project that as
it stands right now is non-tractable.
You can also see the progress in near real time, when
you potentially insert your hardware into the project
you can see the inflection point in the graphs.
And if this were not enough, when going to the 80-bit
key you will definitely need the contribution of a
large team, unless you want to rent a whole building
to fill-up with FPGA boards.
As for the chips, the Virtex get expensive fast (even
above 1000$). I think we should use the smallest that
can implement the architecture. If 50Kgates is not
enough, then 100K.
At the high-prices it will be a proof of concept only,
because actually the pentium and ppc chips are doing a
really good job.
This is what convinces me that the bit-serial
architecture is not so hot.


--- Dan Oetting <dan_oetting at uswest.net> wrote:
> An uncoordinated search is not as efficient as a
> well coordinated 
> search. The efficiency in terms of the average time
> to find the key is 
> something like 1/e due to overlaps. So on average it
> would take 2.7 
> times as long to search and there is no guaranteed
> maximum time. But on 
> the plus side, the winner gets the full $10,000.00,
> there is no 
> advantage for anyone to run a bogus client, and
> there is no overhead 
> managing the key space.
> The efficiencies of the coordinated search don't
> kick in until a 
> significant fraction of the key space has been
> searched. So users could 
> jump in today and start searching random blocks with
> very little chance 
> of overlap. As long as the block structure of the
> key space is defined 
> early, the random blocks searched can be merged into
> the coordinated 
> search later.
> Assuming that the cash reward is the only reason for
> participating and 
> the d.net payoff schedual remains the same and not
> counting the initial 
> cost of hardware and operating costs, at what
> percentage of the 
> keyspace searched does it become advantageous to
> stop random searching 
> and join the coordinated search? [PS: this is just a
> math problem and 
> not intended to reflect reality]
> > I read the pointer to the Virtex device; in order
> for
> > it to be worth-while, I think it ideally needs to
> cost
> > about20$. This is about the cost of the Spartan
> chips?
> Using some reasonable value for cracking rate and
> operating cost for 
> some piece of hardware, what is the break-even
> acquisition cost for 
> that hardware. [PS: this is trick question, what is
> the trick?]
> _______________________________________________
> Hardware mailing list
> Hardware at lists.distributed.net

More information about the Hardware mailing list