# Progress on Sheeva -- actually got a half-way functional flow-graph working

It can only handle two channels, and it takes 78% of the CPU while doing
it, but it does work!

Marcus L.
Principal Investigator

On 04/08/2010 10:30 PM, Marcus D. Leech wrote:

It can only handle two channels, and it takes 78% of the CPU while doing
it, but it does work!

How about calculating a \$ per sample number?

Philip

Hmmm, let’s see.

The Sheeva Plug was \$99.00
The USB Audio subsystem I had to buy for it was \$79.00

So, \$178.00 divided by 96000 samples/sec

That’s 0.2cents/sample/sec

Let’s see how that scales using a USRP1 and a “mildly pimped out” PC

USRP1 is \$700
PC is probably another \$700

So, \$1400.00 divided by 8megasamples/sec

That’s roughly 0.02cents/sample/sec

The USRP1 + PC is roughly an order-of-magnitude better value, when
looked at in that way.

Marcus L.
Principal Investigator

I did the same sort of thing comparing the 2 channel 1Gs/s roach boards
and the 4 channel 1.5Gs/s PX15500’s when every one was ranting about the
roach boards and how cheap they where. It worked out a persample dollar
of a 3rd less or so for the more expensive one.

What it really comes down to is, what do you really need, and will you
ever have any need of the capabilities. And more importantly do you have
the could hard cash to get the more “expensive one”

Bruce

From:
discuss-gnu[email protected]
[discuss-gnur[email protected]]
on behalf of Marcus D. Leech [[email protected]]
Sent: Friday, 9 April 2010 10:39 PM
To: Philip B.; GNURadio D.ion List
Subject: Re: [Discuss-gnuradio] Progress on Sheeva – actually got a
half-way functional flow-graph working

Hmmm, let’s see.

The Sheeva Plug was \$99.00
The USB Audio subsystem I had to buy for it was \$79.00

So, \$178.00 divided by 96000 samples/sec

That’s 0.2cents/sample/sec

Let’s see how that scales using a USRP1 and a “mildly pimped out” PC

USRP1 is \$700
PC is probably another \$700

So, \$1400.00 divided by 8megasamples/sec

That’s roughly 0.02cents/sample/sec

The USRP1 + PC is roughly an order-of-magnitude better value, when
looked at in that way.

Marcus L.
Principal Investigator