SETI

A perfect example of a DSP application, again based on FFTs, a Cell will boost my SETI@home [SETI] score no end! As mentioned elsewhere I estimate a single Cell will complete unit in under 5 minutes [SETI Calc]. Numerous other distributed applications will also benefit from the Cell.


http://www.blachford.info/computer/Cells/Cell1.html

---

<img src="http://boinc.mundayweb.com/seti2/stats.php?userID=1103">

I see you've read Nicholas Blachford's article on the new Cell processors.
It will be interesting to see if the final result matches up to it's estimated potential.

---

Grant
Darwin NT

Hm...

Even if the expected wonder-CPU could do so, what about the needed RAM bandwidth to feed it with the required data ?

IMHO, a System wanting to complete a WU in 5 Minutes would either require ~16MB Cache, or a RAM bandwidth in excess of 75GB/sec.

Anyway, sounds like an interesting CPU to come.

---

> IMHO, a System wanting to complete a WU in 5 Minutes would either require
> ~16MB Cache, or a RAM bandwidth in excess of 75GB/sec.

hmm, it has 36Mb of cache ram :)

---

<img src="http://boinc.mundayweb.com/seti2/stats.php?userID=1103">

Aren't PS3s due out Jan. '06?

---

>
> > IMHO, a System wanting to complete a WU in 5 Minutes would either
> require
> > ~16MB Cache, or a RAM bandwidth in excess of 75GB/sec.
>
> hmm, it has 36Mb of cache ram :)
>

Since the huge seti FFTs (1MB) won't fit into the 128KB local RAM of the cell's PU, it won't run at full speed with seti.

On the other side, there are 8 independent main memory channels, one for each PU.

I guess each PU will run as fast as a decent x86 PC, adding up to a total computing power similar to 32 PCs for a PS3 :o)


Regards Hans

P.S:

This still adds up to 5 - 10 minutes per WU!

---

> >
> > > IMHO, a System wanting to complete a WU in 5 Minutes would either
> > require
> > > ~16MB Cache, or a RAM bandwidth in excess of 75GB/sec.
> >
> > hmm, it has 36Mb of cache ram :)
> >
>
> Since the huge seti FFTs (1MB) won't fit into the 128KB local RAM of the
> cell's PU, it won't run at full speed with seti.
>
> On the other side, there are 8 independent main memory channels, one for each
> PU.
>
> I guess each PU will run as fast as a decent x86 PC, adding up to a total
> computing power similar to 32 PCs for a PS3 :o)
>
>
> Regards Hans
>
> P.S:
>
> This still adds up to 5 - 10 minutes per WU!


So fast in fact, that you'd NEED to sign up for multiple projects and run them all on the PS3. You'd run into the 100 WU per day limit within a few hours if you only ran Seti.

---

You will be assimilated...bunghole!

Is this maybe the first precursor to artificial intelligence running amok throughout the world ;)

Food for thought.

---

<img border="0" src="http://boinc.mundayweb.com/one/stats.php?userID=268&amp;prj=1&amp;trans=off" /><img border="0" src="http://boinc.mundayweb.com/one/stats.php?userID=268&amp;prj=4&amp;trans=off" />

> Is this maybe the first precursor to artificial intelligence running amok
> throughout the world ;)
>
> Food for thought.
>
>
>

If Artificial Intelligence is as selective as Natural Intelligence, then it'll probably self-destruct anyways before it poses any real threat to us. :)

---

You will be assimilated...bunghole!

>
> So fast in fact, that you'd NEED to sign up for multiple projects and run them
> all on the PS3. You'd run into the 100 WU per day limit within a few hours if
> you only ran Seti.
>
>

A way to make Climateprediction take less than a couple weeks for a WU? That is truely impressive.

Kinnison- It treads awfully close to the "trapper keeper" episode of South Park. The threatening type of AI would require that sort of power.

---

Still looking for something profound or inspirational to place here.

> Is this maybe the first precursor to artificial intelligence running amok
> throughout the world ;)
>
> Food for thought.
>
>

OK, time for some math :o)

A neuron works (very simplified) by doing weighted sum of it's synaptic input (1 multiply, 1 addition per synapse).

An average neuron in the human brain has 10000 synapses, and it updates it's "state" roughly 100 times per second.

This adds up to 2 million flops per second.

The number of neurons in the human brain is estimated to be around 100 billions.

So the human brain is capable of 200 Million GFlops :o)

Let's compare this to the PS3: Absolute maximum for 4 cells would be 1000 GFlops.

Result: You need 200,000 PS3's to emulate a human brain.
That's way less than the number of PS3's that will be sold.

Creepy :o)

Regards Hans

---

Hans,

30 million PS2's have been sold worldwide (to May 2004). I'm sure that Sony is expecting the same kind of sales for PS3 if they are putting so many billions of dollars into the cell. Just hope that the security on things is good, imagine what you can do with all 30 million PS3's linked via the internet - 100,000 work units per second (apart from the huge delay in internet transfers and other tech stuff that I know nothing about).

Keep up the interesting posts Hans.
Cheers

http://www.psreporter.com/playstation_2_sales.html