Nope, it is turned off.
They do not have all their air conditioning installed yet, so when it runs they have to turn off a bunch of other gear in the room.

To do testing we had them rent some big fan units to push air around the room. It's a BIG room!

OK, so I am busted by the "grammar police".

The unit takes about 89KW of power, and requires 25 tons of refrigeration.

Electricity cost for running it and related cooling is about $120 per hour.

 

Looks like a nice setup, beautifully installed. Too bad it's not IBM equipment. :D

 

$120 and hr for cooling? Yikes! but wait that's 8,280,000,000,000,000.00 point operations. It may be worth it!(note: Legal Disclaimer, the above number might be wrong, writer can not be accountable for errors). Thanks, the writer.

 

So, just wondering what something like that cost?

 

About 1.9 million.

 

IBM is nice, but then it would cost twice as much!

And I would not get a slice of it..

 

$120 for power AND cooling.

You got the right number of zeros.
8,280,000,000,000,000 floating point ops, and about half as many integer ops. At the same time.

Let's say you used this and a good algorithm to crack encryption keys. With a 256bit key it would take about 3 minutes to brute force solve it.

Los Alamos just got a similar style setup in the fall, but with 4 times as many CPUs. Around 8 or 9TF

We are installing a cluster at the U of Toronto this spring that is 385 nodes, 770 CPUs at about 50% higher clock rate than the one we just did.
About 3.5 tflops.

They figure that by 2008 the bigger installations will be breaking a petaflop. That is one quadrillion ops per second.

Kind of humbling, isn't it?

 

Maurice hasn't publicly revealed this, but every Stage I unit is equipped with an 802.11gggggg wireless interface that ties our ECUs into the Utah installation and all his others. When the switch is thrown we will be the world's first mobile distributed computing network that can pull .91 G's. Will be used to break the Mazda ECU authentication and encryption, lowering the price of future work.

 

....talk about thread creep; really we're all just giddy to when our units will arrive.

......must be decimal floating point error on Utah's 101th CPU; check with NASA. They've had a similar flash memory error.

BTW: I think I heard that the memory chips for some of the cameras on the rovers are only like 128MB or 256MB; present day cameras have past this level already. Technology is cool - and fast.

 

As you know, I was just giving you grief Maurice. :D

 

Actually, I think the PCM piggybacks do indeed have a wireless interface, all tied into one of Maurice's mainframes... from which....he will someday flip a switch..... and...
AND RULE THE WORLD!!!!!!!!HAHAHAHAHhahahahahahahhahahahahahahahahahaha

We're on to you, Maurice. :D

 

Maurice, I thought you said that you had all this scaled down so it can fit into the stock ECU box?

 

Uh, yeah.
You are pullingmy leg right?

OK, here goes:
The unit weighs about 1.3 tons and comes with a small trailer which you pull behind the car.

Most of the contents of the box are for show. The only active component is a little chip about the size of your small fingernail.

It is powered by a small experimental fusion reactor, and the control codes actually come by wireless communications from our headquarters in a cave deep in the Rocky mountains.

If someone is good, then we reprogram your car to make lots of power.

If you are bad we turn it all off.

Say bad things and troll on forums?

No HP for you!

 

Maurice.....The Horse Power ****!

NO HORSE POWER FOR YOU!

S!

 

Yes it is, what's next, the speed of thought? Wait a minute we can't possibly think that fast!

 

Oki, 3 minutes for a 256bit key? Let's do the maths.

Normally you solve a 256bit key in 2^255 tries.
Assuming your computer does 10.000.000.000.000.000 keys per second (that is if you can try 1 key per 1 CPU operation, again not very likely as it is going to take a LOT more operations do check 1 key)

But again, lets assume you can check 1 key for every operation, that leads us to about 1.8 * 10^53 years crack one message.

Assuming your computer cluster is 1 million times faster than what you specified then it is still going to take 1.8 * 10^47 years to solve it. To put things into perspective; the universe is belived to be 11.5 * 10^9 years old.

/twospoons, just putting things into perspective

 

The brain is still much quicker than any computer out there, but the gap is closing quickly.

/twospoons

 

i seem to remember a time when you would have almost had to turn the switch off on yourself, Maurice :p :D i for one am glad that cooler heads prevailed. now to just find $600 (or is it still $750)


you missed where he said " this and a good algorithm" .
:p

 

It is now officially $600 for one unit, and $500 for 10 and up.
Go ahead, do a group buy, make my day and save me paperwork!
Shipping is lots cheaper per unit too!

Over in the poll thread I became convinced that is the correct path.

Thanks!